RED-9010: remove redaction log

Red 9974: improce headline classification, fix font size calculation

See merge request fforesight/layout-parser!226

.gitattributes

@@ -0,0 +1 @@
+*.pdf filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -38,3 +38,9 @@ build/
 
 ### VS Code ###
 .vscode/
+gradlew.bat
+gradlew
+gradle.properties
+gradle/
+.DS_Store
+.DS_Store/

.gitlab-ci.yml

@@ -1,3 +1,7 @@
+variables:
+  #  SONAR_PROJECT_KEY: 'fforesight_layout-parser_AYd5quv2mRkBOCG22hvF'
+  GIT_SUBMODULE_STRATEGY: recursive
+  GIT_SUBMODULE_FORCE_HTTPS: 'true'
 include:
   - project: 'gitlab/gitlab'
     ref: 'main'
@@ -10,7 +14,7 @@ deploy:
   script:
     - echo "Building with gradle version ${BUILDVERSION}"
     - gradle -Pversion=${BUILDVERSION} publish
-    - gradle bootBuildImage --cleanCache --publishImage -PbuildbootDockerHostNetwork=true -Pversion=${BUILDVERSION}
+    - gradle bootBuildImage --publishImage -PbuildbootDockerHostNetwork=true -Pversion=${BUILDVERSION}
     - echo "BUILDVERSION=$BUILDVERSION" >> version.env
   artifacts:
     reports:
@@ -18,4 +22,4 @@ deploy:
   rules:
     - if: $CI_COMMIT_BRANCH == $CI_DEFAULT_BRANCH 
     - if: $CI_COMMIT_BRANCH =~ /^release/
-    - if: $CI_COMMIT_TAG
+    - if: $CI_COMMIT_TAG

.gitmodules

@@ -0,0 +1,8 @@
+[submodule "layoutparser-service/layoutparser-service-server/src/test/resources/files/basf"]
+	path = layoutparser-service/layoutparser-service-server/src/test/resources/files/basf
+	url = ssh://git@git.knecon.com:22222/fforesight/documents/basf.git
+	update = merge
+[submodule "layoutparser-service/layoutparser-service-server/src/test/resources/files/syngenta"]
+	path = layoutparser-service/layoutparser-service-server/src/test/resources/files/syngenta
+	url = ssh://git@git.knecon.com:22222/fforesight/documents/syngenta.git
+	update = merge

README.md

@@ -1 +1,89 @@
+# PDF Layout Parser Micro-Service: layout-parser  
 
+## Introduction
+The layout-parser micro-service is a powerful tool designed to efficiently extract structured information from PDF documents. Written in Java and utilizing Spring Boot 3, Apache PDFBox, and RabbitMQ, this micro-service excels at parsing PDFs and organizing their content into a meaningful and coherent layout structure. Notably, the layout-parser micro-service distinguishes itself by relying solely on advanced algorithms, rather than machine learning techniques.
+
+### Key Steps in the PDF Layout Parsing Process:
+
+* **Text Position Extraction:**  
+The micro-service leverages Apache PDFBox to extract precise text positions for each individual character within the PDF document.
+
+* **Word Segmentation and Text Block Formation:**  
+Employing an array of diverse algorithms, the micro-service initially identifies and segments words, creating distinct text blocks.
+
+* **Text Block Classification:**  
+The segmented text blocks are then subjected to classification algorithms. These algorithms categorize the text blocks based on their content and visual properties, distinguishing between sections, subsections, headlines, paragraphs, images, tables, table cells, headers, and footers.
+
+* **Layout Coherence Establishment:**  
+The classified text blocks are subsequently orchestrated into a cohesive layout structure. This process involves arranging sections, subsections, paragraphs, images, and other elements in a logical and structured manner.
+
+* **Output Generation in Various Formats:**  
+Once the layout structure is established, the micro-service generates output in multiple formats. These formats are designed for seamless integration with downstream micro-services. The supported formats include JSON, XML, and others, ensuring flexibility in downstream data consumption.
+
+### Optional Enhancements:
+
+* **ML-Based Table Extraction:**  
+For enhanced results, users have the option to incorporate machine learning-based table extraction. This feature can be activated by providing ML-generated results as a JSON file, which are then integrated seamlessly into the layout structure.
+
+* **Image Classification using ML:**  
+Additionally, for more accurate image classification, users can optionally feed ML-generated image classification results into the micro-service. Similar to the table extraction option, the micro-service processes the pre-parsed results in JSON format, thus optimizing the accuracy of image content identification.
+
+In conclusion, the layout-parser micro-service is a versatile PDF layout parsing solution crafted entirely around advanced algorithms, without reliance on machine learning. It proficiently extracts text positions, segments content into meaningful blocks, classifies these blocks, arranges them coherently, and outputs structured data for downstream micro-services. Optional integration with ML-generated table extractions and image classifications further enhances its capabilities.
+
+
+
+
+
+
+## Installation
+
+### Prerequisites
+
+Before building and using the layout-parser micro-service, please ensure you have the following software and tools installed:
+
+Java Development Kit (JDK) 17 or later
+Gradle build tool (preinstalled)
+Build and Test
+To build and test the micro-service, follow these steps:
+
+### Clone the Repository:
+
+bash
+```
+git clone ssh://git@git.knecon.com:22222/fforesight/layout-parser.git
+cd layout-parser
+```
+### Build the Project:
+Use the following command to build the project using Gradle:
+
+```
+gradle clean build
+```
+### Run Tests:
+Run the test suite using the following command:
+```
+gradle test
+```
+## Building a Custom Docker Image
+To create a custom Docker image for the layout-parser micro-service, execute the provided script:
+
+### Ensure Docker is Installed:
+Ensure that Docker is installed and running on your system.
+
+### Run the Image Building Script:
+Execute the publish-custom-image script in the project directory:
+
+```
+./publish-custom-image
+```
+## Publishing to Internal Maven Repository
+To publish the layout-parser micro-service to your internal Maven repository, execute the following command:
+
+```
+gradle -Pversion=buildVersion publish
+```
+Replace buildVersion with the desired version number.
+
+## Additional Notes
+Make sure to configure any necessary application properties before deploying the micro-service.
+For advanced usage and configurations, refer to Kilian or Dom or preferably the source code.

buildSrc/src/main/kotlin/com.knecon.fforesight.java-conventions.gradle.kts

@@ -24,6 +24,8 @@ tasks.named<Test>("test") {
     reports {
         junitXml.outputLocation.set(layout.buildDirectory.dir("reports/junit"))
     }
+    minHeapSize = "512m"
+    maxHeapSize = "2048m"
 }
 
 tasks.test {
@@ -40,6 +42,19 @@ tasks.jacocoTestReport {
 }
 
 allprojects {
+
+    tasks.withType<Javadoc> {
+        options {
+            this as StandardJavadocDocletOptions
+            addBooleanOption("Xdoclint:none", true)
+            addStringOption("Xmaxwarns", "1")
+        }
+    }
+
+    pmd {
+        setConsoleOutput(true)
+    }
+
     publishing {
         publications {
             create<MavenPublication>(name) {
@@ -62,6 +77,7 @@ java {
     withJavadocJar()
 }
 
+
 repositories {
     mavenLocal()
     mavenCentral()

config/pmd/pmd.xml

@@ -9,12 +9,13 @@
     </description>
 
     <rule ref="category/java/errorprone.xml">
-        <exclude name="DataflowAnomalyAnalysis"/>
         <exclude name="MissingSerialVersionUID"/>
-        <exclude name="NullAssignment"/>
         <exclude name="AvoidLiteralsInIfCondition"/>
         <exclude name="AvoidDuplicateLiterals"/>
-        <exclude name="AvoidFieldNameMatchingMethodName"/>
+        <exclude name="NullAssignment"/>
+        <exclude name="AssignmentInOperand"/>
+        <exclude name="BeanMembersShouldSerialize"/>
     </rule>
 
-</ruleset>
+</ruleset>
+

config/pmd/test_pmd.xml

@@ -10,14 +10,14 @@
 
 
     <rule ref="category/java/errorprone.xml">
-        <exclude name="DataflowAnomalyAnalysis"/>
         <exclude name="MissingSerialVersionUID"/>
-        <exclude name="NullAssignment"/>
         <exclude name="AvoidLiteralsInIfCondition"/>
         <exclude name="AvoidDuplicateLiterals"/>
         <exclude name="AvoidFieldNameMatchingMethodName"/>
-        <exclude name="AvoidFieldNameMatchingTypeName"/>
+        <exclude name="NullAssignment"/>
+        <exclude name="AssignmentInOperand"/>
         <exclude name="TestClassWithoutTestCases"/>
+        <exclude name="BeanMembersShouldSerialize"/>
     </rule>
 
-</ruleset>
+</ruleset>

layoutparser-service/layoutparser-service-internal-api/build.gradle.kts

@@ -1,6 +1,10 @@
 plugins {
     id("com.knecon.fforesight.java-conventions")
-    id("io.freefair.lombok") version "8.2.2"
+    id("io.freefair.lombok") version "8.4"
 }
 
 description = "layoutparser-service-internal-api"
+
+dependencies {
+    implementation("io.swagger.core.v3:swagger-annotations:2.2.15")
+}

layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/redaction/DocumentData.java

@@ -1,5 +1,8 @@
 package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
 
+import java.io.Serializable;
+
+import io.swagger.v3.oas.annotations.media.Schema;
 import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
 import lombok.Builder;
@@ -10,11 +13,16 @@ import lombok.experimental.FieldDefaults;
 @Builder
 @AllArgsConstructor
 @FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
-public class DocumentData {
+@Schema(description = "Object containing the complete document layout parsing information. It is split into 4 categories, structure, text, positions and pages: " + "The document tree structure of SemanticNodes such as Section, Paragraph, Headline, etc. " + "The text, which is stored as separate blocks of data. " + "The text positions, which are also stored as separate blocks. The Blocks are equal to the text blocks in length and order. " + "The page information.")
+public class DocumentData implements Serializable {
 
+    @Schema(description = "Contains information about the document's pages.")
     DocumentPage[] documentPages;
+    @Schema(description = "Contains information about the document's text.")
     DocumentTextData[] documentTextData;
+    @Schema(description = "Contains information about the document's text positions.")
     DocumentPositionData[] documentPositions;
+    @Schema(description = "Contains information about the document's semantic structure.")
     DocumentStructure documentStructure;
 
 }

layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/redaction/DocumentPage.java

@@ -1,5 +1,8 @@
 package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
 
+import java.io.Serializable;
+
+import io.swagger.v3.oas.annotations.media.Schema;
 import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
 import lombok.Builder;
@@ -12,11 +15,16 @@ import lombok.experimental.FieldDefaults;
 @NoArgsConstructor
 @AllArgsConstructor
 @FieldDefaults(level = AccessLevel.PRIVATE)
-public class DocumentPage {
+@Schema(description = "Object containing information about the document's pages.")
+public class DocumentPage implements Serializable {
 
+    @Schema(description = "The page number, starting with 1.")
     int number;
+    @Schema(description = "The page height in PDF user units.", example = "792")
     int height;
+    @Schema(description = "The page width in PDF user units.", example = "694")
     int width;
+    @Schema(description = "The page rotation as specified by the PDF.", example = "90", allowableValues = {"0", "90", "180", "270"})
     int rotation;
 
 }

layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/redaction/DocumentPositionData.java

@@ -1,5 +1,8 @@
 package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
 
+import java.io.Serializable;
+
+import io.swagger.v3.oas.annotations.media.Schema;
 import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
 import lombok.Builder;
@@ -12,10 +15,14 @@ import lombok.experimental.FieldDefaults;
 @NoArgsConstructor
 @AllArgsConstructor
 @FieldDefaults(level = AccessLevel.PRIVATE)
-public class DocumentPositionData {
+@Schema(description = "Object containing text positional information of a specific text block. A document is split into multiple text blocks, which are supposed to be read in order. Every text block can only occur on a single page.")
+public class DocumentPositionData implements Serializable {
 
+    @Schema(description = "Identifier of the text block.")
     Long id;
+    @Schema(description = "For each string coordinate in the search text of the text block, the array contains an entry relating the string coordinate to the position coordinate. This is required due to the text and position coordinates not being equal.")
     int[] stringIdxToPositionIdx;
+    @Schema(description = "The bounding box for each glyph as a rectangle. This matrix is of size (n,4), where n is the number of glyphs in the text block. The second dimension specifies the rectangle with the value x, y, width, height, with x, y specifying the lower left corner. In order to access this information, the stringIdxToPositionIdx array must be used to transform the coordinates.")
     float[][] positions;
 
 }

layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/redaction/DocumentStructure.java

@@ -1,11 +1,14 @@
 package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
 
 import java.awt.geom.Rectangle2D;
+import java.io.Serializable;
 import java.util.Arrays;
 import java.util.List;
 import java.util.Map;
+import java.util.Set;
 import java.util.stream.Stream;
 
+import io.swagger.v3.oas.annotations.media.Schema;
 import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
 import lombok.Builder;
@@ -18,28 +21,34 @@ import lombok.experimental.FieldDefaults;
 @NoArgsConstructor
 @AllArgsConstructor
 @FieldDefaults(level = AccessLevel.PRIVATE)
-public class DocumentStructure {
+@Schema(description = "Object containing information about the parsed tree structure of the SemanticNodes, such as Section, Paragraph, Headline etc inside of the document.")
+public class DocumentStructure implements Serializable {
 
+    @Schema(description = "The root EntryData represents the Document.")
     EntryData root;
 
-
-    public static class TableProperties {
+    @Schema(description = "Object containing the extra field names, a table has in its properties field.")
+    public static class TableProperties implements Serializable {
 
         public static final String NUMBER_OF_ROWS = "numberOfRows";
         public static final String NUMBER_OF_COLS = "numberOfCols";
 
     }
 
-    public static class ImageProperties {
+    @Schema(description = "Object containing the extra field names, an Image has in its properties field.")
+    public static class ImageProperties implements Serializable {
 
         public static final String TRANSPARENT = "transparent";
         public static final String IMAGE_TYPE = "imageType";
         public static final String POSITION = "position";
         public static final String ID = "id";
 
+        public static final String REPRESENTATION_HASH = "representationHash";
+
     }
 
-    public static class TableCellProperties {
+    @Schema(description = "Object containing the extra field names, a table cell has in its properties field.")
+    public static class TableCellProperties implements Serializable {
 
         public static final String B_BOX = "bBox";
         public static final String ROW = "row";
@@ -48,15 +57,37 @@ public class DocumentStructure {
 
     }
 
+    @Schema(description = "Object containing the extra field names, a duplicate paragraph has in its properties field.")
+    public static class DuplicateParagraphProperties implements Serializable {
+
+        public static final String UNSORTED_TEXTBLOCK_ID = "utbid";
+
+    }
+
     public static final String RECTANGLE_DELIMITER = ";";
 
+
     public static Rectangle2D parseRectangle2D(String bBox) {
 
-        List<Float> floats = Arrays.stream(bBox.split(RECTANGLE_DELIMITER)).map(Float::parseFloat).toList();
+        List<Float> floats = Arrays.stream(bBox.split(RECTANGLE_DELIMITER))
+                .map(Float::parseFloat)
+                .toList();
         return new Rectangle2D.Float(floats.get(0), floats.get(1), floats.get(2), floats.get(3));
     }
 
 
+    public static double[] parseRepresentationVector(String representationHash) {
+
+        String[] stringArray = representationHash.split("[,\\s]+");
+        double[] doubleArray = new double[stringArray.length];
+        for (int i = 0; i < stringArray.length; i++) {
+            doubleArray[i] = Double.parseDouble(stringArray[i]);
+        }
+
+        return doubleArray;
+    }
+
+
     public EntryData get(List<Integer> tocId) {
 
         if (tocId.isEmpty()) {
@@ -72,19 +103,24 @@ public class DocumentStructure {
 
     public Stream<EntryData> streamAllEntries() {
 
-        return Stream.concat(Stream.of(root), root.children.stream()).flatMap(DocumentStructure::flatten);
+        return Stream.concat(Stream.of(root), root.children.stream())
+                .flatMap(DocumentStructure::flatten);
     }
 
 
     public String toString() {
 
-        return String.join("\n", streamAllEntries().map(EntryData::toString).toList());
+        return String.join("\n",
+                           streamAllEntries().map(EntryData::toString)
+                                   .toList());
     }
 
 
     private static Stream<EntryData> flatten(EntryData entry) {
 
-        return Stream.concat(Stream.of(entry), entry.children.stream().flatMap(DocumentStructure::flatten));
+        return Stream.concat(Stream.of(entry),
+                             entry.children.stream()
+                                     .flatMap(DocumentStructure::flatten));
     }
 
 
@@ -93,14 +129,23 @@ public class DocumentStructure {
     @NoArgsConstructor
     @AllArgsConstructor
     @FieldDefaults(level = AccessLevel.PRIVATE)
-    public static class EntryData {
+    @Schema(description = "Object containing information of a SemanticNode and also structuring the layout with children.")
+    public static class EntryData implements Serializable {
 
+        @Schema(description = "Type of the semantic node.", allowableValues = {"DOCUMENT", "SECTION", "PARAGRAPH", "HEADLINE", "TABLE", "TABLE_CELL", "HEADER", "FOOTER", "IMAGE"})
         NodeType type;
+        @Schema(description = "Specifies the position in the parsed tree structure.", example = "[1, 0, 2]")
         int[] treeId;
+        @Schema(description = "Specifies the text block IDs associated with this semantic node. The value should be joined with the DocumentTextData/DocumentPositionData. Is empty, if no text block is directly associated with this semantic node. Only Paragraph, Headline, Header or Footer is directly associated with a text block.", example = "[1]")
         Long[] atomicBlockIds;
+        @Schema(description = "Specifies the pages this semantic node appears on. The value should be joined with the PageData.", example = "[1, 2, 3]")
         Long[] pageNumbers;
+        @Schema(description = "Some semantic nodes have additional information, this information is stored in this Map. The extra fields are specified by the Properties subclasses.", example = "For a Table: {\"numberOfRows\": 3, \"numberOfCols\": 4}")
         Map<String, String> properties;
+        @Schema(description = "All child Entries of this Entry.", example = "[1, 2, 3]")
         List<EntryData> children;
+        @Schema(description = "Describes the origin of the semantic node", example = "[ALGORITHM]")
+        Set<LayoutEngine> engines;
 
 
         @Override

layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/redaction/DocumentTextData.java

@@ -1,7 +1,8 @@
 package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
 
+import java.io.Serializable;
 
-
+import io.swagger.v3.oas.annotations.media.Schema;
 import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
 import lombok.Builder;
@@ -14,14 +15,22 @@ import lombok.experimental.FieldDefaults;
 @NoArgsConstructor
 @AllArgsConstructor
 @FieldDefaults(level = AccessLevel.PRIVATE)
-public class DocumentTextData {
+@Schema(description = "Object containing text information of a specific text block. A document is split into multiple text blocks, which are supposed to be read in order. Every text block can only occur on a single page.")
+public class DocumentTextData implements Serializable {
 
+    @Schema(description = "Identifier of the text block.")
     Long id;
+    @Schema(description = "The page the text block occurs on.")
     Long page;
+    @Schema(description = "The text the text block.")
     String searchText;
+    @Schema(description = "Each text block is assigned a number on a page, starting from 0.")
     int numberOnPage;
+    @Schema(description = "The text blocks are ordered, this number represents the start of the text block as a string offset.")
     int start;
+    @Schema(description = "The text blocks are ordered, this number represents the end of the text block as a string offset.")
     int end;
+    @Schema(description = "The line breaks in the text of this semantic node in string offsets. They are exclusive end. At the end of each semantic node there is an implicit linebreak.", example = "[5, 10]")
     int[] lineBreaks;
 
 }

layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/redaction/LayoutEngine.java

@@ -0,0 +1,7 @@
+package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
+
+public enum LayoutEngine {
+    ALGORITHM,
+    AI,
+    OUTLINE
+}

layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/redaction/NodeType.java

@@ -1,10 +1,12 @@
 package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
 
+import java.io.Serializable;
 import java.util.Locale;
 
-public enum NodeType {
+public enum NodeType implements Serializable {
     DOCUMENT,
     SECTION,
+    SUPER_SECTION,
     HEADLINE,
     PARAGRAPH,
     TABLE,

layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/redaction/SimplifiedSectionText.java

@@ -1,5 +1,6 @@
 package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
 
+import io.swagger.v3.oas.annotations.media.Schema;
 import lombok.AllArgsConstructor;
 import lombok.Builder;
 import lombok.Data;
@@ -9,9 +10,12 @@ import lombok.NoArgsConstructor;
 @Builder
 @NoArgsConstructor
 @AllArgsConstructor
+@Schema(description = "Object containing a simplified version, which contains almost exclusively text, of the document structure Section class.")
 public class SimplifiedSectionText {
 
-    private int sectionNumber;
+    @Schema(description = "The number of this Section. This is used to map the simplified section text back to the original Section.")
+    private String sectionNumber;
+    @Schema(description = "The text in this Section.")
     private String text;
 
 }

layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/redaction/SimplifiedText.java

@@ -3,6 +3,7 @@ package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
 import java.util.ArrayList;
 import java.util.List;
 
+import io.swagger.v3.oas.annotations.media.Schema;
 import lombok.AllArgsConstructor;
 import lombok.Builder;
 import lombok.Data;
@@ -12,9 +13,22 @@ import lombok.NoArgsConstructor;
 @Builder
 @NoArgsConstructor
 @AllArgsConstructor
+@Schema(description = "Object containing a simplified version, which contains almost exclusively text, of the document structure.")
 public class SimplifiedText {
 
+    @Schema(description = "Number of pages in the entire document.")
     private int numberOfPages;
+    @Schema(description = "A List of simplified Sections, which contains almost exclusively text.")
+    @Builder.Default
     private List<SimplifiedSectionText> sectionTexts = new ArrayList<>();
+    @Schema(description = "A list of the main section numbers ")
+    @Builder.Default
+    private List<String> mainSectionNumbers = new ArrayList<>();
+    @Schema(description = "A list of the header section numbers ")
+    @Builder.Default
+    private List<String> headerSectionNumbers = new ArrayList<>();
+    @Schema(description = "A list of the footer section numbers ")
+    @Builder.Default
+    private List<String> footerSectionNumbers = new ArrayList<>();
 
 }

layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/taas/ParagraphData.java

@@ -2,20 +2,29 @@ package com.knecon.fforesight.service.layoutparser.internal.api.data.taas;
 
 import java.util.List;
 
+import io.swagger.v3.oas.annotations.media.Schema;
 import lombok.Builder;
 import lombok.Data;
 
 @Data
 @Builder
+@Schema(description = "Object containing information about a Paragraph/Headline/Header/Footer.")
 public class ParagraphData {
 
+    @Schema(description = "The text of this Semantic Node, without any linebreaks.", example = "This is some text.")
     private String text;
+    @Schema(description = "A list of text ranges in string offsets. Every character in any of the ranges is bold.", example = "[0, 15]")
     List<Range> boldTextBoundaries;
+    @Schema(description = "A list of text ranges in string offsets. Every character in any of the ranges is italic.", example = "[0, 15]")
     List<Range> italicTextBoundaries;
+    @Schema(description = "The line breaks in the text of this semantic node in string offsets. They are exclusive end. At the end of each semantic node there is an implicit linebreak.", example = "[5, 10]")
     List<Integer> linebreaks;
+    @Schema(description = "The classification of this Paragraph.", allowableValues = "{paragraph, headline, header, footer}")
     private String classification;
 
+    @Schema(description = "Describes the text orientation of this semantic node. Any semantic node only has a single text orientation.", allowableValues = "{ZERO, QUARTER_CIRCLE, HALF_CIRCLE, THREE_QUARTER_CIRCLE}")
     private String orientation;
+    @Schema(description = "Describes the text direction in degrees of this semantic node. Any semantic node only has a single text direction.", minimum = "0", maximum = "359")
     private int textDirection;
 
 }

layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/taas/Range.java

@@ -1,5 +1,8 @@
 package com.knecon.fforesight.service.layoutparser.internal.api.data.taas;
 
+import io.swagger.v3.oas.annotations.media.Schema;
+
+@Schema(description = "Object specifying the start and end offsets of a text range in string offsets.")
 public record Range(int start, int end) {
 
 }

layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/taas/ResearchDocumentData.java

@@ -2,6 +2,7 @@ package com.knecon.fforesight.service.layoutparser.internal.api.data.taas;
 
 import java.util.List;
 
+import io.swagger.v3.oas.annotations.media.Schema;
 import lombok.AllArgsConstructor;
 import lombok.Builder;
 import lombok.Data;
@@ -9,8 +10,12 @@ import lombok.Data;
 @Builder
 @Data
 @AllArgsConstructor
+@Schema(description = "Object containing a simplified version of the document structure. This simplified form only knows Paragraphs and Tables. The Paragraph Objects might be a Paragraph, Headline, Header or Footer.")
 public class ResearchDocumentData {
 
+    @Schema(description = "File name of the original uploaded file.")
     String originalFile;
+    @Schema(description = "A List of all paragraphs/headline or table objects, that have been parsed in this document.")
     List<StructureObject> structureObjects;
+
 }

layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/taas/RowData.java

@@ -2,14 +2,19 @@ package com.knecon.fforesight.service.layoutparser.internal.api.data.taas;
 
 import java.util.List;
 
+import io.swagger.v3.oas.annotations.media.Schema;
 import lombok.AllArgsConstructor;
 import lombok.Data;
 
 @Data
 @AllArgsConstructor
+@Schema(description = "Object containing information about a Table Row.")
 public class RowData {
 
+    @Schema(description = "Boolean indicating whether this table row is classified as a header row.")
     boolean header;
+    @Schema(description = "A list of Objects containing information about the text in each cell of this row.")
     List<ParagraphData> cellText;
+    @Schema(description = "The bounding box of this StructureObject. Is always exactly 4 values representing x, y, w, h, where x, y specify the lower left corner.")
     float[] bBox;
 }

layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/taas/StructureObject.java

@@ -1,5 +1,8 @@
 package com.knecon.fforesight.service.layoutparser.internal.api.data.taas;
 
+import java.util.List;
+
+import io.swagger.v3.oas.annotations.media.Schema;
 import lombok.AllArgsConstructor;
 import lombok.Builder;
 import lombok.Data;
@@ -7,13 +10,22 @@ import lombok.Data;
 @Data
 @Builder
 @AllArgsConstructor
+@Schema(description = "Object containing information about either a Paragraph/Headline/Header/Footer or a Table.")
 public class StructureObject {
 
+    @Schema(description = "The ID of this StructureObject.")
     Integer structureObjectNumber;
+    @Schema(description = "The Tree ID of this StructureObject.")
+    List<Integer> treeId;
+    @Schema(description = "This value indicates the start of the string offsets in this Object, with respect to the reading order.")
     int page;
+    @Schema(description = "This stringOffset indicates the start of the string offsets in this Object, with respect to the reading order of the entire document. It is equal to the previous' StructureObject stringOffset + its length.")
     int stringOffset;
+    @Schema(description = "The bounding box of this StructureObject. Is always exactly 4 values representing x, y, w, h, where x, y specify the lower left corner.", example = "[100, 100, 50, 50]")
     float[] boundingBox;
+    @Schema(description = "Object containing information about a Paragraph/Headline/Header/Footer. Either this or table is null.")
     ParagraphData paragraph;
+    @Schema(description = "Object containing information about a Table. Either this or paragraph is null.")
     TableData table;
 
 }

layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/taas/TableData.java

@@ -2,15 +2,20 @@ package com.knecon.fforesight.service.layoutparser.internal.api.data.taas;
 
 import java.util.List;
 
+import io.swagger.v3.oas.annotations.media.Schema;
 import lombok.AllArgsConstructor;
 import lombok.Data;
 
 @Data
 @AllArgsConstructor
+@Schema(description = "Object containing information about a Table.")
 public class TableData {
 
+    @Schema(description = "A list of Objects containing information about all rows in this table.")
     List<RowData> rowData;
+    @Schema(description = "Number of columns in this table.")
     Integer numberOfCols;
+    @Schema(description = "Number of rows in this table.")
     Integer numberOfRows;
 
 }

layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/queue/LayoutParsingFinishedEvent.java

@@ -2,9 +2,19 @@ package com.knecon.fforesight.service.layoutparser.internal.api.queue;
 
 import java.util.Map;
 
+import io.swagger.v3.oas.annotations.media.Schema;
 import lombok.Builder;
 
 @Builder
-public record LayoutParsingFinishedEvent(Map<String, String> identifier, long duration, int numberOfPages, String message) {
+@Schema(description = "Object containing information about the layout parsing.")
+public record LayoutParsingFinishedEvent(
+        @Schema(description = "General purpose identifier. It is returned exactly the same way it is inserted with the LayoutParsingRequest.")
+        Map<String, String> identifier,//
+        @Schema(description = "The duration of a single layout parsing in ms.")
+        long duration,//
+        @Schema(description = "The number of pages of the parsed document.")
+        int numberOfPages,//
+        @Schema(description = "A general message. It contains some information useful for a developer, like the paths where the files are stored. Not meant to be machine readable.")
+        String message) {
 
 }

layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/queue/LayoutParsingQueueNames.java

@@ -2,7 +2,9 @@ package com.knecon.fforesight.service.layoutparser.internal.api.queue;
 
 public class LayoutParsingQueueNames {
 
-    public static final String LAYOUT_PARSING_REQUEST_QUEUE = "layout_parsing_request_queue";
-    public static final String LAYOUT_PARSING_DLQ = "layout_parsing_dead_letter_queue";
-    public static final String LAYOUT_PARSING_FINISHED_EVENT_QUEUE = "layout_parsing_response_queue";
+    public static final String LAYOUT_PARSING_REQUEST_QUEUE_PREFIX = "layout_parsing_request";
+    public static final String LAYOUT_PARSING_REQUEST_EXCHANGE = "layout_parsing_request_exchange";
+    public static final String LAYOUT_PARSING_RESPONSE_QUEUE_PREFIX = "layout_parsing_response";
+    public static final String LAYOUT_PARSING_RESPONSE_EXCHANGE = "layout_parsing_response_exchange";
+    public static final String LAYOUT_PARSING_DLQ = "layout_parsing_error";
 }

layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/queue/LayoutParsingRequest.java

@@ -3,23 +3,45 @@ package com.knecon.fforesight.service.layoutparser.internal.api.queue;
 import java.util.Map;
 import java.util.Optional;
 
+import io.swagger.v3.oas.annotations.media.Schema;
 import lombok.Builder;
 import lombok.NonNull;
 
 @Builder
+@Schema(description = "Object containing all storage paths the service needs to know.")
 public record LayoutParsingRequest(
+        @Schema(description = "Enum specifying the type of layout parsing to be performed.", allowableValues = "{RedactManager, DocuMine, TAAS}")//
         @NonNull LayoutParsingType layoutParsingType,
+
+        @Schema(description = "General purpose identifiers. They are not changed by the service at all and are returned as is in the response queue.")//
         Map<String, String> identifier,
-        @NonNull String originFileStorageId,
-        Optional<String> tablesFileStorageId,
-        Optional<String> imagesFileStorageId,
-        @NonNull String structureFileStorageId,
-        String researchDocumentStorageId,
-        @NonNull String textBlockFileStorageId,
-        @NonNull String positionBlockFileStorageId,
-        @NonNull String pageFileStorageId,
-        @NonNull String simplifiedTextStorageId,
-        @NonNull String viewerDocumentStorageId,
-        @NonNull String sectionGridStorageId) {
+
+        @Schema(description = "Path to the original PDF file.")//
+        @NonNull String originFileStorageId,//
+
+        @Schema(description = "Optional Path to the table extraction file.")//
+        Optional<String> tablesFileStorageId,//
+        @Schema(description = "Optional Path to the image classification file.")//
+        Optional<String> imagesFileStorageId,//
+
+        @Schema(description = "Optional Path to the the visual layout parsing service file") Optional<String> visualLayoutParsingFileId,//
+
+        @Schema(description = "Path where the Document Structure File will be stored.")//
+        @NonNull String structureFileStorageId,//
+        @Schema(description = "Path where the Research Data File will be stored.")//
+        String researchDocumentStorageId,//
+        @Schema(description = "Path where the Document Text File will be stored.")//
+        @NonNull String textBlockFileStorageId,//
+        @Schema(description = "Path where the Document Positions File will be stored.")//
+        @NonNull String positionBlockFileStorageId,//
+        @Schema(description = "Path where the Document Pages File will be stored.")//
+        @NonNull String pageFileStorageId,//
+        @Schema(description = "Path where the Document Markdown File will be stored.")//
+        Optional<String> documentMarkdownFileStorageId,//
+        @Schema(description = "Path where the Simplified Text File will be stored.")//
+        @NonNull String simplifiedTextStorageId,//
+        @Schema(description = "Path where the Viewer Document PDF will be stored.")//
+        @NonNull String viewerDocumentStorageId
+) {
 
 }

layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/queue/LayoutParsingType.java

@@ -2,6 +2,11 @@ package com.knecon.fforesight.service.layoutparser.internal.api.queue;
 
 public enum LayoutParsingType {
     REDACT_MANAGER,
-    TAAS,
-    DOCUMINE
+    REDACT_MANAGER_OLD,
+    REDACT_MANAGER_PARAGRAPH_DEBUG,
+    REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH,
+    DOCUMINE,
+    DOCUMINE_OLD,
+    CLARIFYND,
+    CLARIFYND_PARAGRAPH_DEBUG
 }

layoutparser-service/layoutparser-service-processor/build.gradle.kts

@@ -1,26 +1,37 @@
 plugins {
     id("com.knecon.fforesight.java-conventions")
-    id("io.freefair.lombok") version "8.2.2"
+    id("io.freefair.lombok") version "8.4"
 }
 
 description = "layoutparser-service-processor"
 
 val jacksonVersion = "2.15.2"
-val pdfBoxVersion = "3.0.0-RC1"
+val pdfBoxVersion = "3.0.0"
 
 dependencies {
     implementation(project(":layoutparser-service-internal-api"))
+    implementation(project(":viewer-doc-processor"))
 
-    implementation("com.iqser.red.service:persistence-service-shared-api-v1:2.131.0") {
+    implementation("com.iqser.red.service:persistence-service-shared-api-v1:2.564.0-RED9010.0") {
         exclude("org.springframework.boot", "spring-boot-starter-security")
         exclude("org.springframework.boot", "spring-boot-starter-validation")
     }
-    implementation("com.knecon.fforesight:tenant-commons:0.10.0")
-    implementation("com.iqser.red.commons:storage-commons:2.33.0")
+    implementation("com.knecon.fforesight:tenant-commons:0.30.0")
+    implementation("com.iqser.red.commons:storage-commons:2.45.0")
 
     implementation("org.apache.pdfbox:pdfbox:${pdfBoxVersion}")
     implementation("org.apache.pdfbox:pdfbox-tools:${pdfBoxVersion}")
     implementation("com.fasterxml.jackson.module:jackson-module-afterburner:${jacksonVersion}")
     implementation("com.fasterxml.jackson.datatype:jackson-datatype-jsr310:${jacksonVersion}")
-    implementation("org.springframework.boot:spring-boot-starter-web:3.1.2")
+    implementation("org.springframework.boot:spring-boot-starter-web:3.1.3")
+    implementation("org.jgrapht:jgrapht-core:1.5.2")
+    implementation("org.apache.pdfbox:jbig2-imageio:3.0.4")
+    implementation("com.github.jai-imageio:jai-imageio-core:1.4.0")
+    implementation("com.github.jai-imageio:jai-imageio-jpeg2000:1.4.0")
+    implementation("org.tinspin:tinspin-indexes:2.1.3")
+    implementation("org.commonmark:commonmark:0.22.0")
+    implementation("org.commonmark:commonmark-ext-gfm-tables:0.22.0")
+    implementation("com.pdftron:PDFNet:10.11.0")
+    implementation("org.apache.commons:commons-text:1.12.0")
+
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/LayoutParsingPipeline.java

@@ -2,142 +2,268 @@ package com.knecon.fforesight.service.layoutparser.processor;
 
 import static java.lang.String.format;
 
-import java.io.ByteArrayOutputStream;
+import java.awt.geom.Rectangle2D;
+import java.io.File;
 import java.io.IOException;
+import java.nio.file.Files;
 import java.util.ArrayList;
+import java.util.HashMap;
+import java.util.HashSet;
 import java.util.List;
 import java.util.Map;
+import java.util.concurrent.atomic.AtomicReference;
 
+import org.apache.pdfbox.Loader;
 import org.apache.pdfbox.pdmodel.PDDocument;
 import org.apache.pdfbox.pdmodel.PDPage;
 import org.apache.pdfbox.pdmodel.common.PDRectangle;
+import org.apache.pdfbox.pdmodel.documentinterchange.markedcontent.PDMarkedContent;
 import org.springframework.stereotype.Service;
 
+import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
 import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingFinishedEvent;
 import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingRequest;
 import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingType;
+import com.knecon.fforesight.service.layoutparser.processor.services.mapper.MarkdownMapper;
 import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationDocument;
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.ImageType;
 import com.knecon.fforesight.service.layoutparser.processor.model.image.ClassifiedImage;
+import com.knecon.fforesight.service.layoutparser.processor.model.outline.OutlineExtractorService;
+import com.knecon.fforesight.service.layoutparser.processor.model.outline.OutlineObject;
+import com.knecon.fforesight.service.layoutparser.processor.model.outline.OutlineValidationService;
+import com.knecon.fforesight.service.layoutparser.processor.model.outline.TOCEnrichmentService;
+import com.knecon.fforesight.service.layoutparser.processor.model.outline.TableOfContents;
+import com.knecon.fforesight.service.layoutparser.processor.model.table.Cell;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRulings;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.adapter.CvTableParsingAdapter;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.adapter.ImageServiceResponseAdapter;
+import com.knecon.fforesight.service.layoutparser.processor.python_api.adapter.VisualLayoutParsingAdapter;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.image.ImageServiceResponse;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.TableCells;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.TableServiceResponse;
+import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.VisualLayoutParsingResponse;
 import com.knecon.fforesight.service.layoutparser.processor.services.BodyTextFrameService;
 import com.knecon.fforesight.service.layoutparser.processor.services.RulingCleaningService;
-import com.knecon.fforesight.service.layoutparser.processor.services.SectionGridCreatorService;
 import com.knecon.fforesight.service.layoutparser.processor.services.SectionsBuilderService;
 import com.knecon.fforesight.service.layoutparser.processor.services.SimplifiedSectionTextService;
 import com.knecon.fforesight.service.layoutparser.processor.services.TableExtractionService;
+import com.knecon.fforesight.service.layoutparser.processor.services.TextRulingsClassifier;
+import com.knecon.fforesight.service.layoutparser.processor.services.blockification.BlockificationPostprocessingService;
+import com.knecon.fforesight.service.layoutparser.processor.services.blockification.DocstrumBlockificationService;
 import com.knecon.fforesight.service.layoutparser.processor.services.blockification.DocuMineBlockificationService;
 import com.knecon.fforesight.service.layoutparser.processor.services.blockification.RedactManagerBlockificationService;
-import com.knecon.fforesight.service.layoutparser.processor.services.blockification.TaasBlockificationService;
+import com.knecon.fforesight.service.layoutparser.processor.services.classification.ClarifyndClassificationService;
 import com.knecon.fforesight.service.layoutparser.processor.services.classification.DocuMineClassificationService;
 import com.knecon.fforesight.service.layoutparser.processor.services.classification.RedactManagerClassificationService;
-import com.knecon.fforesight.service.layoutparser.processor.services.classification.TaasClassificationService;
 import com.knecon.fforesight.service.layoutparser.processor.services.factory.DocumentGraphFactory;
+import com.knecon.fforesight.service.layoutparser.processor.services.graphics.Box;
+import com.knecon.fforesight.service.layoutparser.processor.services.graphics.GraphicExtractorService;
 import com.knecon.fforesight.service.layoutparser.processor.services.mapper.DocumentDataMapper;
 import com.knecon.fforesight.service.layoutparser.processor.services.mapper.TaasDocumentDataMapper;
 import com.knecon.fforesight.service.layoutparser.processor.services.parsing.PDFLinesTextStripper;
-import com.knecon.fforesight.service.layoutparser.processor.services.visualization.ViewerDocumentService;
+import com.knecon.fforesight.service.layoutparser.processor.services.visualization.LayoutGridService;
+import com.knecon.fforesight.service.layoutparser.processor.utils.MarkedContentUtils;
+import com.knecon.fforesight.service.layoutparser.processor.utils.PageInformation;
+import com.knecon.fforesight.service.layoutparser.processor.utils.TextPositionOperations;
 
+import io.micrometer.observation.Observation;
+import io.micrometer.observation.ObservationRegistry;
+import io.micrometer.observation.annotation.Observed;
+import lombok.AccessLevel;
 import lombok.RequiredArgsConstructor;
 import lombok.SneakyThrows;
+import lombok.experimental.FieldDefaults;
 import lombok.extern.slf4j.Slf4j;
 
+@SuppressWarnings("PMD.CloseResource")
 @Slf4j
 @Service
 @RequiredArgsConstructor
+@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
 public class LayoutParsingPipeline {
 
-    private final ImageServiceResponseAdapter imageServiceResponseAdapter;
-    private final CvTableParsingAdapter cvTableParsingAdapter;
-    private final LayoutParsingStorageService layoutParsingStorageService;
-    private final SectionsBuilderService sectionsBuilderService;
-    private final SectionGridCreatorService sectionGridCreatorService;
-    private final TaasClassificationService taasClassificationService;
-    private final RedactManagerClassificationService redactManagerClassificationService;
-    private final DocuMineClassificationService docuMineClassificationService;
-    private final SimplifiedSectionTextService simplifiedSectionTextService;
-    private final BodyTextFrameService bodyTextFrameService;
-    private final RulingCleaningService rulingCleaningService;
-    private final TableExtractionService tableExtractionService;
-    private final TaasBlockificationService taasBlockificationService;
-    private final DocuMineBlockificationService docuMineBlockificationService;
-    private final RedactManagerBlockificationService redactManagerBlockificationService;
-    private final ViewerDocumentService viewerDocumentService;
+    ImageServiceResponseAdapter imageServiceResponseAdapter;
+    CvTableParsingAdapter cvTableParsingAdapter;
+    LayoutParsingStorageService layoutParsingStorageService;
+    SectionsBuilderService sectionsBuilderService;
+    RedactManagerClassificationService redactManagerClassificationService;
+    DocuMineClassificationService docuMineClassificationService;
+    SimplifiedSectionTextService simplifiedSectionTextService;
+    BodyTextFrameService bodyTextFrameService;
+    RulingCleaningService rulingCleaningService;
+    TableExtractionService tableExtractionService;
+    DocuMineBlockificationService docuMineBlockificationService;
+    RedactManagerBlockificationService redactManagerBlockificationService;
+    BlockificationPostprocessingService blockificationPostprocessingService;
+    DocstrumBlockificationService docstrumBlockificationService;
+    LayoutGridService layoutGridService;
+    ObservationRegistry observationRegistry;
+    VisualLayoutParsingAdapter visualLayoutParsingAdapter;
+    ClarifyndClassificationService clarifyndClassificationService;
+    GraphicExtractorService graphicExtractorService;
+    OutlineExtractorService outlineExtractorService;
+    OutlineValidationService outlineValidationService;
+    TOCEnrichmentService tocEnrichmentService;
+    LayoutparserSettings settings;
 
 
     public LayoutParsingFinishedEvent parseLayoutAndSaveFilesToStorage(LayoutParsingRequest layoutParsingRequest) throws IOException {
 
         long start = System.currentTimeMillis();
+        log.info("Starting layout parsing for {}", layoutParsingRequest.identifier());
 
-        try (PDDocument originDocument = layoutParsingStorageService.getOriginFile(layoutParsingRequest.originFileStorageId())) {
-            ImageServiceResponse imageServiceResponse = new ImageServiceResponse();
-            if (layoutParsingRequest.imagesFileStorageId().isPresent()) {
-                imageServiceResponse = layoutParsingStorageService.getImagesFile(layoutParsingRequest.imagesFileStorageId().get());
-            }
+        File originFile = layoutParsingStorageService.getOriginFile(layoutParsingRequest.originFileStorageId());
+        File viewerDocumentFile = layoutParsingStorageService.getViewerDocFile(layoutParsingRequest.viewerDocumentStorageId()).orElse(originFile);
 
-            TableServiceResponse tableServiceResponse = new TableServiceResponse();
-            if (layoutParsingRequest.tablesFileStorageId().isPresent()) {
-                tableServiceResponse = layoutParsingStorageService.getTablesFile(layoutParsingRequest.tablesFileStorageId().get());
-            }
+        VisualLayoutParsingResponse visualLayoutParsingResponse = layoutParsingRequest.visualLayoutParsingFileId()
+                .map(layoutParsingStorageService::getVisualLayoutParsingFile).orElse(new VisualLayoutParsingResponse());
+        ImageServiceResponse imageServiceResponse = layoutParsingRequest.imagesFileStorageId()
+                .map(layoutParsingStorageService::getImagesFile).orElse(new ImageServiceResponse());
+        TableServiceResponse tableServiceResponse = layoutParsingRequest.tablesFileStorageId()
+                .map(layoutParsingStorageService::getTablesFile).orElse(new TableServiceResponse());
 
-            ClassificationDocument classificationDocument = parseLayout(layoutParsingRequest.layoutParsingType(), originDocument, imageServiceResponse, tableServiceResponse);
-            Document documentGraph = DocumentGraphFactory.buildDocumentGraph(classificationDocument);
+        ClassificationDocument classificationDocument = parseLayout(settings.getLayoutParsingTypeOverride() == null //
+                                                                            ? layoutParsingRequest.layoutParsingType() : settings.getLayoutParsingTypeOverride(),
+                                                                    originFile,
+                                                                    imageServiceResponse,
+                                                                    tableServiceResponse,
+                                                                    visualLayoutParsingResponse,
+                                                                    layoutParsingRequest.identifier());
 
-            int numberOfPages = originDocument.getNumberOfPages();
+        log.info("Building document graph for {}", layoutParsingRequest.identifier());
 
-            layoutParsingStorageService.storeSectionGrid(layoutParsingRequest, sectionGridCreatorService.createSectionGrid(documentGraph));
-            layoutParsingStorageService.storeDocumentData(layoutParsingRequest, DocumentDataMapper.toDocumentData(documentGraph));
-            layoutParsingStorageService.storeSimplifiedText(layoutParsingRequest, simplifiedSectionTextService.toSimplifiedText(documentGraph));
+        Document documentGraph = observeBuildDocumentGraph(settings.getLayoutParsingTypeOverride() == null //
+                                                                   ? layoutParsingRequest.layoutParsingType() : settings.getLayoutParsingTypeOverride(), classificationDocument);
 
-            try (var out = new ByteArrayOutputStream()) {
-                viewerDocumentService.createViewerDocument(originDocument, documentGraph, out);
-                layoutParsingStorageService.storeViewerDocument(layoutParsingRequest, out);
-            }
+        log.info("Creating viewer document for {}", layoutParsingRequest.identifier());
 
-            if (layoutParsingRequest.layoutParsingType().equals(LayoutParsingType.TAAS)) {
-                var researchDocumentData = TaasDocumentDataMapper.fromDocument(documentGraph);
-                layoutParsingStorageService.storeResearchDocumentData(layoutParsingRequest, researchDocumentData);
-            }
+        layoutGridService.addLayoutGrid(viewerDocumentFile, documentGraph, viewerDocumentFile, false);
 
-            return LayoutParsingFinishedEvent.builder()
-                    .identifier(layoutParsingRequest.identifier())
-                    .numberOfPages(numberOfPages)
-                    .duration(System.currentTimeMillis() - start)
-                    .message(format("Layout parsing is finished and files have been saved with Ids:\n Structure: %s\nText: %s\nPositions: %s\nPageData: %s",
-                            layoutParsingRequest.structureFileStorageId(),
-                            layoutParsingRequest.textBlockFileStorageId(),
-                            layoutParsingRequest.positionBlockFileStorageId(),
-                            layoutParsingRequest.pageFileStorageId()))
-                    .build();
+        log.info("Storing resulting files for {}", layoutParsingRequest.identifier());
+
+        layoutParsingStorageService.storeDocumentData(layoutParsingRequest, DocumentDataMapper.toDocumentData(documentGraph));
+        if (layoutParsingRequest.documentMarkdownFileStorageId().isPresent()) {
+            layoutParsingStorageService.storeMarkdownFile(layoutParsingRequest.documentMarkdownFileStorageId().get(), new MarkdownMapper().toMarkdownContent(documentGraph));
         }
+        layoutParsingStorageService.storeSimplifiedText(layoutParsingRequest, simplifiedSectionTextService.toSimplifiedText(documentGraph));
+        layoutParsingStorageService.storeViewerDocument(layoutParsingRequest, viewerDocumentFile);
+
+        if (layoutParsingRequest.researchDocumentStorageId() != null) {
+            log.info("Building research document data for {}", layoutParsingRequest.identifier());
+            var researchDocumentData = TaasDocumentDataMapper.fromDocument(documentGraph);
+            layoutParsingStorageService.storeResearchDocumentData(layoutParsingRequest, researchDocumentData);
+        }
+
+        if (!viewerDocumentFile.equals(originFile)) {
+            viewerDocumentFile.delete();
+        }
+        originFile.delete();
+
+        return LayoutParsingFinishedEvent.builder()
+                .identifier(layoutParsingRequest.identifier())
+                .numberOfPages(documentGraph.getNumberOfPages())
+                .duration(System.currentTimeMillis() - start)
+                .message(format("""
+                                        Layout parsing has finished in %.02f s.
+                                        identifiers: %s
+                                        %s
+                                        Files have been saved with Ids:
+                                        Structure: %s
+                                        Text: %s
+                                        Positions: %s
+                                        PageData: %s
+                                        Simplified Text: %s
+                                        Viewer Doc: %s""",
+                                ((float) (System.currentTimeMillis() - start)) / 1000,
+                                layoutParsingRequest.identifier(),
+                                buildSemanticNodeCountMessage(documentGraph.getNumberOfPages(), documentGraph.buildSemanticNodeCounts()),
+                                layoutParsingRequest.structureFileStorageId(),
+                                layoutParsingRequest.textBlockFileStorageId(),
+                                layoutParsingRequest.positionBlockFileStorageId(),
+                                layoutParsingRequest.pageFileStorageId(),
+                                layoutParsingRequest.simplifiedTextStorageId(),
+                                layoutParsingRequest.viewerDocumentStorageId()))
+                .build();
+
+    }
+
+
+    private Document observeBuildDocumentGraph(LayoutParsingType layoutParsingType, ClassificationDocument classificationDocument) {
+
+        AtomicReference<Document> documentReference = new AtomicReference<>();
+
+        Observation.createNotStarted("LayoutParsingPipeline", observationRegistry)
+                .contextualName("build-document-graph")
+                .observe(() -> documentReference.set(DocumentGraphFactory.buildDocumentGraph(layoutParsingType, classificationDocument)));
+
+        return documentReference.get();
+    }
+
+
+    private String buildSemanticNodeCountMessage(int numberOfPages, Map<NodeType, Long> semanticNodeCounts) {
+
+        return format("%d pages with %d sections, %d headlines, %d paragraphs, %d tables with %d cells, %d headers, and %d footers parsed",
+                      numberOfPages,
+                      semanticNodeCounts.get(NodeType.SECTION) == null ? 0 : semanticNodeCounts.get(NodeType.SECTION),
+                      semanticNodeCounts.get(NodeType.HEADLINE) == null ? 0 : semanticNodeCounts.get(NodeType.HEADLINE),
+                      semanticNodeCounts.get(NodeType.PARAGRAPH) == null ? 0 : semanticNodeCounts.get(NodeType.PARAGRAPH),
+                      semanticNodeCounts.get(NodeType.TABLE) == null ? 0 : semanticNodeCounts.get(NodeType.TABLE),
+                      semanticNodeCounts.get(NodeType.TABLE_CELL) == null ? 0 : semanticNodeCounts.get(NodeType.TABLE_CELL),
+                      semanticNodeCounts.get(NodeType.HEADER) == null ? 0 : semanticNodeCounts.get(NodeType.HEADER),
+                      semanticNodeCounts.get(NodeType.FOOTER) == null ? 0 : semanticNodeCounts.get(NodeType.FOOTER));
     }
 
 
     @SneakyThrows
+    @Observed(name = "LayoutParsingPipeline", contextualName = "parse-layout")
     public ClassificationDocument parseLayout(LayoutParsingType layoutParsingType,
-                                              PDDocument originDocument,
+                                              File originFile,
                                               ImageServiceResponse imageServiceResponse,
-                                              TableServiceResponse tableServiceResponse) {
+                                              TableServiceResponse tableServiceResponse,
+                                              VisualLayoutParsingResponse visualLayoutParsingResponse,
+                                              Map<String, String> identifier) {
+
+        PDDocument originDocument = openDocument(originFile);
+        addNumberOfPagesToTrace(originDocument.getNumberOfPages(), Files.size(originFile.toPath()));
 
         Map<Integer, List<TableCells>> pdfTableCells = cvTableParsingAdapter.buildCvParsedTablesPerPage(tableServiceResponse);
         Map<Integer, List<ClassifiedImage>> pdfImages = imageServiceResponseAdapter.buildClassifiedImagesPerPage(imageServiceResponse);
+        Map<Integer, List<ClassifiedImage>> signatures = visualLayoutParsingAdapter.buildExtractedSignaturesPerPage(visualLayoutParsingResponse);
 
         ClassificationDocument classificationDocument = new ClassificationDocument();
-        List<ClassificationPage> classificationPages = new ArrayList<>();
 
-        originDocument.setAllSecurityToBeRemoved(true);
+        if (settings.isDebug() || identifier.containsKey("debug")) {
+            classificationDocument.getLayoutDebugLayer().setActive(true);
+        }
+
+        List<ClassificationPage> classificationPages = new ArrayList<>();
+        OutlineObject lastProcessedOutlineObject = null;
+
+        // parsing the structure elements could be useful as well
+        if (layoutParsingType != LayoutParsingType.REDACT_MANAGER_OLD) {
+            classificationDocument.setOutlineObjectTree(outlineExtractorService.getOutlineObjectTree(originDocument));
+        }
+
         long pageCount = originDocument.getNumberOfPages();
 
         for (int pageNumber = 1; pageNumber <= pageCount; pageNumber++) {
+
+            if (pageNumber % 100 == 0) {
+                // re-open document every once in a while to save on RAM. This has no significant performance impact.
+                // This is due to PDFBox caching all images and some other stuff with Soft References. This dereferences them and forces the freeing of memory.
+                originDocument.close();
+                originDocument = openDocument(originFile);
+            }
+
+            if (pageNumber % 100 == 0 || pageNumber == pageCount || pageNumber == 1) {
+                log.info("Extracting text on Page {} for {}", pageNumber, identifier);
+            }
+
             classificationDocument.setPages(classificationPages);
             PDFLinesTextStripper stripper = new PDFLinesTextStripper();
             PDPage pdPage = originDocument.getPage(pageNumber - 1);
@@ -145,10 +271,14 @@ public class LayoutParsingPipeline {
             stripper.setStartPage(pageNumber);
             stripper.setEndPage(pageNumber);
             stripper.setPdpage(pdPage);
-            if (layoutParsingType.equals(LayoutParsingType.DOCUMINE)) {
-                stripper.setSortByPosition(true);
-            }
             stripper.getText(originDocument);
+            List<TextPositionSequence> words = stripper.getTextPositionSequences();
+            if (layoutParsingType.equals(LayoutParsingType.DOCUMINE_OLD)) {
+                var lines = TextPositionOperations.groupByLine(new HashSet<>(words));
+                classificationDocument.getLayoutDebugLayer().addLineVisualizationsFromNestedTextPosition(lines, pageNumber);
+                words = TextPositionOperations.sortLines(lines);
+            }
+            classificationDocument.getLayoutDebugLayer().addTextVisualizations(words, pageNumber);
 
             PDRectangle pdr = pdPage.getMediaBox();
 
@@ -156,16 +286,36 @@ public class LayoutParsingPipeline {
             boolean isLandscape = pdr.getWidth() > pdr.getHeight() && (rotation == 0 || rotation == 180) || pdr.getHeight() > pdr.getWidth() && (rotation == 90 || rotation == 270);
 
             PDRectangle cropbox = pdPage.getCropBox();
-            CleanRulings cleanRulings = rulingCleaningService.getCleanRulings(pdfTableCells.get(pageNumber),
-                    stripper.getRulings(),
-                    stripper.getMinCharWidth(),
-                    stripper.getMaxCharHeight());
+            classificationDocument.getLayoutDebugLayer().addRulingVisualization(stripper.getRulings(), pageNumber);
+            CleanRulings cleanRulings = rulingCleaningService.deduplicateAndStraightenRulings(pdfTableCells.get(pageNumber), stripper.getRulings());
+
+            PageInformation pageInformation = PageInformation.fromPDPage(pageNumber, pdPage);
+            List<Cell> emptyTableCells = TableExtractionService.findCells(cleanRulings.getHorizontals(), cleanRulings.getVerticals(), pageInformation);
+            classificationDocument.getLayoutDebugLayer().addCellVisualizations(emptyTableCells, pageNumber);
+
+            TextRulingsClassifier.classifyUnderlinedAndStrikethroughText(words, cleanRulings);
+
+            List<Box> graphics = graphicExtractorService.extractPathElementGraphics(originDocument, pdPage, pageNumber, cleanRulings, stripper.getTextPositionSequences(), false);
+
+            pdfImages.computeIfAbsent(pageNumber, x -> new ArrayList<>())
+                    .addAll(graphics.stream()
+                                    .map(g -> new ClassifiedImage(new Rectangle2D.Double(g.x1, g.y1, g.width(), g.height()),
+                                                                  ImageType.GRAPHIC,
+                                                                  false,
+                                                                  stripper.getPageNumber(),
+                                                                  ""))
+                                    .toList());
 
             ClassificationPage classificationPage = switch (layoutParsingType) {
-                case REDACT_MANAGER -> redactManagerBlockificationService.blockify(stripper.getTextPositionSequences(), cleanRulings.getHorizontal(), cleanRulings.getVertical());
-                case TAAS -> taasBlockificationService.blockify(stripper.getTextPositionSequences(), cleanRulings.getHorizontal(), cleanRulings.getVertical());
-                case DOCUMINE -> docuMineBlockificationService.blockify(stripper.getTextPositionSequences(), cleanRulings.getHorizontal(), cleanRulings.getVertical());
+                case REDACT_MANAGER_OLD ->
+                        redactManagerBlockificationService.blockify(stripper.getTextPositionSequences(), cleanRulings, classificationDocument.getLayoutDebugLayer());
+                case DOCUMINE_OLD -> docuMineBlockificationService.blockify(words, cleanRulings);
+                case DOCUMINE, REDACT_MANAGER, REDACT_MANAGER_PARAGRAPH_DEBUG, REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH ->
+                        docstrumBlockificationService.blockify(words, cleanRulings, true, classificationDocument.getLayoutDebugLayer(), layoutParsingType);
+                case CLARIFYND, CLARIFYND_PARAGRAPH_DEBUG ->
+                        docstrumBlockificationService.blockify(words, cleanRulings, false, classificationDocument.getLayoutDebugLayer(), layoutParsingType);
             };
+
             classificationPage.setCleanRulings(cleanRulings);
             classificationPage.setRotation(rotation);
             classificationPage.setLandscape(isLandscape);
@@ -173,39 +323,119 @@ public class LayoutParsingPipeline {
             classificationPage.setPageWidth(cropbox.getWidth());
             classificationPage.setPageHeight(cropbox.getHeight());
 
+            if (layoutParsingType != LayoutParsingType.REDACT_MANAGER_OLD) {
+                List<OutlineObject> outlineObjects = classificationDocument.getOutlineObjectTree().getOutlineObjectsPerPage().getOrDefault(pageNumber, new ArrayList<>());
+
+                OutlineObject notFoundOutlineObject = null;
+                if (lastProcessedOutlineObject != null && !lastProcessedOutlineObject.isFound()) {
+                    lastProcessedOutlineObject.resetPoint();
+                    notFoundOutlineObject = lastProcessedOutlineObject;
+                }
+                if (!outlineObjects.isEmpty()) {
+                    classificationPage.setOutlineObjects(outlineObjects);
+                    lastProcessedOutlineObject = blockificationPostprocessingService.sanitizeOutlineBlocks(classificationPage, notFoundOutlineObject);
+                }
+                classificationDocument.getLayoutDebugLayer().addOutlineObjects(outlineObjects, pageInformation);
+            }
+
+            classificationDocument.getLayoutDebugLayer().addMarkedContentVisualizations(stripper.getMarkedContents(), pageNumber);
+            // MarkedContent needs to be converted at this point, otherwise it leads to GC Problems in Pdfbox.
+            classificationPage.setMarkedContentBboxPerType(convertMarkedContents(stripper.getMarkedContents()));
+
             // If images is ocr needs to be calculated before textBlocks are moved into tables, otherwise findOcr algorithm needs to be adopted.
-            if (pdfImages != null && pdfImages.containsKey(pageNumber)) {
+            if (pdfImages.containsKey(pageNumber)) {
                 classificationPage.setImages(pdfImages.get(pageNumber));
                 imageServiceResponseAdapter.findOcr(classificationPage);
             }
 
-            tableExtractionService.extractTables(cleanRulings, classificationPage, layoutParsingType);
+            if (signatures.containsKey(pageNumber)) {
+                if (classificationPage.getImages() == null || classificationPage.getImages().isEmpty()) {
+                    classificationPage.setImages(signatures.get(pageNumber));
+                } else {
+                    classificationPage.getImages().addAll(signatures.get(pageNumber));
+                }
+            }
+
+            tableExtractionService.extractTables(emptyTableCells, classificationPage);
+
             buildPageStatistics(classificationPage);
             increaseDocumentStatistics(classificationPage, classificationDocument);
 
             classificationPages.add(classificationPage);
         }
 
-        bodyTextFrameService.setBodyTextFrames(classificationDocument, layoutParsingType);
+        originDocument.close();
 
+        log.info("Calculating BodyTextFrame for {}", identifier);
+        bodyTextFrameService.setBodyTextFrames(classificationDocument, layoutParsingType);
+        for (ClassificationPage page : classificationDocument.getPages()) {
+            classificationDocument.getLayoutDebugLayer().addCleanRulingVisualization(page.getCleanRulings(), page.getPageNumber());
+        }
+        log.info("Classify TextBlocks for {}", identifier);
         switch (layoutParsingType) {
-            case TAAS -> taasClassificationService.classifyDocument(classificationDocument);
-            case DOCUMINE -> docuMineClassificationService.classifyDocument(classificationDocument);
-            case REDACT_MANAGER -> redactManagerClassificationService.classifyDocument(classificationDocument);
+            case REDACT_MANAGER, REDACT_MANAGER_PARAGRAPH_DEBUG, REDACT_MANAGER_OLD, CLARIFYND_PARAGRAPH_DEBUG, REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH ->
+                    redactManagerClassificationService.classifyDocument(classificationDocument);
+            case DOCUMINE_OLD, DOCUMINE -> docuMineClassificationService.classifyDocument(classificationDocument);
+            case CLARIFYND -> clarifyndClassificationService.classifyDocument(classificationDocument);
         }
 
-        sectionsBuilderService.buildSections(classificationDocument);
-        sectionsBuilderService.addImagesToSections(classificationDocument);
+        if (layoutParsingType.equals(LayoutParsingType.DOCUMINE_OLD)) {
+            for (ClassificationPage page : classificationDocument.getPages()) {
+                docuMineBlockificationService.mergeblocks(page, page.getCleanRulings().withoutTextRulings(), 0, 10);
+            }
+        }
+
+        List<TextPageBlock> headlines = classificationDocument.getPages()
+                .stream()
+                .flatMap(classificationPage -> classificationPage.getTextBlocks()
+                        .stream()
+                        .filter(tb -> tb instanceof TextPageBlock && tb.getClassification() != null && tb.getClassification().isHeadline())
+                        .map(tb -> (TextPageBlock) tb))
+                .toList();
+        TableOfContents tableOfContents = outlineValidationService.createToC(headlines);
+        classificationDocument.setTableOfContents(tableOfContents);
+
+        log.info("Building Sections for {}", identifier);
+
+        switch (layoutParsingType) {
+            case CLARIFYND_PARAGRAPH_DEBUG, REDACT_MANAGER_PARAGRAPH_DEBUG -> sectionsBuilderService.buildParagraphDebugSections(classificationDocument);
+            default -> tocEnrichmentService.assignSectionBlocksAndImages(classificationDocument);
+        }
 
         return classificationDocument;
     }
 
 
+    private void addNumberOfPagesToTrace(int numberOfPages, long size) {
+
+        if (observationRegistry.getCurrentObservation() != null) {
+            observationRegistry.getCurrentObservation().highCardinalityKeyValue("numberOfPages", String.valueOf(numberOfPages));
+            observationRegistry.getCurrentObservation().highCardinalityKeyValue("fileSize", String.valueOf(size));
+        }
+    }
+
+
+    @SneakyThrows
+    private PDDocument openDocument(File originFile) {
+
+        PDDocument document = Loader.loadPDF(originFile);
+        document.setAllSecurityToBeRemoved(true);
+        return document;
+    }
+
+
+    private Map<String, List<Rectangle2D>> convertMarkedContents(List<PDMarkedContent> pdMarkedContents) {
+
+        Map<String, List<Rectangle2D>> markedContentBboxes = new HashMap<>();
+        markedContentBboxes.put(MarkedContentUtils.HEADER, MarkedContentUtils.getMarkedContentBboxPerLine(pdMarkedContents, MarkedContentUtils.HEADER));
+        markedContentBboxes.put(MarkedContentUtils.FOOTER, MarkedContentUtils.getMarkedContentBboxPerLine(pdMarkedContents, MarkedContentUtils.FOOTER));
+        return markedContentBboxes;
+    }
+
+
     private void increaseDocumentStatistics(ClassificationPage classificationPage, ClassificationDocument document) {
 
-        if (!classificationPage.isLandscape()) {
-            document.getFontSizeCounter().addAll(classificationPage.getFontSizeCounter().getCountPerValue());
-        }
+        document.getFontSizeCounter().addAll(classificationPage.getFontSizeCounter().getCountPerValue());
         document.getFontCounter().addAll(classificationPage.getFontCounter().getCountPerValue());
         document.getTextHeightCounter().addAll(classificationPage.getTextHeightCounter().getCountPerValue());
         document.getFontStyleCounter().addAll(classificationPage.getFontStyleCounter().getCountPerValue());

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/LayoutParsingServiceProcessorConfiguration.java

@@ -1,10 +1,23 @@
 package com.knecon.fforesight.service.layoutparser.processor;
 
+import org.springframework.beans.factory.annotation.Autowired;
+import org.springframework.context.annotation.Bean;
 import org.springframework.context.annotation.ComponentScan;
 import org.springframework.context.annotation.Configuration;
 
+import com.knecon.fforesight.service.viewerdoc.service.PDFTronViewerDocumentService;
+
+import io.micrometer.observation.ObservationRegistry;
+
 @Configuration
 @ComponentScan
 public class LayoutParsingServiceProcessorConfiguration {
 
+    @Bean
+    @Autowired
+    public PDFTronViewerDocumentService viewerDocumentService(ObservationRegistry registry) {
+
+        return new PDFTronViewerDocumentService(registry);
+    }
+
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/LayoutParsingStorageService.java

@@ -1,24 +1,20 @@
 package com.knecon.fforesight.service.layoutparser.processor;
 
 import java.io.ByteArrayInputStream;
-import java.io.ByteArrayOutputStream;
 import java.io.File;
-import java.io.FileOutputStream;
+import java.io.FileInputStream;
 import java.io.IOException;
 import java.io.InputStream;
+import java.nio.charset.StandardCharsets;
 import java.nio.file.Files;
 import java.nio.file.Path;
 import java.nio.file.Paths;
 import java.nio.file.StandardOpenOption;
+import java.util.Optional;
 
-import org.apache.commons.io.IOUtils;
-import org.apache.pdfbox.Loader;
-import org.apache.pdfbox.io.MemoryUsageSetting;
-import org.apache.pdfbox.pdmodel.PDDocument;
 import org.springframework.stereotype.Service;
 
 import com.fasterxml.jackson.databind.ObjectMapper;
-import com.iqser.red.service.persistence.service.v1.api.shared.model.redactionlog.section.SectionGrid;
 import com.iqser.red.storage.commons.service.StorageService;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentData;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.SimplifiedText;
@@ -26,8 +22,11 @@ import com.knecon.fforesight.service.layoutparser.internal.api.data.taas.Researc
 import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingRequest;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.image.ImageServiceResponse;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.TableServiceResponse;
+import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.VisualLayoutParsingResponse;
+import com.knecon.fforesight.service.viewerdoc.service.ViewerDocVersioningUtility;
 import com.knecon.fforesight.tenantcommons.TenantContext;
 
+import io.micrometer.observation.annotation.Observed;
 import lombok.RequiredArgsConstructor;
 import lombok.SneakyThrows;
 import lombok.extern.slf4j.Slf4j;
@@ -41,20 +40,35 @@ public class LayoutParsingStorageService {
     private final ObjectMapper objectMapper;
 
 
-    public PDDocument getOriginFile(String storageId) throws IOException {
+    @Observed(name = "LayoutParsingStorageService", contextualName = "get-origin-file")
+    public File getOriginFile(String storageId) throws IOException {
 
-        try (var originDocumentInputStream = getObject(storageId)) {
-            File tempFile = createTempFile("document", ".pdf");
-            try (var tempFileOutputStream = new FileOutputStream(tempFile)) {
-                IOUtils.copy(originDocumentInputStream, tempFileOutputStream);
-                originDocumentInputStream.close();
-            }
-            return Loader.loadPDF(tempFile, MemoryUsageSetting.setupMixed(67108864L));
-        }
+        File tempFile = createTempFile("document", ".pdf");
+        storageService.downloadTo(TenantContext.getTenantId(), storageId, tempFile);
+        return tempFile;
     }
 
 
-    public ImageServiceResponse getImagesFile(String storageId) throws IOException {
+    @Observed(name = "LayoutParsingStorageService", contextualName = "get-viewer-doc-file")
+    public Optional<File> getViewerDocFile(String storageId) throws IOException {
+
+        if (!storageService.objectExists(TenantContext.getTenantId(), storageId)) {
+            return Optional.empty();
+        }
+        File tempFile = createTempFile("viewerDocument", ".pdf");
+        storageService.downloadTo(TenantContext.getTenantId(), storageId, tempFile);
+
+        if (!ViewerDocVersioningUtility.isCurrentVersion(tempFile)) {
+            assert tempFile.delete();
+            return Optional.empty();
+        }
+
+        return Optional.of(tempFile);
+    }
+
+
+    @SneakyThrows
+    public ImageServiceResponse getImagesFile(String storageId) {
 
         try (InputStream inputStream = getObject(storageId)) {
 
@@ -65,7 +79,8 @@ public class LayoutParsingStorageService {
     }
 
 
-    public TableServiceResponse getTablesFile(String storageId) throws IOException {
+    @SneakyThrows
+    public TableServiceResponse getTablesFile(String storageId) {
 
         try (var tableClassificationStream = getObject(storageId)) {
 
@@ -76,6 +91,16 @@ public class LayoutParsingStorageService {
     }
 
 
+@SneakyThrows
+    public VisualLayoutParsingResponse getVisualLayoutParsingFile(String storageId) {
+
+        try (InputStream inputStream = getObject(storageId)) {
+            return objectMapper.readValue(inputStream, VisualLayoutParsingResponse.class);
+        }
+    }
+
+
+    @Observed(name = "LayoutParsingStorageService", contextualName = "store-document-data")
     public void storeDocumentData(LayoutParsingRequest layoutParsingRequest, DocumentData documentData) {
 
         storageService.storeJSONObject(TenantContext.getTenantId(), layoutParsingRequest.structureFileStorageId(), documentData.getDocumentStructure());
@@ -85,12 +110,6 @@ public class LayoutParsingStorageService {
     }
 
 
-    public void storeSectionGrid(LayoutParsingRequest layoutParsingRequest, SectionGrid sectionGrid) {
-
-        storageService.storeJSONObject(TenantContext.getTenantId(), layoutParsingRequest.sectionGridStorageId(), sectionGrid);
-    }
-
-
     public void storeResearchDocumentData(LayoutParsingRequest layoutParsingRequest, ResearchDocumentData researchDocumentData) {
 
         storageService.storeJSONObject(TenantContext.getTenantId(), layoutParsingRequest.researchDocumentStorageId(), researchDocumentData);
@@ -122,6 +141,7 @@ public class LayoutParsingStorageService {
     }
 
 
+    @Observed(name = "LayoutParsingStorageService", contextualName = "store-simplified-text")
     public void storeSimplifiedText(LayoutParsingRequest layoutParsingRequest, SimplifiedText simplifiedText) {
 
         storageService.storeJSONObject(TenantContext.getTenantId(), layoutParsingRequest.simplifiedTextStorageId(), simplifiedText);
@@ -139,12 +159,24 @@ public class LayoutParsingStorageService {
 
 
     @SneakyThrows
-    public void storeViewerDocument(LayoutParsingRequest layoutParsingRequest, ByteArrayOutputStream out) {
-
-        try (var in = new ByteArrayInputStream(out.toByteArray())) {
+    @Observed(name = "LayoutParsingStorageService", contextualName = "store-viewer-document")
+    public void storeViewerDocument(LayoutParsingRequest layoutParsingRequest, File out) {
 
+        try (var in = new FileInputStream(out)) {
             storageService.storeObject(TenantContext.getTenantId(), layoutParsingRequest.viewerDocumentStorageId(), in);
         }
     }
 
+
+    @SneakyThrows
+    @Observed(name = "LayoutParsingStorageService", contextualName = "store-markdown-file")
+    public void storeMarkdownFile(String markdownFileStorageId, String markdownContent) {
+
+        try (InputStream inputStream = new ByteArrayInputStream(markdownContent.getBytes(StandardCharsets.UTF_8))) {
+
+            storageService.storeObject(TenantContext.getTenantId(), markdownFileStorageId, inputStream);
+        }
+
+    }
+
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/LayoutparserSettings.java

@@ -0,0 +1,21 @@
+package com.knecon.fforesight.service.layoutparser.processor;
+
+import org.springframework.boot.context.properties.ConfigurationProperties;
+import org.springframework.context.annotation.Configuration;
+
+import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingType;
+
+import lombok.AccessLevel;
+import lombok.Data;
+import lombok.experimental.FieldDefaults;
+
+@Data
+@Configuration
+@ConfigurationProperties("layoutparser")
+@FieldDefaults(level = AccessLevel.PRIVATE)
+public class LayoutparserSettings {
+
+    boolean debug;
+    LayoutParsingType layoutParsingTypeOverride;
+    String pdftronLicense;
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/docstrum/DocstrumSegmentationService.java

@@ -0,0 +1,102 @@
+package com.knecon.fforesight.service.layoutparser.processor.docstrum;
+
+import java.util.ArrayList;
+import java.util.EnumMap;
+import java.util.List;
+import java.util.stream.Collectors;
+
+import org.springframework.stereotype.Service;
+
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Character;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Line;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Zone;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.service.LineBuilderService;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.service.NearestNeighbourService;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.service.ReadingOrderService;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.service.SpacingService;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.service.ZoneBuilderService;
+import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRulings;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.RedTextPosition;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.TextDirection;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
+import com.knecon.fforesight.service.layoutparser.processor.visualization.LayoutDebugLayer;
+
+import lombok.RequiredArgsConstructor;
+
+@Service
+@RequiredArgsConstructor
+public class DocstrumSegmentationService {
+
+    public static final double SAME_DIRECTION_THRESHOLD = 0.9;
+    private final NearestNeighbourService nearestNeighbourService;
+    private final SpacingService spacingService;
+    private final LineBuilderService lineBuilderService;
+    private final ZoneBuilderService zoneBuilderService;
+    private final ReadingOrderService readingOrderService;
+
+
+    public List<Zone> segmentPage(List<TextPositionSequence> textPositions, boolean xyOrder, CleanRulings usedRulings, LayoutDebugLayer visualizations) {
+
+        EnumMap<TextDirection, Integer> directionCounts = new EnumMap<>(TextDirection.class);
+
+        List<Zone> newZones = computeZones(textPositions, usedRulings, visualizations, TextDirection.ZERO);
+        directionCounts.put(TextDirection.ZERO, newZones.size());
+        List<Zone> zones = new ArrayList<>(newZones);
+
+        newZones = computeZones(textPositions, usedRulings, visualizations, TextDirection.QUARTER_CIRCLE);
+        directionCounts.put(TextDirection.QUARTER_CIRCLE, newZones.size());
+        zones.addAll(newZones);
+
+        newZones = computeZones(textPositions, usedRulings, visualizations, TextDirection.HALF_CIRCLE);
+        directionCounts.put(TextDirection.HALF_CIRCLE, newZones.size());
+        zones.addAll(newZones);
+
+        newZones = computeZones(textPositions, usedRulings, visualizations, TextDirection.THREE_QUARTER_CIRCLE);
+        directionCounts.put(TextDirection.THREE_QUARTER_CIRCLE, newZones.size());
+        zones.addAll(newZones);
+
+        return readingOrderService.resolve(zones, xyOrder, mostSameDirection(directionCounts));
+    }
+
+
+    private boolean mostSameDirection(EnumMap<TextDirection, Integer> directionCounts) {
+
+        int total = directionCounts.values()
+                .stream()
+                .mapToInt(i -> i).sum();
+
+        if ((double) directionCounts.get(TextDirection.ZERO) / total > SAME_DIRECTION_THRESHOLD) {
+            return true;
+        } else if ((double) directionCounts.get(TextDirection.QUARTER_CIRCLE) / total > SAME_DIRECTION_THRESHOLD) {
+            return true;
+        } else if ((double) directionCounts.get(TextDirection.HALF_CIRCLE) / total > SAME_DIRECTION_THRESHOLD) {
+            return true;
+        } else if ((double) directionCounts.get(TextDirection.THREE_QUARTER_CIRCLE) / total > SAME_DIRECTION_THRESHOLD) {
+            return true;
+        }
+        return false;
+    }
+
+
+    private List<Zone> computeZones(List<TextPositionSequence> textPositions, CleanRulings rulings, LayoutDebugLayer visualizations, TextDirection direction) {
+
+        List<RedTextPosition> positions = textPositions.stream()
+                .filter(t -> t.getDir() == direction)
+                .map(TextPositionSequence::getTextPositions)
+                .flatMap(List::stream)
+                .toList();
+
+        List<Character> characters = positions.stream()
+                .map(Character::new)
+                .collect(Collectors.toList());
+
+        nearestNeighbourService.findNearestNeighbors(characters);
+
+        double characterSpacing = spacingService.computeCharacterSpacing(characters);
+        double lineSpacing = Math.min(spacingService.computeLineSpacing(characters), 20);
+
+        List<Line> lines = lineBuilderService.buildLines(characters, characterSpacing, lineSpacing, rulings);
+        return zoneBuilderService.buildZones(lines, characterSpacing, lineSpacing, rulings);
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/docstrum/model/AngleFilter.java

@@ -0,0 +1,31 @@
+package com.knecon.fforesight.service.layoutparser.processor.docstrum.model;
+
+public class AngleFilter {
+
+    protected double lowerAngle;
+    protected double upperAngle;
+
+
+    public AngleFilter(double lowerAngle, double upperAngle) {
+
+        this.lowerAngle = lowerAngle < -Math.PI / 2 ? lowerAngle + Math.PI : lowerAngle;
+        this.upperAngle = upperAngle >= Math.PI / 2 ? upperAngle - Math.PI : upperAngle;
+    }
+
+
+    public boolean matches(Neighbor neighbor) {
+
+        return matches(neighbor.getAngle());
+    }
+
+
+    public boolean matches(double angle) {
+
+        if (lowerAngle <= upperAngle) {
+            return lowerAngle <= angle && angle < upperAngle;
+        } else {
+            return lowerAngle <= angle || angle < upperAngle;
+        }
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/docstrum/model/BoundingBox.java

@@ -0,0 +1,281 @@
+package com.knecon.fforesight.service.layoutparser.processor.docstrum.model;
+
+import java.awt.geom.Rectangle2D;
+import java.util.Comparator;
+import java.util.List;
+
+import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
+
+import lombok.Data;
+import lombok.NoArgsConstructor;
+import lombok.experimental.SuperBuilder;
+
+@Data
+@SuperBuilder
+@NoArgsConstructor
+public abstract class BoundingBox {
+
+    // Java coordinate system: (0, 0) is always upper left, x is increasing left to right and y is increasing from top to bottom.
+    // should be used when determining reading order or other tasks which require coordinates in a harmonized system.
+    protected Rectangle2D bBox; // I would not trust this coordinate when comparing rulings and text, due to the text positions being slightly off.
+
+    // PDF coordinate system: depends on page rotation, (0, 0) is lower left corner, x is increasing left to right and y from bottom to top.
+    // This rotates completely in 90 degree steps with page rotation.
+    // Needs to be used when writing to a PDF.
+    // Also, these are definitely correct and should be used whenever possible.
+    protected Rectangle2D bBoxPdf;
+
+    protected static final float VERTICAL_COMPARISON_THRESHOLD = 0.4f;
+
+
+    public double getX() {
+
+        return bBox.getX();
+    }
+
+
+    public double getY() {
+
+        return bBox.getY();
+    }
+
+
+    public double getMinX() {
+
+        return bBox.getMinX();
+    }
+
+
+    public double getMinY() {
+
+        return bBox.getMinY();
+    }
+
+
+    public double getPdfMinX() {
+
+        return bBoxPdf.getMinX();
+    }
+
+
+    public double getPdfMaxX() {
+
+        return bBoxPdf.getMaxX();
+    }
+
+
+    public double getPdfMinY() {
+
+        return bBoxPdf.getMinY();
+    }
+
+
+    public double getPdfMaxY() {
+
+        return bBoxPdf.getMaxY();
+    }
+
+
+    public double getWidth() {
+
+        return bBox.getWidth();
+    }
+
+
+    public double getHeight() {
+
+        return bBox.getHeight();
+    }
+
+
+    public double getMaxX() {
+
+        return bBox.getMaxX();
+    }
+
+
+    public double getMaxY() {
+
+        return bBox.getMaxY();
+    }
+
+
+    public double getArea() {
+
+        return (bBox.getHeight() * bBox.getWidth());
+    }
+
+
+    public boolean contains(BoundingBox contained) {
+
+        return contains(contained, 0);
+    }
+
+
+    public boolean contains(BoundingBox contained, double tolerance) {
+
+        return getPdfMinX() <= contained.getPdfMinX() + tolerance
+               && getPdfMinY() <= contained.getPdfMinY() + tolerance
+               && getPdfMaxX() >= contained.getPdfMaxX() - tolerance
+               && getPdfMaxY() >= contained.getPdfMaxY() - tolerance;
+    }
+
+
+    public boolean intersects(BoundingBox other) {
+
+        return this.intersectsX(other) && this.intersectsY(other);
+    }
+
+
+    public boolean intersects(BoundingBox other, float yThreshold, float xThreshold) {
+
+        return this.intersectsX(other, xThreshold) && this.intersectsY(other, yThreshold);
+    }
+
+
+    public boolean intersectsX(BoundingBox other, float threshold) {
+
+        return this.getX() - threshold <= other.getMaxX() && this.getMaxX() + threshold >= other.getX();
+    }
+
+
+    public boolean intersectsPdf(BoundingBox other) {
+
+        return this.intersectsXPdf(other) && this.intersectsYPdf(other);
+    }
+
+
+    public boolean intersectsPdf(BoundingBox other, float yThreshold, float xThreshold) {
+
+        return this.intersectsXPdf(other, xThreshold) && this.intersectsYPdf(other, yThreshold);
+    }
+
+
+    public boolean intersectsYPdf(BoundingBox other) {
+
+        return this.getPdfMinY() <= other.getPdfMaxY() && this.getPdfMaxY() >= other.getPdfMinY();
+    }
+
+
+    public boolean intersectsY(BoundingBox other) {
+
+        return this.getY() <= other.getMaxY() && this.getMaxY() >= other.getY();
+    }
+
+
+    public boolean intersectsY(BoundingBox other, float threshold) {
+
+        return this.getY() - threshold <= other.getMaxY() && this.getMaxY() + threshold >= other.getY();
+    }
+
+
+    public boolean intersectsYPdf(BoundingBox other, float threshold) {
+
+        return this.getPdfMinY() - threshold <= other.getPdfMaxY() && this.getPdfMaxY() + threshold >= other.getPdfMinY();
+    }
+
+
+    public boolean intersectsXPdf(BoundingBox other) {
+
+        return this.getPdfMinX() <= other.getPdfMaxX() && this.getPdfMaxX() >= other.getPdfMinX();
+    }
+
+
+    public boolean intersectsX(BoundingBox other) {
+
+        return this.getX() <= other.getMaxX() && this.getMaxX() >= other.getMinX();
+    }
+
+
+    public boolean intersectsXPdf(BoundingBox other, float threshold) {
+
+        return this.getPdfMinX() - threshold <= other.getPdfMaxX() && this.getMaxX() + threshold >= other.getPdfMinX();
+    }
+
+
+    public void setToBBoxOfComponents(List<? extends BoundingBox> components) {
+
+        this.bBox = components.stream()
+                .map(BoundingBox::getBBox)
+                .collect(RectangleTransformations.collectBBox());
+        this.bBoxPdf = components.stream()
+                .map(BoundingBox::getBBoxPdf)
+                .collect(RectangleTransformations.collectBBox());
+    }
+
+
+    public double verticalOverlap(BoundingBox other) {
+
+        return Math.max(0, Math.min(this.getPdfMaxY(), other.getPdfMaxY()) - Math.max(this.getPdfMinY(), other.getPdfMinY()));
+    }
+
+
+    public static final Comparator<BoundingBox> ILL_DEFINED_ORDER = (o1, o2) -> {
+
+        if (o1.equals(o2)) {
+            return 0;
+        }
+        if (o1.verticalOverlap(o2) > VERTICAL_COMPARISON_THRESHOLD * ((o1.getHeight() + o2.getHeight()) / 2)) {
+            return Double.compare(o1.getPdfMinX(), o2.getPdfMinX());
+        } else {
+            return Double.compare(o1.getPdfMaxY(), o2.getPdfMaxY());
+        }
+    };
+
+
+    public double horizontalDistance(BoundingBox other) {
+
+        Rectangle2D left;
+        Rectangle2D right;
+        if (this.leftOf(other)) {
+            left = this.getBBox();
+            right = other.getBBox();
+        } else {
+            left = other.getBBox();
+            right = this.getBBox();
+        }
+
+        return Math.max(0, right.getMinX() - left.getMaxX());
+    }
+
+
+    public double verticalDistance(BoundingBox other) {
+
+        Rectangle2D bottom;
+        Rectangle2D top;
+        if (this.isAbove(other)) {
+            top = this.getBBox();
+            bottom = other.getBBox();
+        } else {
+            bottom = this.getBBox();
+            top = other.getBBox();
+        }
+
+        return Math.max(0, bottom.getMinY() - top.getMaxY());
+    }
+
+
+    public boolean rightOf(BoundingBox other) {
+
+        return this.intersectsY(other) && other.getMaxX() <= this.getMinX();
+    }
+
+
+    public boolean leftOf(BoundingBox other) {
+
+        return this.intersectsY(other) && other.getMinX() >= this.getMaxX();
+    }
+
+
+    public boolean isAbove(BoundingBox other) {
+
+        return this.intersectsX(other) && other.getMinY() >= this.getMaxY();
+    }
+
+
+    public boolean isBelow(BoundingBox other) {
+
+        return this.intersectsX(other) && this.getMinY() >= other.getMaxY();
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/docstrum/model/Character.java

@@ -0,0 +1,86 @@
+package com.knecon.fforesight.service.layoutparser.processor.docstrum.model;
+
+import java.util.ArrayList;
+import java.util.Arrays;
+import java.util.List;
+
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.utils.FastAtan2;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.RedTextPosition;
+
+import lombok.Data;
+import lombok.EqualsAndHashCode;
+import lombok.Setter;
+
+@Data
+@EqualsAndHashCode(onlyExplicitlyIncluded = true)
+public class Character {
+
+    @EqualsAndHashCode.Include
+    private final double x;
+    @EqualsAndHashCode.Include
+    private final double y;
+    private final RedTextPosition textPosition;
+
+    @Setter
+    private List<Neighbor> neighbors = new ArrayList<>();
+
+
+    public Character(RedTextPosition chunk) {
+
+        this.x = chunk.getBBoxDirAdj().getCenterX();
+        this.y = chunk.getBBoxDirAdj().getCenterY();
+        this.textPosition = chunk;
+    }
+
+
+    public double getHeight() {
+
+        return textPosition.getHeightDirAdj();
+    }
+
+
+    public double distance(Character character) {
+
+        double dx = getX() - character.getX();
+        double dy = getY() - character.getY();
+        return Math.sqrt(dx * dx + dy * dy);
+    }
+
+
+    public double horizontalDistance(Character character) {
+
+        return Math.abs(getX() - character.getX());
+    }
+
+
+    public double verticalDistance(Character character) {
+
+        return Math.abs(getY() - character.getY());
+    }
+
+
+    public double overlappingDistance(Character other) {
+
+        double[] xs = new double[4];
+        double s = Math.sin(-0);
+        double c = Math.cos(-0);
+        xs[0] = c * x - s * y;
+        xs[1] = c * (x + textPosition.getWidthDirAdj()) - s * (y + textPosition.getHeightDirAdj());
+        xs[2] = c * other.x - s * other.y;
+        xs[3] = c * (other.x + other.textPosition.getWidthDirAdj()) - s * (other.y + other.textPosition.getHeightDirAdj());
+        boolean overlapping = xs[1] >= xs[2] && xs[3] >= xs[0];
+        Arrays.sort(xs);
+        return Math.abs(xs[2] - xs[1]) * (overlapping ? 1 : -1);
+    }
+
+
+    public double angle(Character character) {
+
+        if (getX() > character.getX()) {
+            return FastAtan2.fastAtan2(getY() - character.getY(), getX() - character.getX());
+        } else {
+            return FastAtan2.fastAtan2(character.getY() - getY(), character.getX() - getX());
+        }
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/docstrum/model/ColumnDetector.java

@@ -0,0 +1,324 @@
+package com.knecon.fforesight.service.layoutparser.processor.docstrum.model;
+
+import java.util.ArrayList;
+import java.util.Collections;
+import java.util.Comparator;
+import java.util.HashSet;
+import java.util.LinkedList;
+import java.util.List;
+import java.util.Set;
+
+
+/*
+WIP, mostly working, needs to be tested a bit more
+ */
+public class ColumnDetector {
+
+    public static final double MAX_VALUE_THRESHOLD = 0.5;
+    final static int bins_num = 512;
+    final static int globalStartIdx = 0; // ignore outer parts completely, we don't expect columns there
+    final static int globalEndIdx = bins_num; // i chose 7, since thirds seems a likely split for columns, therefore divided by 6 would eliminate those.
+    public static final double DERIVATIVE_ZERO_THRESHOLD = 1e-10;
+    public static final double MINIMUM_THRESHOLD_FOR_COLUMNS = 0.05;
+    public static final double NEAR_GLOBAL_THRESHOLD = 0.5;
+    double minY;
+    double maxY;
+    double midY;
+    double[] histogram;
+    double min;
+    double max;
+    double resolution;
+    double sum;
+    int N;
+
+
+    public ColumnDetector(double min, double max, double minY, double maxY) {
+
+        this.min = min;
+        this.max = max;
+        this.minY = minY;
+        this.maxY = maxY;
+        this.midY = maxY - minY;
+        this.resolution = (max - min) / bins_num;
+        this.histogram = new double[bins_num];
+    }
+
+
+    public void add(BoundingBox zone) {
+
+        N++;
+        double weight = computeWeight(zone);
+        int start = (int) ((zone.getMinX() - min) / resolution);
+        int end = (int) ((zone.getMaxX() - min) / resolution);
+        for (int i = start; i < end; i++) {
+            histogram[i] += weight;
+            sum += histogram[i];
+        }
+    }
+
+
+    private double computeWeight(BoundingBox zone) {
+
+        double areaWeight = zone.getBBox().getHeight();
+
+        double relativeDistance = relativeDistanceToMiddle(zone.getBBox().getCenterY());
+
+        double distanceWeight;
+        if (relativeDistance < 0.6) {
+            distanceWeight = 1;
+        } else if (relativeDistance < 0.8) {
+            distanceWeight = 0.8;
+        } else {
+            distanceWeight = 0.1;
+        }
+
+        return areaWeight * distanceWeight;
+    }
+
+
+    private double relativeDistanceToMiddle(double y) {
+
+        double range = (maxY - minY) / 2;
+        double mid = minY + range;
+
+        return Math.abs(y - mid) / range;
+    }
+
+
+    public double[] computeDerivative() {
+
+        int length = histogram.length;
+        double[] derivative = new double[length];
+
+        for (int i = 0; i < length; i++) {
+            if (i == 0) {
+                derivative[i] = (histogram[i + 1] - histogram[i]) / resolution;
+            } else if (i == length - 1) {
+                derivative[i] = (histogram[i] - histogram[i - 1]) / resolution;
+            } else {
+                derivative[i] = (histogram[i + 1] - histogram[i - 1]) / (2 * resolution);
+            }
+        }
+
+        return derivative;
+    }
+
+
+    public double calcMean(double[] arr, int start, int end) {
+
+        if (start == end) {
+            return 0;
+        }
+        double sum = 0;
+        for (int i = start; i < end; i++) {
+            sum += arr[i];
+        }
+        return sum / (end - start);
+    }
+
+
+    /*
+    Find columns, by finding all local maxima/minima of the derivative. Filtering them for the ones with the biggest values.
+    For each found minima, we will step to the right until we hit a 0 in the derivative, this indicates a minimum in the main histogram. If this minimum is below a threshold, it is deemed a column divider.
+    Same goes for maxima, but stepping to the left now, since minima in the function will always be to the left of a maximum in its derivative.
+     */
+    public List<Double> determineColumnsWithDerivative(double[] derivative) {
+
+        assert derivative.length == histogram.length;
+
+        Set<Integer> columnIndices = new HashSet<>();
+        double mean = calcMean(histogram, 0, histogram.length);
+        double maxDvValue = calcMax(derivative);
+        double minDvValue = calcMin(derivative);
+
+        if (maxDvValue - minDvValue < mean * MAX_VALUE_THRESHOLD) {
+            Collections.emptyList();
+        }
+
+        Extrema derivativeExtrema = calculateNearGlobalExtrema(derivative, maxDvValue, minDvValue);
+
+        List<Integer> columnsRightOfMinima = findZerosToTheRightOfMinima(derivative, derivativeExtrema.minima(), mean);
+        columnIndices.addAll(columnsRightOfMinima);
+
+        List<Integer> columnsLeftOfMaxima = findZerosToTheLeftOfMaxima(derivative, derivativeExtrema.maxima(), mean);
+        columnIndices.addAll(columnsLeftOfMaxima);
+
+        return columnIndices.stream()
+                .sorted(Comparator.naturalOrder())
+                .map(this::calculateXCoordinateFromIdx)
+                .toList();
+    }
+
+
+    private List<Integer> findZerosToTheLeftOfMaxima(double[] derivative, List<Integer> derivativeMaxima, double mean) {
+
+        List<Integer> columnsLeftOfMaxima = new ArrayList<>();
+
+        for (int i = 0; i < derivativeMaxima.size(); i++) {
+            List<Integer> consecutiveZeroes = new LinkedList<>();
+            boolean maximumFound = false;
+            int maximaIdx = derivativeMaxima.get(i) - 1; // the highest derivative will always be at least one step away from the lowest value.
+            int endIdx = (int) Math.max(globalStartIdx,
+                                        Math.min(maximaIdx - 1,
+                                                 maximaIdx - 0.1 * bins_num)); // search through 10% of array to the right, but at least one step and at most to the left edge;
+
+            for (int j = maximaIdx; j >= endIdx; j--) {
+                if (derivative[j] < DERIVATIVE_ZERO_THRESHOLD) {
+                    maximumFound = true;
+                    consecutiveZeroes.add(j);
+                } else if (maximumFound) {
+                    break;
+                }
+            }
+            if (maximumFound) {
+                int midIdx = consecutiveZeroes.size() / 2;
+                int middleMinimumIdx = consecutiveZeroes.get(midIdx);
+                if (histogram[middleMinimumIdx] < mean * MINIMUM_THRESHOLD_FOR_COLUMNS) {
+                    columnsLeftOfMaxima.add(middleMinimumIdx);
+                }
+            }
+        }
+        return columnsLeftOfMaxima;
+    }
+
+
+    private List<Integer> findZerosToTheRightOfMinima(double[] derivative, List<Integer> derivativeMinima, double mean) {
+
+        List<Integer> columnIndixes = new LinkedList<>();
+        for (int i = 0; i < derivativeMinima.size(); i++) {
+            List<Integer> consecutiveZeroes = new LinkedList<>();
+            boolean minimumFound = false;
+            int minimaIdx = derivativeMinima.get(i) + 1; // the highest derivative will always be at least one step earlier than the lowest value.
+            int endIdx = (int) Math.min(globalEndIdx,
+                                        Math.max(minimaIdx + 1,
+                                                 minimaIdx + 0.1 * bins_num)); // search through 10% of array to the right, but at least one step and at most to the right edge;
+
+            for (int j = minimaIdx; j < endIdx; j++) {
+                if (derivative[j] < DERIVATIVE_ZERO_THRESHOLD) {
+                    minimumFound = true;
+                    consecutiveZeroes.add(j);
+                } else if (minimumFound) {
+                    break;
+                }
+            }
+            if (minimumFound) {
+                int midIdx = consecutiveZeroes.size() / 2;
+                int middleMinimumIdx = consecutiveZeroes.get(midIdx);
+                if (histogram[middleMinimumIdx] < mean * MINIMUM_THRESHOLD_FOR_COLUMNS) {
+                    columnIndixes.add(middleMinimumIdx);
+                }
+            }
+        }
+        return columnIndixes;
+    }
+
+
+    private double calcMax(double[] array) {
+
+        double max = Double.NEGATIVE_INFINITY;
+        for (int i = 0; i < array.length; i++) {
+            if (array[i] > max) {
+                max = array[i];
+            }
+        }
+        return max;
+    }
+
+
+    private double calcMin(double[] array) {
+
+        double min = Double.POSITIVE_INFINITY;
+        for (int i = 0; i < array.length; i++) {
+            if (array[i] < min) {
+                min = array[i];
+            }
+        }
+        return min;
+    }
+
+
+    private Extrema calculateNearGlobalExtrema(double[] derivative, double maxDvValue, double minDvValue) {
+
+        List<Integer> nearGlobalDvMaximaIdx = new LinkedList<>();
+        List<Integer> nearGlobalDvMinimaIdx = new LinkedList<>();
+        for (int i = globalStartIdx; i < globalEndIdx; i++) {
+            if (derivative[i] <= minDvValue * NEAR_GLOBAL_THRESHOLD) {
+                nearGlobalDvMinimaIdx.add(i);
+            }
+            if (derivative[i] >= maxDvValue * NEAR_GLOBAL_THRESHOLD) {
+                nearGlobalDvMaximaIdx.add(i);
+            }
+        }
+
+        nearGlobalDvMinimaIdx = removeConsecutive(nearGlobalDvMinimaIdx);
+        nearGlobalDvMaximaIdx = removeConsecutive(nearGlobalDvMaximaIdx);
+
+        return new Extrema(nearGlobalDvMaximaIdx, nearGlobalDvMinimaIdx);
+    }
+
+
+    private record Extrema(List<Integer> maxima, List<Integer> minima) {
+
+    }
+
+
+    private Double calculateXCoordinateFromIdx(int globalMinIdx) {
+
+        return min + ((globalMinIdx + 1) * resolution);
+    }
+
+
+    public static List<Integer> removeConsecutive(List<Integer> numbers) {
+
+        List<Integer> result = new ArrayList<>();
+        if (numbers == null || numbers.isEmpty()) {
+            return result;
+        }
+
+        result.add(numbers.get(0)); // Add the first number
+
+        for (int i = 1; i < numbers.size(); i++) {
+            if (numbers.get(i) != numbers.get(i - 1) + 1) {
+                result.add(numbers.get(i)); // Add non-consecutive numbers
+            }
+        }
+
+        return result;
+    }
+
+
+    public void kernelSmooth(double[] kernel) {
+
+        double[] newFrequencies = new double[histogram.length];
+        int shift = (kernel.length - 1) / 2;
+        for (int i = 0; i < kernel.length; i++) {
+            int jStart = Math.max(0, i - shift);
+            int jEnd = Math.min(histogram.length, histogram.length + i - shift);
+            for (int j = jStart; j < jEnd; j++) {
+                newFrequencies[j - i + shift] += kernel[i] * histogram[j];
+            }
+        }
+        histogram = newFrequencies;
+    }
+
+
+    public double[] createGaussianKernel(int length, double stdDeviation) {
+
+        int r = length / 2;
+
+        int size = 2 * r + 1;
+        double[] kernel = new double[size];
+        double sum = 0;
+        double b = 2 * (stdDeviation) * (stdDeviation);
+        double a = 1 / Math.sqrt(Math.PI * b);
+        for (int i = 0; i < size; i++) {
+            kernel[i] = a * Math.exp(-(i - r) * (i - r) / b);
+            sum += kernel[i];
+        }
+        for (int i = 0; i < size; i++) {
+            kernel[i] /= sum;
+        }
+        return kernel;
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/docstrum/model/Histogram.java

@@ -0,0 +1,90 @@
+package com.knecon.fforesight.service.layoutparser.processor.docstrum.model;
+
+public class Histogram {
+
+    private static final double EPSILON = 1.0e-6;
+    private final double min;
+    private final double resolution;
+    private double[] frequencies;
+
+
+    public Histogram(double minValue, double maxValue, double resolution) {
+
+        this.min = minValue - EPSILON;
+        double delta = maxValue - minValue + 2 * EPSILON;
+        int size = Math.max(1, (int) Math.round((maxValue - minValue) / resolution));
+        this.resolution = delta / size;
+        this.frequencies = new double[size];
+    }
+
+
+    public void kernelSmooth(double[] kernel) {
+
+        double[] newFrequencies = new double[frequencies.length];
+        int shift = (kernel.length - 1) / 2;
+        for (int i = 0; i < kernel.length; i++) {
+            int jStart = Math.max(0, i - shift);
+            int jEnd = Math.min(frequencies.length, frequencies.length + i - shift);
+            for (int j = jStart; j < jEnd; j++) {
+                newFrequencies[j - i + shift] += kernel[i] * frequencies[j];
+            }
+        }
+        frequencies = newFrequencies;
+    }
+
+
+    public double[] createGaussianKernel(double length, double stdDeviation) {
+
+        int r = (int) Math.round(length / resolution) / 2;
+
+        int size = 2 * r + 1;
+        double[] kernel = new double[size];
+        double sum = 0;
+        double b = 2 * (stdDeviation / resolution) * (stdDeviation / resolution);
+        double a = 1 / Math.sqrt(Math.PI * b);
+        for (int i = 0; i < size; i++) {
+            kernel[i] = a * Math.exp(-(i - r) * (i - r) / b);
+            sum += kernel[i];
+        }
+        for (int i = 0; i < size; i++) {
+            kernel[i] /= sum;
+        }
+        return kernel;
+    }
+
+
+    public void gaussianSmooth(double windowLength, double stdDeviation) {
+
+        kernelSmooth(createGaussianKernel(windowLength, stdDeviation));
+    }
+
+
+    public void add(double value) {
+
+        frequencies[(int) ((value - min) / resolution)] += 1.0;
+    }
+
+
+    public int getSize() {
+
+        return frequencies.length;
+    }
+
+
+    public double getPeakValue() {
+
+        int peakIndex = 0;
+        for (int i = 1; i < frequencies.length; i++) {
+            if (frequencies[i] > frequencies[peakIndex]) {
+                peakIndex = i;
+            }
+        }
+        int peakEndIndex = peakIndex + 1;
+        final double EPS = 0.0001;
+        while (peakEndIndex < frequencies.length && Math.abs(frequencies[peakEndIndex] - frequencies[peakIndex]) < EPS) {
+            peakEndIndex++;
+        }
+        return ((double) peakIndex + peakEndIndex) / 2 * resolution + min;
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/docstrum/model/Line.java

@@ -0,0 +1,195 @@
+package com.knecon.fforesight.service.layoutparser.processor.docstrum.model;
+
+import static com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence.BOLD;
+import static com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence.BOLD_ITALIC;
+import static com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence.ITALIC;
+import static com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence.STANDARD;
+
+import java.util.ArrayList;
+import java.util.Arrays;
+import java.util.Comparator;
+import java.util.EnumMap;
+import java.util.List;
+import java.util.Map;
+import java.util.concurrent.atomic.AtomicInteger;
+
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.utils.FastAtan2;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.FontStyle;
+
+import lombok.Data;
+import lombok.EqualsAndHashCode;
+
+@Data
+@EqualsAndHashCode(onlyExplicitlyIncluded = true, callSuper = false)
+public class Line extends TextBoundingBox {
+
+    private static final double WORD_DISTANCE_MULTIPLIER = 0.18;
+
+    @EqualsAndHashCode.Include
+    private final double x0;
+    @EqualsAndHashCode.Include
+    private final double y0;
+
+    @EqualsAndHashCode.Include
+    private final double x1;
+    @EqualsAndHashCode.Include
+    private final double y1;
+
+    private final double height;
+
+    private FontStyle fontStyle;
+
+    private final List<Character> characters;
+    private final List<TextPositionSequence> words = new ArrayList<>();
+
+
+    public Line(List<Character> characters, double wordSpacing) {
+
+        this.characters = characters;
+
+        if (characters.size() >= 2) {
+            // linear regression
+            double sx = 0.0;
+            double sxx = 0.0;
+            double sxy = 0.0;
+            double sy = 0.0;
+            for (Character character : characters) {
+                sx += character.getX();
+                sxx += character.getX() * character.getX();
+                sxy += character.getX() * character.getY();
+                sy += character.getY();
+            }
+            double b = (characters.size() * sxy - sx * sy) / (characters.size() * sxx - sx * sx);
+            double a = (sy - b * sx) / characters.size();
+
+            this.x0 = characters.get(0).getX();
+            this.y0 = a + b * this.x0;
+            this.x1 = characters.get(characters.size() - 1).getX();
+            this.y1 = a + b * this.x1;
+        } else {
+            Character character = characters.get(0);
+            double dx = character.getTextPosition().getWidthDirAdj() / 3;
+            double dy = dx * Math.tan(0);
+            this.x0 = character.getX() - dx;
+            this.x1 = character.getX() + dx;
+            this.y0 = character.getY() - dy;
+            this.y1 = character.getY() + dy;
+        }
+        height = computeHeight();
+        computeWords(wordSpacing * WORD_DISTANCE_MULTIPLIER);
+        buildBBox();
+        computeFontStyle();
+    }
+
+
+    private void computeFontStyle() {
+
+        EnumMap<FontStyle, AtomicInteger> fontStyleCounter = new EnumMap<>(FontStyle.class);
+        for (FontStyle fontStyle : FontStyle.values()) {
+            fontStyleCounter.put(fontStyle, new AtomicInteger(0));
+        }
+        for (TextPositionSequence word : words) {
+            switch (word.getFontStyle()) {
+                case STANDARD -> fontStyleCounter.get(FontStyle.REGULAR).getAndIncrement();
+                case BOLD -> fontStyleCounter.get(FontStyle.BOLD).getAndIncrement();
+                case ITALIC -> fontStyleCounter.get(FontStyle.ITALIC).getAndIncrement();
+                case BOLD_ITALIC -> fontStyleCounter.get(FontStyle.BOLD_ITALIC).getAndIncrement();
+            }
+        }
+        fontStyle = fontStyleCounter.entrySet()
+                .stream()
+                .max(Comparator.comparing(entry -> entry.getValue().get()))
+                .map(Map.Entry::getKey)
+                .orElse(FontStyle.REGULAR);
+    }
+
+
+    public double getAngle() {
+
+        return FastAtan2.fastAtan2(y1 - y0, x1 - x0);
+    }
+
+
+    public double getLength() {
+
+        return Math.sqrt((x0 - x1) * (x0 - x1) + (y0 - y1) * (y0 - y1));
+    }
+
+
+    private double computeHeight() {
+
+        return characters.stream()
+                       .map(Character::getHeight)
+                       .reduce(0d, Double::sum) / characters.size();
+    }
+
+
+    public double angularDifference(Line j) {
+
+        double diff = Math.abs(getAngle() - j.getAngle());
+        if (diff <= Math.PI / 2) {
+            return diff;
+        } else {
+            return Math.PI - diff;
+        }
+    }
+
+
+    public double horizontalDistance(Line other) {
+
+        double[] xs = new double[4];
+        xs[0] = x0;
+        xs[1] = x1;
+        xs[2] = other.x0;
+        xs[3] = other.x1;
+        boolean overlapping = xs[1] >= xs[2] && xs[3] >= xs[0];
+        Arrays.sort(xs);
+        return Math.abs(xs[2] - xs[1]) * (overlapping ? 1 : -1);
+    }
+
+
+    public double verticalDistance(Line other) {
+
+        double ym = (y0 + y1) / 2;
+        double yn = (other.y0 + other.y1) / 2;
+        return Math.abs(ym - yn);
+    }
+
+
+    private void computeWords(double wordSpacing) {
+
+        TextPositionSequence word = new TextPositionSequence();
+        Character previous = null;
+        for (Character current : characters) {
+            if (previous != null) {
+                double dist = current.getTextPosition().getXDirAdj() - previous.getTextPosition().getXDirAdj() - previous.getTextPosition().getWidthDirAdj();
+                if (dist > wordSpacing) {
+                    words.add(word);
+                    word = new TextPositionSequence();
+                }
+            }
+            word.getTextPositions().add(current.getTextPosition());
+            previous = current;
+        }
+        words.add(word);
+    }
+
+
+    private void buildBBox() {
+
+        this.setToBBoxOfComponents(characters.stream()
+                                           .map(Character::getTextPosition)
+                                           .toList());
+    }
+
+
+    public String toString() {
+
+        StringBuilder sb = new StringBuilder();
+        words.forEach(word -> sb.append(word.toString()).append(" "));
+        return sb.toString().trim();
+    }
+
+}
+

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/docstrum/model/Neighbor.java

@@ -0,0 +1,43 @@
+package com.knecon.fforesight.service.layoutparser.processor.docstrum.model;
+
+import lombok.Getter;
+
+public class Neighbor {
+
+    @Getter
+    private final double distance;
+    private Double angle;
+    private final Character originCharacter;
+    @Getter
+    private final Character character;
+
+
+    public Neighbor(Character neighbor, Character origin) {
+
+        this.distance = neighbor.distance(origin);
+        this.character = neighbor;
+        this.originCharacter = origin;
+    }
+
+
+    public double getHorizontalDistance() {
+
+        return character.horizontalDistance(originCharacter);
+    }
+
+
+    public double getVerticalDistance() {
+
+        return character.verticalDistance(originCharacter);
+    }
+
+
+    public double getAngle() {
+
+        if (angle != null) {
+            return angle;
+        }
+        return this.character.angle(this.originCharacter);
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/docstrum/model/TextBoundingBox.java

@@ -0,0 +1,102 @@
+package com.knecon.fforesight.service.layoutparser.processor.docstrum.model;
+
+import java.awt.geom.Rectangle2D;
+import java.util.List;
+import java.util.Set;
+import java.util.stream.Collectors;
+
+import com.knecon.fforesight.service.layoutparser.processor.model.text.TextDirection;
+import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
+
+import lombok.EqualsAndHashCode;
+import lombok.Getter;
+import lombok.NoArgsConstructor;
+import lombok.Setter;
+import lombok.experimental.SuperBuilder;
+
+@Getter
+@Setter
+@SuperBuilder
+@NoArgsConstructor
+@EqualsAndHashCode(callSuper = false)
+public abstract class TextBoundingBox extends BoundingBox {
+
+    protected Rectangle2D bBoxDirAdj;
+
+    protected TextDirection dir;
+
+
+    @Override
+    public void setToBBoxOfComponents(List<? extends BoundingBox> components) {
+
+        super.setToBBoxOfComponents(components);
+        this.bBoxDirAdj = components.stream()
+                .filter(c -> c instanceof TextBoundingBox)
+                .map(c -> (TextBoundingBox) c)
+                .map(TextBoundingBox::getBBoxDirAdj)
+                .collect(RectangleTransformations.collectBBox());
+
+        Set<TextDirection> textDirections = components.stream()
+                .filter(c -> c instanceof TextBoundingBox)
+                .map(c -> (TextBoundingBox) c)
+                .map(TextBoundingBox::getDir)
+                .collect(Collectors.toSet());
+
+        if (textDirections.isEmpty()) {
+            dir = TextDirection.ZERO;
+        } else if (textDirections.size() > 1) {
+            throw new IllegalArgumentException("More than one text direction found");
+        } else {
+            dir = textDirections.iterator().next();
+        }
+    }
+
+
+    public double getXDirAdj() {
+
+        return this.bBoxDirAdj.getX();
+    }
+
+
+    public double getYDirAdj() {
+
+        return this.bBoxDirAdj.getY();
+    }
+
+
+    public double getWidthDirAdj() {
+
+        return this.bBoxDirAdj.getWidth();
+    }
+
+
+    public double getHeightDirAdj() {
+
+        return this.bBoxDirAdj.getHeight();
+    }
+
+
+    public double getMaxXDirAdj() {
+
+        return this.bBoxDirAdj.getMaxX();
+    }
+
+
+    public double getMaxYDirAdj() {
+
+        return this.bBoxDirAdj.getMaxY();
+    }
+
+
+    public double getCenterYDirAdj() {
+
+        return this.bBoxDirAdj.getCenterY();
+    }
+
+
+    public double getCenterXDirAdj() {
+
+        return this.bBoxDirAdj.getCenterX();
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/docstrum/model/UnionFind.java

@@ -0,0 +1,31 @@
+package com.knecon.fforesight.service.layoutparser.processor.docstrum.model;
+
+import java.util.Collection;
+import java.util.LinkedHashMap;
+import java.util.LinkedHashSet;
+import java.util.Map;
+import java.util.Set;
+
+public class UnionFind<T> extends org.jgrapht.alg.util.UnionFind<T> {
+
+    public UnionFind(Set<T> elements) {
+
+        super(elements);
+    }
+
+
+    public Collection<Set<T>> getGroups() {
+
+        Map<T, Set<T>> setRep = new LinkedHashMap<>();
+        for (T t : getParentMap().keySet()) {
+            T representative = find(t);
+            if (!setRep.containsKey(representative)) {
+                setRep.put(representative, new LinkedHashSet<>());
+            }
+            setRep.get(representative).add(t);
+        }
+
+        return setRep.values();
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/docstrum/model/Zone.java

@@ -0,0 +1,34 @@
+package com.knecon.fforesight.service.layoutparser.processor.docstrum.model;
+
+import java.util.Comparator;
+import java.util.List;
+
+import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
+
+import lombok.Data;
+import lombok.EqualsAndHashCode;
+
+@Data
+@EqualsAndHashCode(callSuper = false)
+public class Zone extends TextBoundingBox {
+
+    private List<Line> lines;
+
+
+    @SuppressWarnings("PMD.ConstructorCallsOverridableMethod")
+    public Zone(List<Line> lines) {
+
+        lines.sort(Comparator.comparingDouble(Line::getY0));
+        this.lines = lines;
+        setToBBoxOfComponents(lines);
+    }
+
+
+    public String toString() {
+
+        StringBuilder sb = new StringBuilder();
+        lines.forEach(line -> sb.append(line.toString()).append("\n"));
+        return sb.toString().trim();
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/docstrum/service/LineBuilderService.java

@@ -0,0 +1,58 @@
+package com.knecon.fforesight.service.layoutparser.processor.docstrum.service;
+
+import java.util.Comparator;
+import java.util.HashSet;
+import java.util.List;
+
+import org.springframework.stereotype.Service;
+
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.AngleFilter;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Character;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Line;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.UnionFind;
+import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRulings;
+
+@Service
+public class LineBuilderService {
+
+    private static final double CHARACTER_SPACING_DISTANCE_MULTIPLIER = 3.5;
+    private static final double LINE_SPACING_THRESHOLD_MULTIPLIER = 0.67;
+    private static final double ANGLE_TOLERANCE = Math.PI / 6;
+
+
+    public List<Line> buildLines(List<Character> characters, double characterSpacing, double lineSpacing, CleanRulings rulings) {
+
+        double maxHorizontalDistance = characterSpacing * CHARACTER_SPACING_DISTANCE_MULTIPLIER;
+        double maxVerticalDistance = lineSpacing * LINE_SPACING_THRESHOLD_MULTIPLIER;
+
+        UnionFind<Character> unionFind = new UnionFind<>(new HashSet<>(characters));
+
+        AngleFilter angleFilter = new AngleFilter(-ANGLE_TOLERANCE, ANGLE_TOLERANCE);
+
+        characters.forEach(character -> {
+            character.getNeighbors()
+                    .forEach(neighbor -> {
+                        double normalizedHorizontalDistance = neighbor.getHorizontalDistance() / maxHorizontalDistance;
+                        double normalizedVerticalDistance = neighbor.getVerticalDistance() / maxVerticalDistance;
+
+                        if (character.getTextPosition().getDir() != neighbor.getCharacter().getTextPosition().getDir() //
+                            || !angleFilter.matches(neighbor) //
+                            || Math.pow(normalizedHorizontalDistance, 2) + Math.pow(normalizedVerticalDistance, 2) > 1 //
+                            || rulings.lineBetween(character.getTextPosition(), neighbor.getCharacter().getTextPosition())) {
+                            return;
+                        }
+
+                        unionFind.union(character, neighbor.getCharacter());
+                    });
+        });
+
+        return unionFind.getGroups()
+                .stream()
+                .map(lineCharacters -> lineCharacters.stream()
+                        .sorted(Comparator.comparingDouble(Character::getX))
+                        .toList())
+                .map(lineCharacters -> new Line(lineCharacters, characterSpacing))
+                .toList();
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/docstrum/service/NearestNeighbourService.java

@@ -0,0 +1,36 @@
+package com.knecon.fforesight.service.layoutparser.processor.docstrum.service;
+
+import java.util.List;
+
+import org.springframework.stereotype.Service;
+import org.tinspin.index.Index;
+import org.tinspin.index.kdtree.KDTree;
+
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Character;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Neighbor;
+
+@Service
+public class NearestNeighbourService {
+
+    private static final int NUMBER_OF_NEIGHBOURS = 8;
+
+
+    public void findNearestNeighbors(List<Character> characters) {
+        KDTree<Character> kdTree = KDTree.create(2);
+        characters.forEach(c -> kdTree.insert(new double[]{c.getX(), c.getY()}, c));
+
+        for(Character c : characters) {
+            Index.PointIteratorKnn<Character> iterator = kdTree.queryKnn(new double[]{c.getX(), c.getY()}, NUMBER_OF_NEIGHBOURS + 1);
+            // skip the first as this is identity
+            if(iterator.hasNext()) {
+                iterator.next();
+            }
+            while(iterator.hasNext()) {
+                c.getNeighbors().add(new Neighbor(iterator.next().value(), c));
+            }
+        }
+
+    }
+
+}
+

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/docstrum/service/ReadingOrderService.java

@@ -0,0 +1,192 @@
+package com.knecon.fforesight.service.layoutparser.processor.docstrum.service;
+
+import java.awt.geom.Rectangle2D;
+import java.util.ArrayList;
+import java.util.Comparator;
+import java.util.HashMap;
+import java.util.List;
+import java.util.ListIterator;
+import java.util.Map;
+import java.util.stream.Collectors;
+
+import org.springframework.stereotype.Service;
+
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.TextBoundingBox;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Zone;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.utils.DoubleUtils;
+
+@Service
+public class ReadingOrderService {
+
+    private static final double THRESHOLD = 5;
+    public static final double MULTI_COLUMN_DETECTION_THRESHOLD = 1.5;
+
+    private static final Comparator<TextBoundingBox> COMPARATOR = //
+            Comparator.comparing(TextBoundingBox::getY, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD))
+                    .thenComparing(TextBoundingBox::getX, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD));
+
+    private static final Comparator<TextBoundingBox> COMPARATOR_DIR_ADJ = //
+            Comparator.comparing(TextBoundingBox::getYDirAdj, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD))
+                    .thenComparing(TextBoundingBox::getXDirAdj, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD));
+
+
+    public List<Zone> resolve(List<Zone> zones, boolean xyReadingOrder, boolean useDirAdjCoords) {
+
+        if (zones.isEmpty() || zones.size() == 1) {
+            return zones;
+        }
+
+        if (xyReadingOrder) {
+            return resolveSingleColumnReadingOrder(zones, useDirAdjCoords);
+        }
+
+        Map<Long, Integer> histogram = new HashMap<>();
+        for (Zone zone : zones) {
+            Rectangle2D bbox = useDirAdjCoords ? zone.getBBoxDirAdj() : zone.getBBox();
+            long minY = Math.round(bbox.getMinY());
+            long maxY = Math.round(bbox.getMaxY());
+            for (long i = minY; i <= maxY; i++) {
+                histogram.put(i, histogram.getOrDefault(i, 0) + 1);
+            }
+        }
+
+        if (histogram.values()
+                    .stream()
+                    .mapToInt(Integer::intValue).average()
+                    .orElse(1) < MULTI_COLUMN_DETECTION_THRESHOLD) {
+            return resolveSingleColumnReadingOrder(zones, useDirAdjCoords);
+        } else {
+
+            return resolveMultiColumnReadingOder(zones, useDirAdjCoords);
+        }
+
+    }
+
+
+    private static List<Zone> resolveSingleColumnReadingOrder(List<Zone> zones, boolean useDirAdjCoords) {
+
+        if (useDirAdjCoords) {
+            return zones.stream()
+                    .collect(Collectors.groupingBy(TextBoundingBox::getDir)).values()
+                    .stream()
+                    .flatMap(words -> words.stream()
+                            .sorted(COMPARATOR_DIR_ADJ))
+                    .toList();
+        }
+
+        zones.sort(COMPARATOR);
+        return zones;
+    }
+
+
+    private List<Zone> resolveMultiColumnReadingOder(List<Zone> zones, boolean useDirAdjCoords) {
+
+        // Simple reading order resolver for multi column page layout as described here : https://pub.towardsai.net/advanced-rag-02-unveiling-pdf-parsing-b84ae866344e
+        // TODO implement a more fancy reading order resolver see https://github.com/BobLd/DocumentLayoutAnalysis/blob/master/README.md#reading-order
+
+        double minX = Double.POSITIVE_INFINITY;
+        double maxX = Double.NEGATIVE_INFINITY;
+
+        for (Zone zone : zones) {
+            Rectangle2D bbox = useDirAdjCoords ? zone.getBBoxDirAdj() : zone.getBBox();
+            if (bbox.getX() < minX) {
+                minX = zone.getXDirAdj();
+            }
+            if (bbox.getMaxX() > maxX) {
+                maxX = zone.getMaxXDirAdj();
+            }
+        }
+
+        double midLineXCoordinate = (minX + maxX) / 2;
+
+        List<Zone> leftOf = new ArrayList<>();
+        List<Zone> rightOf = new ArrayList<>();
+        List<Zone> middle = new ArrayList<>();
+
+        for (Zone zone : zones) {
+            Rectangle2D bbox = useDirAdjCoords ? zone.getBBoxDirAdj() : zone.getBBox();
+            if (bbox.getX() < midLineXCoordinate && bbox.getX() + bbox.getWidth() < midLineXCoordinate) {
+                leftOf.add(zone);
+            } else if (bbox.getX() > midLineXCoordinate && bbox.getX() + bbox.getWidth() > midLineXCoordinate) {
+                rightOf.add(zone);
+            } else {
+                middle.add(zone);
+            }
+        }
+
+        if (useDirAdjCoords) {
+            leftOf.sort(COMPARATOR_DIR_ADJ);
+            rightOf.sort(COMPARATOR_DIR_ADJ);
+            middle.sort(COMPARATOR_DIR_ADJ);
+        } else {
+            leftOf.sort(COMPARATOR);
+            rightOf.sort(COMPARATOR);
+            middle.sort(COMPARATOR);
+        }
+/*
+        List<Zone> leftNotIntersecting = new ArrayList<>();
+        for (Zone leftZone : leftOf) {
+            boolean intersects = false;
+            for (Zone rightZone : rightOf) {
+                if (leftZone.intersectsY(rightZone)) {
+                    intersects = true;
+                    break;
+                }
+                // early stopping
+                if (rightZone.getBBox().getMinY() > leftZone.getBBox().getMaxY()) {
+                    break;
+                }
+            }
+            if (!intersects) {
+                leftNotIntersecting.add(leftZone);
+            }
+        }
+
+        List<Zone> rightNotIntersecting = new ArrayList<>();
+        for (Zone rightZone : rightOf) {
+            boolean intersects = false;
+            for (Zone leftZone : leftOf) {
+                if (rightZone.intersectsY(leftZone)) {
+                    intersects = true;
+                    break;
+                }
+                // early stopping
+                if (leftZone.getBBox().getMinY() > rightZone.getBBox().getMaxY()) {
+                    break;
+                }
+            }
+            if (!intersects) {
+                rightNotIntersecting.add(rightZone);
+            }
+        }
+
+        leftOf.removeAll(leftNotIntersecting);
+        rightOf.removeAll(rightNotIntersecting);
+
+        middle.addAll(leftNotIntersecting);
+        middle.addAll(rightNotIntersecting);
+*/
+        List<Zone> sortedZones = new ArrayList<>();
+        sortedZones.addAll(leftOf);
+        sortedZones.addAll(rightOf);
+
+        ListIterator<Zone> itty = middle.listIterator();
+
+        while (itty.hasNext()) {
+            Zone current = itty.next();
+            Rectangle2D bbox = useDirAdjCoords ? current.getBBoxDirAdj() : current.getBBox();
+            for (int i = 0; i < sortedZones.size(); i++) {
+                if (bbox.getY() < sortedZones.get(i).getY()) {
+                    sortedZones.add(i, current);
+                    itty.remove();
+                    break;
+                }
+            }
+        }
+
+        sortedZones.addAll(middle);
+
+        return sortedZones;
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/docstrum/service/SpacingService.java

@@ -0,0 +1,56 @@
+package com.knecon.fforesight.service.layoutparser.processor.docstrum.service;
+
+import java.util.List;
+
+import org.springframework.stereotype.Service;
+
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.AngleFilter;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Character;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Histogram;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Neighbor;
+
+@Service
+public class SpacingService {
+
+    private static final double SPACING_HISTOGRAM_RESOLUTION = 0.5;
+    private static final double SPACING_HISTOGRAM_SMOOTHING_LENGTH = 2.5;
+    private static final double SPACING_HIST_SMOOTHING_STANDARD_DEVIATION = 0.5;
+    private static final double ANGLE_TOLERANCE = Math.PI / 6;
+
+
+    public double computeCharacterSpacing(List<Character> characters) {
+
+        return computeSpacing(characters, 0);
+    }
+
+
+    public double computeLineSpacing(List<Character> characters) {
+
+        return computeSpacing(characters, Math.PI / 2);
+    }
+
+
+    private double computeSpacing(List<Character> characters, double angle) {
+
+        double maxDistance = Double.NEGATIVE_INFINITY;
+
+        for (Character character : characters) {
+            for (Neighbor neighbor : character.getNeighbors()) {
+                maxDistance = Math.max(maxDistance, neighbor.getDistance());
+            }
+        }
+        Histogram histogram = new Histogram(0, maxDistance, SPACING_HISTOGRAM_RESOLUTION);
+        AngleFilter angleFilter = new AngleFilter(angle - ANGLE_TOLERANCE, angle + ANGLE_TOLERANCE);
+        for (Character character : characters) {
+            for (Neighbor neighbor : character.getNeighbors()) {
+                if (angleFilter.matches(neighbor)) {
+                    histogram.add(neighbor.getDistance());
+                }
+            }
+        }
+
+        histogram.gaussianSmooth(SPACING_HISTOGRAM_SMOOTHING_LENGTH, SPACING_HIST_SMOOTHING_STANDARD_DEVIATION);
+        return histogram.getPeakValue();
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/docstrum/service/ZoneBuilderService.java

@@ -0,0 +1,177 @@
+package com.knecon.fforesight.service.layoutparser.processor.docstrum.service;
+
+import static com.knecon.fforesight.service.layoutparser.processor.model.SectionIdentifier.numericalIdentifierPattern;
+
+import java.util.ArrayList;
+import java.util.Comparator;
+import java.util.HashSet;
+import java.util.List;
+import java.util.Set;
+import java.util.stream.Collectors;
+
+import org.springframework.stereotype.Service;
+
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Character;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Line;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.UnionFind;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Zone;
+import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRulings;
+
+@Service
+public class ZoneBuilderService {
+
+    private static final double MIN_HORIZONTAL_DISTANCE_MULTIPLIER = -0.5;
+    private static final double MAX_VERTICAL_DISTANCE_MULTIPLIER = 1.2;
+
+    private static final double MIN_HORIZONTAL_MERGE_DISTANCE_MULTIPLIER = -7;
+
+    private static final double MAX_VERTICAL_MERGE_DISTANCE_MULTIPLIER = 0.5;
+
+    private static final double MIN_LINE_SIZE_SCALE = 0.9;
+
+    private static final double MAX_LINE_SIZE_SCALE = 2.5;
+
+    private static final double ANGLE_TOLERANCE = Math.PI / 6;
+
+    private static final double MAX_VERTICAL_MERGE_DISTANCE = 0.5;
+
+
+    public List<Zone> buildZones(List<Line> lines, double characterSpacing, double lineSpacing, CleanRulings rulings) {
+
+        double minHorizontalDistance = characterSpacing * MIN_HORIZONTAL_DISTANCE_MULTIPLIER;
+        double maxVerticalDistance = lineSpacing * MAX_VERTICAL_DISTANCE_MULTIPLIER;
+        double minHorizontalMergeDistance = lineSpacing * MIN_HORIZONTAL_MERGE_DISTANCE_MULTIPLIER;
+        double maxVerticalMergeDistance = lineSpacing * MAX_VERTICAL_MERGE_DISTANCE_MULTIPLIER;
+
+        UnionFind<Line> unionFind = new UnionFind<>(new HashSet<>(lines));
+
+        double meanHeight = calculateMeanHeight(lines);
+
+        lines.forEach(outerLine -> {
+            lines.forEach(innerLine -> {
+
+                if (innerLine == outerLine //
+                    || unionFind.inSameSet(outerLine, innerLine)//
+                    || outerLine.angularDifference(innerLine) > ANGLE_TOLERANCE) {
+                    return;
+                }
+
+//                if (!innerLine.getFontStyle().equals(outerLine.getFontStyle()) //
+//                    && !outerLine.intersectsY(innerLine, -2f)) {
+//                    return;
+//                }
+
+                double horizontalScale = Math.min(outerLine.getHeightDirAdj(), innerLine.getHeightDirAdj()) / meanHeight;
+                horizontalScale = Math.max(MIN_LINE_SIZE_SCALE, Math.min(horizontalScale, MAX_LINE_SIZE_SCALE));
+                double verticalScale = horizontalScale;
+
+//                if (innerLine.toString().endsWith(":")
+//                    || outerLine.toString().endsWith(":")
+//                    || numericalIdentifierPattern.matcher(innerLine.toString()).matches()
+//                    || numericalIdentifierPattern.matcher(outerLine.toString()).matches()) {
+//
+//                    horizontalScale *= 5;
+//                    verticalScale /= 10;
+//                }
+
+                double horizontalDistance = outerLine.horizontalDistance(innerLine) / horizontalScale;
+                double verticalDistance = outerLine.verticalDistance(innerLine) / verticalScale;
+
+                if ((!(minHorizontalDistance <= horizontalDistance) || !(verticalDistance <= maxVerticalDistance)) //
+                    && (!(minHorizontalMergeDistance <= horizontalDistance) || !(verticalDistance <= maxVerticalMergeDistance))) {
+                    return;
+                }
+
+                if (rulings.lineBetween(outerLine, innerLine)) {
+                    return;
+                }
+
+                unionFind.union(outerLine, innerLine);
+
+            });
+        });
+
+        return unionFind.getGroups()
+                .stream()
+                .map(group -> mergeLinesInZone(new ArrayList<>(group), characterSpacing, lineSpacing))
+                .toList();
+    }
+
+
+    private double calculateMeanHeight(List<Line> lines) {
+
+        double meanHeight = 0.0;
+        double weights = 0.0;
+        for (Line line : lines) {
+            double weight = line.getLength();
+            meanHeight += line.getHeightDirAdj() * weight;
+            weights += weight;
+        }
+        meanHeight /= weights;
+        return meanHeight;
+    }
+
+
+    private Zone mergeLinesInZone(List<Line> lines, double characterSpacing, double lineSpacing) {
+
+        double maxHorizontalDistance = 0;
+        double minVerticalDistance = 0;
+        double maxVerticalDistance = lineSpacing * MAX_VERTICAL_MERGE_DISTANCE;
+
+        UnionFind<Line> unionFind = new UnionFind<>(new HashSet<>(lines));
+
+        lines.forEach(outer -> {
+            lines.forEach(inner -> {
+                if (inner == outer) {
+                    return;
+                }
+
+                double horizontalDistance = outer.horizontalDistance(inner);
+                double verticalDistance = outer.verticalDistance(inner);
+
+                if (horizontalDistance <= maxHorizontalDistance && minVerticalDistance <= verticalDistance && verticalDistance <= maxVerticalDistance) {
+
+                    unionFind.union(outer, inner);
+
+                } else if (minVerticalDistance <= verticalDistance
+                           && verticalDistance <= maxVerticalDistance
+                           && Math.abs(horizontalDistance - Math.min(outer.getLength(), inner.getLength())) < 0.1) {
+
+                    boolean characterOverlap = false;
+                    int overlappingCount = 0;
+                    for (Character outerCharacter : outer.getCharacters()) {
+                        for (Character innerCharacter : inner.getCharacters()) {
+                            double characterOverlapDistance = outerCharacter.overlappingDistance(innerCharacter);
+                            if (characterOverlapDistance > 2) {
+                                characterOverlap = true;
+                            }
+                            if (characterOverlapDistance > 0) {
+                                overlappingCount++;
+                            }
+                        }
+                    }
+                    if (!characterOverlap && overlappingCount <= 2) {
+                        unionFind.union(outer, inner);
+                    }
+                }
+
+            });
+        });
+
+        List<Line> outputZone = new ArrayList<>();
+        for (Set<Line> group : unionFind.getGroups()) {
+            List<Character> characters = new ArrayList<>();
+            for (Line line : group) {
+                characters.addAll(line.getCharacters());
+            }
+            characters.sort(Comparator.comparingDouble(Character::getX));
+
+            outputZone.add(new Line(characters, characterSpacing));
+        }
+
+        return new Zone(outputZone.stream()
+                                .sorted(Comparator.comparing(Line::getY0))
+                                .collect(Collectors.toList()));
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/docstrum/utils/DoubleUtils.java

@@ -0,0 +1,18 @@
+package com.knecon.fforesight.service.layoutparser.processor.docstrum.utils;
+
+public class DoubleUtils {
+
+    public static int compareDouble(double d1, double d2, double precision) {
+
+        if (Double.isNaN(d1) || Double.isNaN(d2)) {
+            return Double.compare(d1, d2);
+        }
+
+        if (Math.abs(d1 - d2) < precision) {
+            return 0;
+        }
+
+        return Double.compare(d1, d2);
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/docstrum/utils/FastAtan2.java

@@ -0,0 +1,76 @@
+package com.knecon.fforesight.service.layoutparser.processor.docstrum.utils;
+
+public class FastAtan2 {
+
+    static final private int Size_Ac = 1000;
+    static final private int Size_Ar = Size_Ac + 1;
+    static final private double Pi = (float) Math.PI;
+    static final private double Pi_H = Pi / 2;
+
+    static final private double[] Atan2 = new double[Size_Ar];
+    static final private double[] Atan2_PM = new double[Size_Ar];
+    static final private double[] Atan2_MP = new double[Size_Ar];
+    static final private double[] Atan2_MM = new double[Size_Ar];
+
+    static final private double[] Atan2_R = new double[Size_Ar];
+    static final private double[] Atan2_RPM = new double[Size_Ar];
+    static final private double[] Atan2_RMP = new double[Size_Ar];
+    static final private double[] Atan2_RMM = new double[Size_Ar];
+
+    static {
+        for (int i = 0; i <= Size_Ac; i++) {
+            double d = (double) i / Size_Ac;
+            double x = 1;
+            double y = x * d;
+            double v = Math.atan2(y, x);
+            Atan2[i] = v;
+            Atan2_PM[i] = Pi - v;
+            Atan2_MP[i] = -v;
+            Atan2_MM[i] = -Pi + v;
+
+            Atan2_R[i] = Pi_H - v;
+            Atan2_RPM[i] = Pi_H + v;
+            Atan2_RMP[i] = -Pi_H + v;
+            Atan2_RMM[i] = -Pi_H - v;
+
+        }
+    }
+
+    @SuppressWarnings("ParameterAssignment")
+    static public double fastAtan2(double y, double x) {
+
+        if (y < 0) {
+            if (x < 0) {
+                //(y < x) because == (-y > -x)
+                if (y < x) {
+                    return Atan2_RMM[(int) (x / y * Size_Ac)];
+                } else {
+                    return Atan2_MM[(int) (y / x * Size_Ac)];
+                }
+            } else {
+                y = -y;
+                if (y > x) {
+                    return Atan2_RMP[(int) (x / y * Size_Ac)];
+                } else {
+                    return Atan2_MP[(int) (y / x * Size_Ac)];
+                }
+            }
+        } else {
+            if (x < 0) {
+                x = -x;
+                if (y > x) {
+                    return Atan2_RPM[(int) (x / y * Size_Ac)];
+                } else {
+                    return Atan2_PM[(int) (y / x * Size_Ac)];
+                }
+            } else {
+                if (y > x) {
+                    return Atan2_R[(int) (x / y * Size_Ac)];
+                } else {
+                    return Atan2[(int) (y / x * Size_Ac)];
+                }
+            }
+        }
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/AbstractPageBlock.java

@@ -1,28 +1,29 @@
 package com.knecon.fforesight.service.layoutparser.processor.model;
 
+import java.util.HashSet;
+import java.util.Set;
+
 import com.fasterxml.jackson.annotation.JsonIgnore;
-import com.iqser.red.service.persistence.service.v1.api.shared.model.redactionlog.Rectangle;
+import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.BoundingBox;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
 
 import lombok.AllArgsConstructor;
 import lombok.Data;
+import lombok.EqualsAndHashCode;
 import lombok.NoArgsConstructor;
 
 @Data
 @AllArgsConstructor
 @NoArgsConstructor
-public abstract class AbstractPageBlock {
+@EqualsAndHashCode(callSuper = true)
+public abstract class AbstractPageBlock extends BoundingBox {
 
-    @JsonIgnore
-    protected float minX;
-    @JsonIgnore
-    protected float maxX;
-    @JsonIgnore
-    protected float minY;
-    @JsonIgnore
-    protected float maxY;
     @JsonIgnore
     protected PageBlockType classification;
+
+    Set<LayoutEngine> engines = new HashSet<>();
+
     @JsonIgnore
     protected int page;
 
@@ -39,42 +40,6 @@ public abstract class AbstractPageBlock {
     }
 
 
-    public boolean containsBlock(TextPageBlock other) {
-
-        return this.minX <= other.getPdfMinX() && this.maxX >= other.getPdfMaxX() && this.minY >= other.getPdfMinY() && this.maxY <= other.getPdfMaxY();
-    }
-
-
-    public boolean contains(AbstractPageBlock other) {
-
-        return this.minX <= other.minX && this.maxX >= other.maxX && this.minY >= other.minY && this.maxY <= other.maxY;
-    }
-
-
-    public boolean contains(Rectangle other) {
-
-        return page == other.getPage() && this.minX <= other.getTopLeft().getX() && this.maxX >= other.getTopLeft().getX() + other.getWidth() && this.minY <= other.getTopLeft()
-                .getY() && this.maxY >= other.getTopLeft().getY() + other.getHeight();
-    }
-
-
-    @JsonIgnore
-    public float getHeight() {
-
-        return maxY - minY;
-    }
-
-
-    @JsonIgnore
-    public float getWidth() {
-
-        return maxX - minX;
-    }
-
-
-    public boolean intersectsY(AbstractPageBlock atc) {
-
-        return this.minY <= atc.getMaxY() && this.maxY >= atc.getMinY();
-    }
+    public abstract boolean isEmpty();
 
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/ClassificationDocument.java

@@ -3,9 +3,11 @@ package com.knecon.fforesight.service.layoutparser.processor.model;
 import java.util.ArrayList;
 import java.util.List;
 
-import com.iqser.red.service.persistence.service.v1.api.shared.model.redactionlog.section.SectionGrid;
+import com.knecon.fforesight.service.layoutparser.processor.model.outline.OutlineObjectTree;
+import com.knecon.fforesight.service.layoutparser.processor.model.outline.TableOfContents;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.StringFrequencyCounter;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.UnclassifiedText;
+import com.knecon.fforesight.service.layoutparser.processor.visualization.LayoutDebugLayer;
 
 import lombok.Data;
 import lombok.NoArgsConstructor;
@@ -23,9 +25,12 @@ public class ClassificationDocument {
     private FloatFrequencyCounter fontSizeCounter = new FloatFrequencyCounter();
     private StringFrequencyCounter fontCounter = new StringFrequencyCounter();
     private StringFrequencyCounter fontStyleCounter = new StringFrequencyCounter();
+    private LayoutDebugLayer layoutDebugLayer = new LayoutDebugLayer();
     private boolean headlines;
 
-    private SectionGrid sectionGrid = new SectionGrid();
     private long rulesVersion;
 
+    private OutlineObjectTree outlineObjectTree;
+    private TableOfContents tableOfContents;
+
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/ClassificationPage.java

@@ -1,10 +1,14 @@
 package com.knecon.fforesight.service.layoutparser.processor.model;
 
+import java.awt.geom.Rectangle2D;
 import java.util.ArrayList;
+import java.util.HashMap;
 import java.util.List;
+import java.util.Map;
 
 import com.iqser.red.service.persistence.service.v1.api.shared.model.redactionlog.Rectangle;
 import com.knecon.fforesight.service.layoutparser.processor.model.image.ClassifiedImage;
+import com.knecon.fforesight.service.layoutparser.processor.model.outline.OutlineObject;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRulings;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.StringFrequencyCounter;
 
@@ -14,11 +18,16 @@ import lombok.RequiredArgsConstructor;
 
 @Data
 @RequiredArgsConstructor
+
 public class ClassificationPage {
 
     @NonNull
     private List<AbstractPageBlock> textBlocks;
 
+    private List<OutlineObject> outlineObjects = new ArrayList<>();
+
+    private List<AbstractPageBlock> headlines = new ArrayList<>();
+
     private List<ClassifiedImage> images = new ArrayList<>();
 
     private Rectangle bodyTextFrame;
@@ -36,6 +45,8 @@ public class ClassificationPage {
     private float pageWidth;
     private float pageHeight;
 
-    CleanRulings cleanRulings;
+    private CleanRulings cleanRulings;
+
+    private Map<String, List<Rectangle2D>> markedContentBboxPerType = new HashMap<>();
 
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/ClassificationSection.java

@@ -2,6 +2,7 @@ package com.knecon.fforesight.service.layoutparser.processor.model;
 
 import java.util.ArrayList;
 import java.util.List;
+import java.util.stream.Collectors;
 
 import com.knecon.fforesight.service.layoutparser.processor.model.image.ClassifiedImage;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.TablePageBlock;
@@ -11,6 +12,7 @@ import lombok.NoArgsConstructor;
 
 @Data
 @NoArgsConstructor
+@Deprecated
 public class ClassificationSection {
 
     private List<AbstractPageBlock> pageBlocks = new ArrayList<>();
@@ -29,4 +31,10 @@ public class ClassificationSection {
         return tables;
     }
 
+
+    public List<AbstractPageBlock> getNonEmptyPageBlocks() {
+
+        return pageBlocks.stream().filter(pageBlock -> !pageBlock.isEmpty()).collect(Collectors.toList());
+    }
+
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/FloatFrequencyCounter.java

@@ -1,6 +1,5 @@
 package com.knecon.fforesight.service.layoutparser.processor.model;
 
-import java.util.ArrayList;
 import java.util.Collections;
 import java.util.HashMap;
 import java.util.List;
@@ -12,10 +11,14 @@ import lombok.Getter;
 @Getter
 public class FloatFrequencyCounter {
 
-    Map<Float, Integer> countPerValue = new HashMap<>();
+    Map<Double, Integer> countPerValue = new HashMap<>();
+    boolean changed;
+    Double mostPopularCache;
 
 
-    public void add(float value) {
+    public void add(double value) {
+
+        changed = true;
 
         if (!countPerValue.containsKey(value)) {
             countPerValue.put(value, 1);
@@ -25,9 +28,11 @@ public class FloatFrequencyCounter {
     }
 
 
-    public void addAll(Map<Float, Integer> otherCounter) {
+    public void addAll(Map<Double, Integer> otherCounter) {
 
-        for (Map.Entry<Float, Integer> entry : otherCounter.entrySet()) {
+        changed = true;
+
+        for (Map.Entry<Double, Integer> entry : otherCounter.entrySet()) {
             if (countPerValue.containsKey(entry.getKey())) {
                 countPerValue.put(entry.getKey(), countPerValue.get(entry.getKey()) + entry.getValue());
             } else {
@@ -37,36 +42,36 @@ public class FloatFrequencyCounter {
     }
 
 
-    public Float getMostPopular() {
+    public Double getMostPopular() {
 
-        Map.Entry<Float, Integer> mostPopular = null;
-        for (Map.Entry<Float, Integer> entry : countPerValue.entrySet()) {
-            if (mostPopular == null || entry.getValue() >= mostPopular.getValue()) {
-                mostPopular = entry;
+        if (changed) {
+            Map.Entry<Double, Integer> mostPopular = null;
+            for (Map.Entry<Double, Integer> entry : countPerValue.entrySet()) {
+                if (mostPopular == null || entry.getValue() >= mostPopular.getValue()) {
+                    mostPopular = entry;
+                }
             }
+            mostPopularCache = mostPopular != null ? mostPopular.getKey() : null;
+            changed = false;
         }
-        return mostPopular != null ? mostPopular.getKey() : null;
+
+        return mostPopularCache;
     }
 
 
-    public List<Float> getHighterThanMostPopular() {
+    public List<Double> getValuesInReverseOrder() {
 
-        Float mostPopular = getMostPopular();
-        List<Float> higher = new ArrayList<>();
-        for (Float value : countPerValue.keySet()) {
-            if (value > mostPopular) {
-                higher.add(value);
-            }
-        }
-
-        return higher.stream().sorted(Collections.reverseOrder()).collect(Collectors.toList());
+        return countPerValue.keySet()
+                .stream()
+                .sorted(Collections.reverseOrder())
+                .collect(Collectors.toList());
     }
 
 
-    public Float getHighest() {
+    public Double getHighest() {
 
-        Float highest = null;
-        for (Float value : countPerValue.keySet()) {
+        Double highest = null;
+        for (Double value : countPerValue.keySet()) {
             if (highest == null || value > highest) {
                 highest = value;
             }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/PageBlockType.java

@@ -31,6 +31,19 @@ public enum PageBlockType {
     }
 
 
+    public static int getHeadlineNumber(PageBlockType pageBlockType) {
+
+        return switch (pageBlockType) {
+            case H1 -> 1;
+            case H2 -> 2;
+            case H3 -> 3;
+            case H4 -> 4;
+            case H5 -> 5;
+            default -> 6;
+        };
+    }
+
+
     public boolean isHeadline() {
 
         return this.equals(H1) || this.equals(H2) || this.equals(H3) || this.equals(H4) || this.equals(H5) || this.equals(H6);

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/SectionIdentifier.java

@@ -3,26 +3,32 @@ package com.knecon.fforesight.service.layoutparser.processor.model;
 import java.util.Collections;
 import java.util.LinkedList;
 import java.util.List;
+import java.util.Locale;
 import java.util.regex.Matcher;
 import java.util.regex.Pattern;
 
 import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
+import lombok.Getter;
 import lombok.experimental.FieldDefaults;
 
 @AllArgsConstructor
 @FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
 public class SectionIdentifier {
 
-    static Pattern numericalIdentifierPattern = Pattern.compile("^[\\s]?(\\d+)[\\s.,;]?(\\d+)?[\\s.,;]?(\\d+)?[\\s.,;]?(\\d+)?");
+    public static Pattern numericalIdentifierPattern = Pattern.compile("^[\\s]?(\\d+)[\\s.,;]?(\\d+)?[\\s.,;]?(\\d+)?[\\s.,;]?(\\d+)?");
+    public static Pattern alphanumericIdentifierPattern = Pattern.compile("^[\\s]?[A-Za-z][\\s.,;]?(\\d+)[\\s.,;]?(\\d+)?[\\s.,;]?(\\d+)?[\\s.,;]?(\\d+)?[\\s.,;]?");
 
-    private enum Format {
+    public enum Format {
         EMPTY,
         NUMERICAL,
+        ALPHANUMERIC,
         DOCUMENT
     }
 
+    @Getter
     Format format;
+    @Getter
     String identifierString;
     List<Integer> identifiers;
     boolean asChild;
@@ -38,6 +44,10 @@ public class SectionIdentifier {
         if (numericalIdentifierMatcher.find()) {
             return buildNumericalSectionIdentifier(headline, numericalIdentifierMatcher);
         }
+        Matcher alphanumericIdentifierMatcher = alphanumericIdentifierPattern.matcher(headline);
+        if (alphanumericIdentifierMatcher.find()) {
+            return buildAlphanumericSectionIdentifier(headline, alphanumericIdentifierMatcher);
+        }
         // more formats here
         return SectionIdentifier.empty();
     }
@@ -72,7 +82,36 @@ public class SectionIdentifier {
             }
             identifiers.add(Integer.parseInt(numericalIdentifier.trim()));
         }
-        return new SectionIdentifier(Format.NUMERICAL, identifierString, identifiers.stream().toList(), false);
+        return new SectionIdentifier(Format.NUMERICAL,
+                                     identifierString,
+                                     identifiers.stream()
+                                             .toList(),
+                                     false);
+    }
+
+
+    private static SectionIdentifier buildAlphanumericSectionIdentifier(String headline, Matcher alphanumericIdentifierMatcher) {
+
+        String identifierString = headline.substring(alphanumericIdentifierMatcher.start(), alphanumericIdentifierMatcher.end());
+
+        String alphanumericIdentifier = alphanumericIdentifierMatcher.group(0).substring(0, 1).toUpperCase(Locale.ENGLISH);
+        int mappedCharacterValue = alphanumericIdentifier.charAt(0) - 'A' + 1;
+        List<Integer> identifiers = new LinkedList<>();
+        identifiers.add(mappedCharacterValue);
+
+        for (int i = 1; i <= 3; i++) {
+            String numericalIdentifier = alphanumericIdentifierMatcher.group(i);
+            if (numericalIdentifier == null || numericalIdentifier.equals("0") || numericalIdentifier.isEmpty() || numericalIdentifier.isBlank()) {
+                break;
+            }
+            identifiers.add(Integer.parseInt(numericalIdentifier.trim()));
+        }
+
+        return new SectionIdentifier(Format.ALPHANUMERIC,
+                                     identifierString,
+                                     identifiers.stream()
+                                             .toList(),
+                                     false);
     }
 
 
@@ -120,4 +159,22 @@ public class SectionIdentifier {
         return identifierString;
     }
 
+
+    public boolean isEmpty() {
+
+        return this.format.equals(Format.EMPTY);
+    }
+
+
+    public int level() {
+
+        return identifiers.size();
+    }
+
+
+    protected List<Integer> getIdentifiers() {
+
+        return identifiers;
+    }
+
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/graph/AbstractNodeVisitor.java

@@ -0,0 +1,94 @@
+package com.knecon.fforesight.service.layoutparser.processor.model.graph;
+
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Footer;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Header;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Headline;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Image;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Paragraph;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Section;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.SemanticNode;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.SuperSection;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Table;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.TableCell;
+
+public abstract class AbstractNodeVisitor implements NodeVisitor {
+
+    @Override
+    public void visit(Document document) {
+
+        visitChildren(document);
+    }
+
+
+    @Override
+    public void visit(SuperSection superSection) {
+
+        visitChildren(superSection);
+    }
+
+
+    @Override
+    public void visit(Section section) {
+
+        visitChildren(section);
+    }
+
+
+    @Override
+    public void visit(Headline headline) {
+
+        visitChildren(headline);
+    }
+
+
+    @Override
+    public void visit(Paragraph paragraph) {
+
+        visitChildren(paragraph);
+    }
+
+
+    @Override
+    public void visit(Footer footer) {
+
+        visitChildren(footer);
+    }
+
+
+    @Override
+    public void visit(Header header) {
+
+        visitChildren(header);
+    }
+
+
+    @Override
+    public void visit(Image image) {
+
+        visitChildren(image);
+    }
+
+
+    @Override
+    public void visit(Table table) {
+
+        visitChildren(table);
+    }
+
+
+    @Override
+    public void visit(TableCell tableCell) {
+
+        visitChildren(tableCell);
+    }
+
+
+    protected void visitChildren(SemanticNode semanticNode) {
+
+        semanticNode.streamChildren()
+                .forEach(node -> node.accept(this));
+
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/graph/DocumentTree.java

@@ -2,6 +2,7 @@ package com.knecon.fforesight.service.layoutparser.processor.model.graph;
 
 import static java.lang.String.format;
 
+import java.util.ArrayList;
 import java.util.Collections;
 import java.util.LinkedList;
 import java.util.List;
@@ -39,7 +40,10 @@ public class DocumentTree {
 
     public TextBlock buildTextBlock() {
 
-        return allEntriesInOrder().map(Entry::getNode).filter(SemanticNode::isLeaf).map(SemanticNode::getLeafTextBlock).collect(new TextBlockCollector());
+        return allEntriesInOrder().map(Entry::getNode)
+                .filter(SemanticNode::isLeaf)
+                .map(SemanticNode::getLeafTextBlock)
+                .collect(new TextBlockCollector());
     }
 
 
@@ -113,13 +117,16 @@ public class DocumentTree {
 
     public Stream<SemanticNode> childNodes(List<Integer> treeId) {
 
-        return getEntryById(treeId).children.stream().map(Entry::getNode);
+        return getEntryById(treeId).children.stream()
+                .map(Entry::getNode);
     }
 
 
     public Stream<SemanticNode> childNodesOfType(List<Integer> treeId, NodeType nodeType) {
 
-        return getEntryById(treeId).children.stream().filter(entry -> entry.node.getType().equals(nodeType)).map(Entry::getNode);
+        return getEntryById(treeId).children.stream()
+                .filter(entry -> entry.node.getType().equals(nodeType))
+                .map(Entry::getNode);
     }
 
 
@@ -140,8 +147,8 @@ public class DocumentTree {
         if (treeId.isEmpty()) {
             return root;
         }
-        Entry entry = root.children.get(treeId.get(0));
-        for (int id : treeId.subList(1, treeId.size())) {
+        Entry entry = root;
+        for (int id : treeId) {
             entry = entry.children.get(id);
         }
         return entry;
@@ -156,26 +163,32 @@ public class DocumentTree {
 
     public Stream<Entry> allEntriesInOrder() {
 
-        return Stream.of(root).flatMap(DocumentTree::flatten);
+        return Stream.of(root)
+                .flatMap(DocumentTree::flatten);
     }
 
 
     public Stream<Entry> allSubEntriesInOrder(List<Integer> parentId) {
 
-        return getEntryById(parentId).children.stream().flatMap(DocumentTree::flatten);
+        return getEntryById(parentId).children.stream()
+                .flatMap(DocumentTree::flatten);
     }
 
 
     @Override
     public String toString() {
 
-        return String.join("\n", allEntriesInOrder().map(Entry::toString).toList());
+        return String.join("\n",
+                           allEntriesInOrder().map(Entry::toString)
+                                   .toList());
     }
 
 
     private static Stream<Entry> flatten(Entry entry) {
 
-        return Stream.concat(Stream.of(entry), entry.children.stream().flatMap(DocumentTree::flatten));
+        return Stream.concat(Stream.of(entry),
+                             entry.children.stream()
+                                     .flatMap(DocumentTree::flatten));
     }
 
 

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/graph/NodeVisitor.java

@@ -0,0 +1,45 @@
+package com.knecon.fforesight.service.layoutparser.processor.model.graph;
+
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Footer;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Header;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Headline;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Image;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Paragraph;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Section;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.SuperSection;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Table;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.TableCell;
+
+public interface NodeVisitor {
+
+    void visit(Document document);
+
+
+    void visit(SuperSection superSection);
+
+
+    void visit(Section section);
+
+
+    void visit(Headline headline);
+
+
+    void visit(Paragraph paragraph);
+
+
+    void visit(Footer footer);
+
+
+    void visit(Header header);
+
+
+    void visit(Image image);
+
+
+    void visit(Table table);
+
+
+    void visit(TableCell tableCell);
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/graph/TextRange.java

@@ -12,13 +12,14 @@ import lombok.Setter;
 
 @Setter
 @EqualsAndHashCode
-public class Boundary implements Comparable<Boundary> {
+@SuppressWarnings("PMD.AvoidFieldNameMatchingMethodName")
+public class TextRange implements Comparable<TextRange> {
 
     private int start;
     private int end;
 
 
-    public Boundary(int start, int end) {
+    public TextRange(int start, int end) {
 
         if (start > end) {
             throw new IllegalArgumentException(format("start: %d > end: %d", start, end));
@@ -46,15 +47,15 @@ public class Boundary implements Comparable<Boundary> {
     }
 
 
-    public boolean contains(Boundary boundary) {
+    public boolean contains(TextRange textRange) {
 
-        return start <= boundary.start() && boundary.end() <= end;
+        return start <= textRange.start() && textRange.end() <= end;
     }
 
 
-    public boolean containedBy(Boundary boundary) {
+    public boolean containedBy(TextRange textRange) {
 
-        return boundary.contains(this);
+        return textRange.contains(this);
     }
 
 
@@ -78,22 +79,33 @@ public class Boundary implements Comparable<Boundary> {
 
     public boolean contains(int index) {
 
+        return start <= index && index <= end;
+    }
+
+
+    public boolean containsExclusive(int index) {
+
         return start <= index && index < end;
     }
 
 
-    public boolean intersects(Boundary boundary) {
+    public boolean intersects(TextRange textRange) {
 
-        return boundary.start() < this.end && this.start < boundary.end();
+        return textRange.start() < this.end && this.start < textRange.end();
     }
 
 
-    public List<Boundary> split(List<Integer> splitIndices) {
+    public List<TextRange> split(List<Integer> splitIndices) {
 
-        if (splitIndices.stream().anyMatch(idx -> !this.contains(idx))) {
-            throw new IndexOutOfBoundsException(format("%s splitting indices are out of range for %s", splitIndices.stream().filter(idx -> !this.contains(idx)).toList(), this));
+        if (splitIndices.stream()
+                .anyMatch(idx -> !this.containsExclusive(idx))) {
+            throw new IndexOutOfBoundsException(format("%s splitting indices are out of range for %s",
+                                                       splitIndices.stream()
+                                                               .filter(idx -> !this.containsExclusive(idx))
+                                                               .toList(),
+                                                       this));
         }
-        List<Boundary> splitBoundaries = new LinkedList<>();
+        List<TextRange> splitBoundaries = new LinkedList<>();
         int previousIndex = start;
         for (int splitIndex : splitIndices) {
 
@@ -101,23 +113,31 @@ public class Boundary implements Comparable<Boundary> {
             if (splitIndex == previousIndex) {
                 continue;
             }
-            splitBoundaries.add(new Boundary(previousIndex, splitIndex));
+            splitBoundaries.add(new TextRange(previousIndex, splitIndex));
             previousIndex = splitIndex;
         }
-        splitBoundaries.add(new Boundary(previousIndex, end));
+        splitBoundaries.add(new TextRange(previousIndex, end));
         return splitBoundaries;
     }
 
+
     public IntStream intStream() {
 
         return IntStream.range(start, end);
     }
 
-    public static Boundary merge(Collection<Boundary> boundaries) {
 
-        int minStart = boundaries.stream().mapToInt(Boundary::start).min().orElseThrow(IllegalArgumentException::new);
-        int maxEnd = boundaries.stream().mapToInt(Boundary::end).max().orElseThrow(IllegalArgumentException::new);
-        return new Boundary(minStart, maxEnd);
+    public static TextRange merge(Collection<TextRange> boundaries) {
+
+        int minStart = boundaries.stream()
+                .mapToInt(TextRange::start)
+                .min()
+                .orElseThrow(IllegalArgumentException::new);
+        int maxEnd = boundaries.stream()
+                .mapToInt(TextRange::end)
+                .max()
+                .orElseThrow(IllegalArgumentException::new);
+        return new TextRange(minStart, maxEnd);
     }
 
 
@@ -129,12 +149,12 @@ public class Boundary implements Comparable<Boundary> {
 
 
     @Override
-    public int compareTo(Boundary boundary) {
+    public int compareTo(TextRange textRange) {
 
-        if (end < boundary.end() && start < boundary.start()) {
+        if (end < textRange.end() && start < textRange.start()) {
             return -1;
         }
-        if (start > boundary.start() && end > boundary.end()) {
+        if (start > textRange.start() && end > textRange.end()) {
             return 1;
         }
 

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/graph/entity/TextEntity.java

@@ -11,7 +11,7 @@ import java.util.Map;
 import java.util.Set;
 
 import com.iqser.red.service.persistence.service.v1.api.shared.model.redactionlog.Engine;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.Boundary;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.TextRange;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Page;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.SemanticNode;
 import com.knecon.fforesight.service.layoutparser.processor.utils.IdBuilder;
@@ -28,11 +28,11 @@ import lombok.experimental.FieldDefaults;
 @AllArgsConstructor
 @FieldDefaults(level = AccessLevel.PRIVATE)
 @EqualsAndHashCode(onlyExplicitlyIncluded = true)
-public class RedactionEntity {
+public class TextEntity {
 
     // initial values
     @EqualsAndHashCode.Include
-    final Boundary boundary;
+    final TextRange textRange;
     @EqualsAndHashCode.Include
     final String type;
     @EqualsAndHashCode.Include
@@ -47,7 +47,7 @@ public class RedactionEntity {
     boolean dictionaryEntry;
     boolean dossierDictionaryEntry;
     Set<Engine> engines;
-    Set<RedactionEntity> references;
+    Set<TextEntity> references;
     @Builder.Default
     Deque<Integer> matchedRules = new LinkedList<>();
     String redactionReason;
@@ -66,9 +66,9 @@ public class RedactionEntity {
     SemanticNode deepestFullyContainingNode;
 
 
-    public static RedactionEntity initialEntityNode(Boundary boundary, String type, EntityType entityType) {
+    public static TextEntity initialEntityNode(TextRange textRange, String type, EntityType entityType) {
 
-        return RedactionEntity.builder().type(type).entityType(entityType).boundary(boundary).engines(new HashSet<>()).references(new HashSet<>()).build();
+        return TextEntity.builder().type(type).entityType(entityType).textRange(textRange).engines(new HashSet<>()).references(new HashSet<>()).build();
     }
 
 
@@ -132,7 +132,7 @@ public class RedactionEntity {
     public List<RedactionPosition> getRedactionPositionsPerPage() {
 
         if (redactionPositionsPerPage == null || redactionPositionsPerPage.isEmpty()) {
-            Map<Page, List<Rectangle2D>> rectanglesPerLinePerPage = deepestFullyContainingNode.getTextBlock().getPositionsPerPage(boundary);
+            Map<Page, List<Rectangle2D>> rectanglesPerLinePerPage = deepestFullyContainingNode.getTextBlock().getPositionsPerPage(textRange);
 
             Page firstPage = rectanglesPerLinePerPage.keySet()
                     .stream()
@@ -155,21 +155,21 @@ public class RedactionEntity {
     }
 
 
-    public boolean containedBy(RedactionEntity redactionEntity) {
+    public boolean containedBy(TextEntity textEntity) {
 
-        return this.boundary.containedBy(redactionEntity.getBoundary());
+        return this.textRange.containedBy(textEntity.getTextRange());
     }
 
 
-    public boolean contains(RedactionEntity redactionEntity) {
+    public boolean contains(TextEntity textEntity) {
 
-        return this.boundary.contains(redactionEntity.getBoundary());
+        return this.textRange.contains(textEntity.getTextRange());
     }
 
 
-    public boolean intersects(RedactionEntity redactionEntity) {
+    public boolean intersects(TextEntity textEntity) {
 
-        return this.boundary.intersects(redactionEntity.getBoundary());
+        return this.textRange.intersects(textEntity.getTextRange());
     }
 
 
@@ -185,13 +185,13 @@ public class RedactionEntity {
     }
 
 
-    public void addReference(RedactionEntity reference) {
+    public void addReference(TextEntity reference) {
 
         references.add(reference);
     }
 
 
-    public void addReferences(List<RedactionEntity> references) {
+    public void addReferences(List<TextEntity> references) {
 
         this.references.addAll(references);
     }
@@ -210,7 +210,7 @@ public class RedactionEntity {
         sb.append("Entity[\"");
         sb.append(value);
         sb.append("\", ");
-        sb.append(boundary);
+        sb.append(textRange);
         sb.append(", pages[");
         pages.forEach(page -> {
             sb.append(page.getNumber());

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/graph/nodes/AbstractSemanticNode.java

@@ -0,0 +1,74 @@
+package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
+
+import java.awt.geom.Rectangle2D;
+import java.util.HashSet;
+import java.util.List;
+import java.util.Map;
+import java.util.Set;
+
+import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.TextEntity;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
+
+import lombok.AccessLevel;
+import lombok.AllArgsConstructor;
+import lombok.Builder;
+import lombok.Data;
+import lombok.EqualsAndHashCode;
+import lombok.NoArgsConstructor;
+import lombok.experimental.FieldDefaults;
+import lombok.experimental.SuperBuilder;
+import lombok.extern.slf4j.Slf4j;
+
+@Slf4j
+@Data
+@SuperBuilder
+@AllArgsConstructor
+@NoArgsConstructor
+@FieldDefaults(level = AccessLevel.PRIVATE)
+public abstract class AbstractSemanticNode implements GenericSemanticNode {
+
+    @Builder.Default
+    Set<LayoutEngine> engines = new HashSet<>(Set.of(LayoutEngine.ALGORITHM));
+    List<Integer> treeId;
+
+    TextBlock textBlock;
+    @EqualsAndHashCode.Exclude
+    DocumentTree documentTree;
+
+    @Builder.Default
+    @EqualsAndHashCode.Exclude
+    Set<TextEntity> entities = new HashSet<>();
+
+    @EqualsAndHashCode.Exclude
+    Map<Page, Rectangle2D> bBoxCache;
+
+
+    @Override
+    public TextBlock getTextBlock() {
+
+        if (textBlock == null) {
+            textBlock = GenericSemanticNode.super.getTextBlock();
+        }
+        return textBlock;
+    }
+
+
+    @Override
+    public String toString() {
+
+        return treeId.toString() + ": " + getType() + ": " + this.getTextBlock().buildSummary();
+    }
+
+
+    @Override
+    public Map<Page, Rectangle2D> getBBox() {
+
+        if (bBoxCache == null) {
+            bBoxCache = GenericSemanticNode.super.getBBox();
+        }
+        return bBoxCache;
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/graph/nodes/Document.java

@@ -3,40 +3,38 @@ package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
 import java.awt.geom.Rectangle2D;
 import java.util.Collections;
 import java.util.HashMap;
-import java.util.HashSet;
 import java.util.List;
 import java.util.Map;
 import java.util.Set;
 import java.util.stream.Collectors;
 import java.util.stream.Stream;
 
-import com.amazonaws.services.kms.model.NotFoundException;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlockCollector;
+import com.knecon.fforesight.service.layoutparser.processor.visualization.LayoutDebugLayer;
 
 import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
-import lombok.Builder;
 import lombok.Data;
+import lombok.EqualsAndHashCode;
 import lombok.NoArgsConstructor;
 import lombok.experimental.FieldDefaults;
+import lombok.experimental.SuperBuilder;
 
 @Data
-@Builder
+@SuperBuilder
 @AllArgsConstructor
 @NoArgsConstructor
+@EqualsAndHashCode(callSuper = true)
 @FieldDefaults(level = AccessLevel.PRIVATE)
-public class Document implements GenericSemanticNode {
+public class Document extends AbstractSemanticNode {
 
     Set<Page> pages;
-    DocumentTree documentTree;
     Integer numberOfPages;
-    TextBlock textBlock;
-    @Builder.Default
-    Set<RedactionEntity> entities = new HashSet<>();
+
+    LayoutDebugLayer layoutDebugLayer;
 
 
     @Override
@@ -46,24 +44,72 @@ public class Document implements GenericSemanticNode {
     }
 
 
-    public TextBlock getTextBlock() {
+    /**
+     * Gets the sections of the document as a list.
+     *
+     * @return A list of all sections within the document.
+     */
+    public List<Section> getAllSections() {
 
-        if (textBlock == null) {
-            textBlock = streamTerminalTextBlocksInOrder().collect(new TextBlockCollector());
-        }
-        return textBlock;
+        return streamAllSubNodesOfType(NodeType.SECTION).map(node -> (Section) node)
+                .collect(Collectors.toList());
     }
 
 
+    /**
+     * Gets the main sections of the document as a list.
+     *
+     * @return A list of main sections within the document
+     * @deprecated This method is marked for removal.
+     * Use {@link #streamChildrenOfType(NodeType)} instead,
+     * or {@link #getChildrenOfTypeSectionOrSuperSection()} which returns children of type SECTION as well as SUPER_SECTION.
+     */
+    @Deprecated(forRemoval = true)
     public List<Section> getMainSections() {
 
-        return streamChildrenOfType(NodeType.SECTION).map(node -> (Section) node).collect(Collectors.toList());
+        return streamChildrenOfType(NodeType.SECTION).map(node -> (Section) node)
+                .collect(Collectors.toList());
+    }
+
+
+    /**
+     * Gets the direct children of type SECTION or SUPER_SECTION of the document as a list of SemanticNode objects.
+     *
+     * @return A list of all children of type SECTION or SUPER_SECTION.
+     */
+    public List<SemanticNode> getChildrenOfTypeSectionOrSuperSection() {
+
+        return streamChildren().filter(semanticNode -> semanticNode.getType().equals(NodeType.SECTION) || semanticNode.getType().equals(NodeType.SUPER_SECTION))
+                .toList();
+    }
+
+
+    public List<Header> getHeaders() {
+
+        return streamChildrenOfType(NodeType.HEADER).map(node -> (Header) node)
+                .collect(Collectors.toList());
+    }
+
+
+    public List<Footer> getFooters() {
+
+        return streamChildrenOfType(NodeType.FOOTER).map(node -> (Footer) node)
+                .collect(Collectors.toList());
+    }
+
+
+    @Override
+    public Headline getHeadline() {
+
+        return streamAllSubNodesOfType(NodeType.HEADLINE).map(node -> (Headline) node)
+                .findFirst().orElse(Headline.builder().build());
     }
 
 
     public Stream<TextBlock> streamTerminalTextBlocksInOrder() {
 
-        return streamAllNodes().filter(SemanticNode::isLeaf).map(SemanticNode::getLeafTextBlock);
+        return streamAllNodes().filter(SemanticNode::isLeaf)
+                .map(SemanticNode::getTextBlock);
     }
 
 
@@ -81,16 +127,10 @@ public class Document implements GenericSemanticNode {
     }
 
 
-    @Override
-    public Headline getHeadline() {
-
-        return streamAllSubNodesOfType(NodeType.HEADLINE).map(node -> (Headline) node).findFirst().orElseThrow(() -> new NotFoundException("No Headlines found in this document!"));
-    }
-
-
     private Stream<SemanticNode> streamAllNodes() {
 
-        return documentTree.allEntriesInOrder().map(DocumentTree.Entry::getNode);
+        return getDocumentTree().allEntriesInOrder()
+                .map(DocumentTree.Entry::getNode);
     }
 
 
@@ -100,6 +140,12 @@ public class Document implements GenericSemanticNode {
     }
 
 
+    public Map<NodeType, Long> buildSemanticNodeCounts() {
+
+        return streamAllSubNodes().collect(Collectors.groupingBy(SemanticNode::getType, Collectors.counting()));
+    }
+
+
     @Override
     public String toString() {
 
@@ -117,4 +163,11 @@ public class Document implements GenericSemanticNode {
         return bBox;
     }
 
+
+    @Override
+    public void accept(NodeVisitor visitor) {
+
+        visitor.visit(this);
+    }
+
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/graph/nodes/DuplicatedParagraph.java

@@ -0,0 +1,28 @@
+package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
+
+import java.util.stream.Stream;
+
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlockCollector;
+
+import lombok.Data;
+import lombok.EqualsAndHashCode;
+import lombok.experimental.SuperBuilder;
+
+@Data
+@EqualsAndHashCode(callSuper = true)
+@SuperBuilder
+public class DuplicatedParagraph extends Paragraph {
+
+    TextBlock unsortedLeafTextBlock;
+
+
+    @Override
+    public TextBlock getTextBlock() {
+
+        return Stream.of(super.getLeafTextBlock(), unsortedLeafTextBlock)
+                .collect(new TextBlockCollector());
+
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/graph/nodes/Footer.java

@@ -1,39 +1,25 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
 
-import java.util.HashSet;
-import java.util.List;
-import java.util.Set;
-
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
 
 import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
-import lombok.Builder;
 import lombok.Data;
 import lombok.EqualsAndHashCode;
-import lombok.NoArgsConstructor;
 import lombok.experimental.FieldDefaults;
+import lombok.experimental.SuperBuilder;
 
 @Data
-@Builder
+@SuperBuilder
 @AllArgsConstructor
-@NoArgsConstructor
+@EqualsAndHashCode(callSuper = true)
 @FieldDefaults(level = AccessLevel.PRIVATE)
-public class Footer implements GenericSemanticNode {
+public class Footer extends AbstractSemanticNode {
 
-    List<Integer> treeId;
     TextBlock leafTextBlock;
 
-    @EqualsAndHashCode.Exclude
-    DocumentTree documentTree;
-
-    @Builder.Default
-    @EqualsAndHashCode.Exclude
-    Set<RedactionEntity> entities = new HashSet<>();
-
 
     @Override
     public NodeType getType() {
@@ -49,6 +35,12 @@ public class Footer implements GenericSemanticNode {
     }
 
 
+    @Override
+    public void accept(NodeVisitor visitor) {
+        visitor.visit(this);
+    }
+
+
     @Override
     public TextBlock getTextBlock() {
 
@@ -59,7 +51,7 @@ public class Footer implements GenericSemanticNode {
     @Override
     public String toString() {
 
-        return treeId + ": " + NodeType.FOOTER + ": " + leafTextBlock.buildSummary();
+        return getTreeId() + ": " + NodeType.FOOTER + ": " + leafTextBlock.buildSummary();
     }
 
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/graph/nodes/Header.java

@@ -1,39 +1,25 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
 
-import java.util.HashSet;
-import java.util.List;
-import java.util.Set;
-
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
 
 import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
-import lombok.Builder;
 import lombok.Data;
 import lombok.EqualsAndHashCode;
-import lombok.NoArgsConstructor;
 import lombok.experimental.FieldDefaults;
+import lombok.experimental.SuperBuilder;
 
 @Data
-@Builder
+@SuperBuilder
 @AllArgsConstructor
-@NoArgsConstructor
+@EqualsAndHashCode(callSuper = true)
 @FieldDefaults(level = AccessLevel.PRIVATE)
-public class Header implements GenericSemanticNode {
+public class Header extends AbstractSemanticNode {
 
-    List<Integer> treeId;
     TextBlock leafTextBlock;
 
-    @EqualsAndHashCode.Exclude
-    DocumentTree documentTree;
-
-    @Builder.Default
-    @EqualsAndHashCode.Exclude
-    Set<RedactionEntity> entities = new HashSet<>();
-
 
     @Override
     public boolean isLeaf() {
@@ -42,6 +28,13 @@ public class Header implements GenericSemanticNode {
     }
 
 
+    @Override
+    public void accept(NodeVisitor visitor) {
+
+        visitor.visit(this);
+    }
+
+
     @Override
     public NodeType getType() {
 
@@ -59,7 +52,7 @@ public class Header implements GenericSemanticNode {
     @Override
     public String toString() {
 
-        return treeId + ": " + NodeType.HEADER + ": " + leafTextBlock.buildSummary();
+        return getTreeId() + ": " + NodeType.HEADER + ": " + leafTextBlock.buildSummary();
     }
 
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/graph/nodes/Headline.java

@@ -1,39 +1,25 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
 
-import java.util.HashSet;
-import java.util.List;
-import java.util.Set;
-
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
 
 import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
-import lombok.Builder;
 import lombok.Data;
 import lombok.EqualsAndHashCode;
-import lombok.NoArgsConstructor;
 import lombok.experimental.FieldDefaults;
+import lombok.experimental.SuperBuilder;
 
 @Data
-@Builder
+@SuperBuilder
 @AllArgsConstructor
-@NoArgsConstructor
+@EqualsAndHashCode(callSuper = true)
 @FieldDefaults(level = AccessLevel.PRIVATE)
-public class Headline implements GenericSemanticNode {
+public class Headline extends AbstractSemanticNode {
 
-    List<Integer> treeId;
     TextBlock leafTextBlock;
 
-    @EqualsAndHashCode.Exclude
-    DocumentTree documentTree;
-
-    @Builder.Default
-    @EqualsAndHashCode.Exclude
-    Set<RedactionEntity> entities = new HashSet<>();
-
 
     @Override
     public NodeType getType() {
@@ -49,6 +35,13 @@ public class Headline implements GenericSemanticNode {
     }
 
 
+    @Override
+    public void accept(NodeVisitor visitor) {
+
+        visitor.visit(this);
+    }
+
+
     @Override
     public TextBlock getTextBlock() {
 
@@ -59,7 +52,7 @@ public class Headline implements GenericSemanticNode {
     @Override
     public String toString() {
 
-        return treeId + ": " + NodeType.HEADLINE + ": " + leafTextBlock.buildSummary();
+        return getTreeId() + ": " + NodeType.HEADLINE + ": " + leafTextBlock.buildSummary();
     }
 
 

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/graph/nodes/Image.java

@@ -3,16 +3,12 @@ package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
 import java.awt.geom.Rectangle2D;
 import java.util.Collections;
 import java.util.HashMap;
-import java.util.HashSet;
-import java.util.List;
 import java.util.Map;
 import java.util.Set;
 
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlockCollector;
 
 import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
@@ -21,21 +17,26 @@ import lombok.Data;
 import lombok.EqualsAndHashCode;
 import lombok.NoArgsConstructor;
 import lombok.experimental.FieldDefaults;
+import lombok.experimental.SuperBuilder;
 
 @Data
-@Builder
+@SuperBuilder
 @AllArgsConstructor
 @NoArgsConstructor
+@EqualsAndHashCode(callSuper = true)
 @FieldDefaults(level = AccessLevel.PRIVATE)
-public class Image implements GenericSemanticNode {
+public class Image extends AbstractSemanticNode {
 
-    List<Integer> treeId;
     String id;
 
+    String representationHash;
+
     ImageType imageType;
     boolean transparent;
     Rectangle2D position;
 
+    TextBlock leafTextBlock;
+
     boolean redaction;
     boolean ignored;
     @Builder.Default
@@ -48,13 +49,6 @@ public class Image implements GenericSemanticNode {
     @EqualsAndHashCode.Exclude
     Page page;
 
-    @EqualsAndHashCode.Exclude
-    DocumentTree documentTree;
-
-    @Builder.Default
-    @EqualsAndHashCode.Exclude
-    Set<RedactionEntity> entities = new HashSet<>();
-
 
     @Override
     public NodeType getType() {
@@ -66,7 +60,7 @@ public class Image implements GenericSemanticNode {
     @Override
     public TextBlock getTextBlock() {
 
-        return streamAllSubNodes().filter(SemanticNode::isLeaf).map(SemanticNode::getLeafTextBlock).collect(new TextBlockCollector());
+        return leafTextBlock;
     }
 
 
@@ -80,7 +74,7 @@ public class Image implements GenericSemanticNode {
     @Override
     public String toString() {
 
-        return treeId + ": " + NodeType.IMAGE + ": " + imageType.toString() + " " + position;
+        return getTreeId() + ": " + NodeType.IMAGE + ": " + imageType.toString() + " " + position;
     }
 
 
@@ -92,4 +86,30 @@ public class Image implements GenericSemanticNode {
         return bBoxPerPage;
     }
 
+
+    @Override
+    public void accept(NodeVisitor visitor) {
+
+        visitor.visit(this);
+    }
+
+
+    @Override
+    public boolean isLeaf() {
+
+        return true;
+    }
+
+
+    public double getArea() {
+
+        return position.getWidth() * position.getHeight();
+    }
+
+
+    public boolean isFullPageImage() {
+
+        return imageType.equals(ImageType.OCR) || getArea() >= 0.5 * page.getArea();
+    }
+
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/graph/nodes/ImageType.java

@@ -6,8 +6,10 @@ public enum ImageType {
     LOGO,
     FORMULA,
     SIGNATURE,
+    SIGNATURE_VISUAL,
     OTHER,
-    OCR;
+    OCR,
+    GRAPHIC;
 
 
     public static ImageType fromString(String imageType) {
@@ -17,6 +19,7 @@ public enum ImageType {
             case "formula" -> ImageType.FORMULA;
             case "signature" -> ImageType.SIGNATURE;
             case "ocr" -> ImageType.OCR;
+            case "graphic" -> ImageType.GRAPHIC;
             default -> ImageType.OTHER;
         };
     }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/graph/nodes/Page.java

@@ -1,12 +1,15 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
 
+import java.util.Comparator;
 import java.util.HashSet;
 import java.util.LinkedList;
 import java.util.List;
 import java.util.Set;
+import java.util.stream.Stream;
 
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.TextEntity;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.AtomicTextBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlockCollector;
 
@@ -29,9 +32,8 @@ public class Page {
     Integer height;
     Integer width;
     Integer rotation;
-
     @EqualsAndHashCode.Exclude
-    List<SemanticNode> mainBody;
+    List<AtomicTextBlock> textBlocksOnPage;
     @EqualsAndHashCode.Exclude
     Header header;
     @EqualsAndHashCode.Exclude
@@ -39,7 +41,7 @@ public class Page {
 
     @Builder.Default
     @EqualsAndHashCode.Exclude
-    Set<RedactionEntity> entities = new HashSet<>();
+    Set<TextEntity> entities = new HashSet<>();
 
     @Builder.Default
     @EqualsAndHashCode.Exclude
@@ -53,14 +55,41 @@ public class Page {
                 .width((int) classificationPage.getPageWidth())
                 .number(classificationPage.getPageNumber())
                 .rotation(classificationPage.getRotation())
-                .mainBody(new LinkedList<>())
+                .textBlocksOnPage(new LinkedList<>())
                 .build();
     }
 
 
+    /**
+     * Constructs and returns a {@link TextBlock} representing the concatenated text of all leaf semantic nodes in the main body.
+     *
+     * @return The main body text block.
+     */
     public TextBlock getMainBodyTextBlock() {
 
-        return mainBody.stream().filter(SemanticNode::isLeaf).map(SemanticNode::getLeafTextBlock).collect(new TextBlockCollector());
+        return textBlocksOnPage.stream()
+                .filter(atb -> !atb.isEmpty())
+                .collect(new TextBlockCollector());
+    }
+
+
+    public List<SemanticNode> getMainBody() {
+
+        return textBlocksOnPage.stream()
+                .map(AtomicTextBlock::getParent)
+                .map(this::getHighestParentOnPage)
+                .distinct()
+                .toList();
+    }
+
+
+    private SemanticNode getHighestParentOnPage(SemanticNode node) {
+
+        SemanticNode currentNode = node;
+        while (currentNode.hasParent() && currentNode.getParent().onlyOnPage(this)) {
+            currentNode = currentNode.getParent();
+        }
+        return currentNode;
     }
 
 
@@ -84,4 +113,10 @@ public class Page {
         return o instanceof Page && o.hashCode() == this.hashCode();
     }
 
+
+    public double getArea() {
+
+        return height * width;
+    }
+
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/graph/nodes/Paragraph.java

@@ -1,37 +1,25 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
 
-import java.util.HashSet;
-import java.util.List;
-import java.util.Set;
-
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
 
 import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
-import lombok.Builder;
 import lombok.Data;
 import lombok.EqualsAndHashCode;
 import lombok.experimental.FieldDefaults;
+import lombok.experimental.SuperBuilder;
 
 @Data
-@Builder
+@SuperBuilder
 @AllArgsConstructor
-@FieldDefaults(level = AccessLevel.PRIVATE)
-public class Paragraph implements GenericSemanticNode {
+@EqualsAndHashCode(callSuper = true)
+@FieldDefaults(level = AccessLevel.PROTECTED)
+public class Paragraph extends AbstractSemanticNode {
 
-    List<Integer> treeId;
     TextBlock leafTextBlock;
 
-    @EqualsAndHashCode.Exclude
-    DocumentTree documentTree;
-
-    @Builder.Default
-    @EqualsAndHashCode.Exclude
-    Set<RedactionEntity> entities = new HashSet<>();
-
 
     @Override
     public NodeType getType() {
@@ -47,17 +35,17 @@ public class Paragraph implements GenericSemanticNode {
     }
 
 
+    @Override
+    public void accept(NodeVisitor visitor) {
+
+        visitor.visit(this);
+    }
+
+
     @Override
     public TextBlock getTextBlock() {
 
         return leafTextBlock;
     }
 
-
-    @Override
-    public String toString() {
-
-        return treeId + ": " + NodeType.PARAGRAPH + ": " + leafTextBlock.buildSummary();
-    }
-
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/graph/nodes/Section.java

@@ -1,40 +1,21 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
 
-import java.util.HashSet;
-import java.util.List;
-import java.util.Set;
-
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlockCollector;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
 
 import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
-import lombok.Builder;
 import lombok.Data;
 import lombok.EqualsAndHashCode;
 import lombok.experimental.FieldDefaults;
-import lombok.extern.slf4j.Slf4j;
+import lombok.experimental.SuperBuilder;
 
-@Slf4j
 @Data
-@Builder
+@SuperBuilder
 @AllArgsConstructor
 @FieldDefaults(level = AccessLevel.PRIVATE)
-public class Section implements GenericSemanticNode {
-
-    List<Integer> treeId;
-
-    TextBlock textBlock;
-    @EqualsAndHashCode.Exclude
-    DocumentTree documentTree;
-
-    @Builder.Default
-    @EqualsAndHashCode.Exclude
-    Set<RedactionEntity> entities = new HashSet<>();
-
+@EqualsAndHashCode(callSuper = true)
+public class Section extends AbstractSemanticNode {
 
     @Override
     public NodeType getType() {
@@ -43,35 +24,30 @@ public class Section implements GenericSemanticNode {
     }
 
 
+    public Headline getHeadline() {
+
+        return streamChildrenOfType(NodeType.HEADLINE).map(node -> (Headline) node)
+                .findFirst().orElseGet(() -> getParent().getHeadline());
+    }
+
+
+    @Override
+    public void accept(NodeVisitor visitor) {
+
+        visitor.visit(this);
+    }
+
+
     public boolean hasTables() {
 
         return streamAllSubNodesOfType(NodeType.TABLE).findAny().isPresent();
     }
 
 
-    @Override
-    public TextBlock getTextBlock() {
-
-        if (textBlock == null) {
-            textBlock = streamAllSubNodes().filter(SemanticNode::isLeaf).map(SemanticNode::getLeafTextBlock).collect(new TextBlockCollector());
-        }
-        return textBlock;
-    }
-
-
     @Override
     public String toString() {
 
-        return treeId.toString() + ": " + NodeType.SECTION + ": " + this.getTextBlock().buildSummary();
-    }
-
-
-    public Headline getHeadline() {
-
-        return streamChildrenOfType(NodeType.HEADLINE)//
-                .map(node -> (Headline) node)//
-                .findFirst()//
-                .orElseGet(() -> getParent().getHeadline());
+        return getTreeId() + ": " + NodeType.SECTION + ": " + this.getTextBlock().buildSummary();
     }
 
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/graph/nodes/SemanticNode.java

@@ -12,14 +12,18 @@ import java.util.Set;
 import java.util.stream.Collectors;
 import java.util.stream.Stream;
 
+import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.Boundary;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.TextRange;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.EntityType;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.TextEntity;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.AtomicTextBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlockCollector;
+import com.knecon.fforesight.service.layoutparser.processor.utils.BBoxMergingUtility;
 import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
 
 public interface SemanticNode {
 
@@ -38,7 +42,12 @@ public interface SemanticNode {
      *
      * @return TextBlock containing all AtomicTextBlocks that are located under this Node.
      */
-    TextBlock getTextBlock();
+    default TextBlock getTextBlock() {
+
+        return streamAllSubNodes().filter(SemanticNode::isLeaf)
+                .map(SemanticNode::getTextBlock)
+                .collect(new TextBlockCollector());
+    }
 
 
     /**
@@ -47,7 +56,7 @@ public interface SemanticNode {
      *
      * @return Set of all Entities associated with this Node
      */
-    Set<RedactionEntity> getEntities();
+    Set<TextEntity> getEntities();
 
 
     /**
@@ -63,7 +72,10 @@ public interface SemanticNode {
 
     default Page getFirstPage() {
 
-        return getTextBlock().getPages().stream().min(Comparator.comparingInt(Page::getNumber)).orElseThrow(() -> new IllegalStateException("SemanticNode has no Page!"));
+        return getTextBlock().getPages()
+                .stream()
+                .min(Comparator.comparingInt(Page::getNumber))
+                .orElseThrow(() -> new IllegalStateException("SemanticNode has no Page!"));
     }
 
 
@@ -72,18 +84,19 @@ public interface SemanticNode {
      *
      * @return Set of PageNodes this node appears on.
      */
-    default Set<Page> getPages(Boundary boundary) {
+    default Set<Page> getPages(TextRange textRange) {
 
-        if (!getBoundary().contains(boundary)) {
-            throw new IllegalArgumentException(format("%s which was used to query for pages is not contained in the %s of this node!", boundary, getBoundary()));
+        if (!getBoundary().contains(textRange)) {
+            throw new IllegalArgumentException(format("%s which was used to query for pages is not contained in the %s of this node!", textRange, getBoundary()));
         }
-        return getTextBlock().getPages(boundary);
+        return getTextBlock().getPages(textRange);
     }
 
 
     default boolean isOnPage(int pageNumber) {
 
-        return getPages().stream().anyMatch(page -> page.getNumber() == pageNumber);
+        return getPages().stream()
+                .anyMatch(page -> page.getNumber() == pageNumber);
     }
 
 
@@ -198,7 +211,9 @@ public interface SemanticNode {
      */
     default boolean hasEntitiesOfType(String type) {
 
-        return getEntities().stream().filter(entity -> entity.getEntityType().equals(EntityType.ENTITY)).anyMatch(redactionEntity -> redactionEntity.getType().equals(type));
+        return getEntities().stream()
+                .filter(entity -> entity.getEntityType().equals(EntityType.ENTITY))
+                .anyMatch(redactionEntity -> redactionEntity.getType().equals(type));
     }
 
 
@@ -208,9 +223,11 @@ public interface SemanticNode {
      * @param type string representing the type of entities to return
      * @return List of RedactionEntities of any the type
      */
-    default List<RedactionEntity> getEntitiesOfType(String type) {
+    default List<TextEntity> getEntitiesOfType(String type) {
 
-        return getEntities().stream().filter(redactionEntity -> redactionEntity.getType().equals(type)).toList();
+        return getEntities().stream()
+                .filter(redactionEntity -> redactionEntity.getType().equals(type))
+                .toList();
     }
 
 
@@ -220,9 +237,11 @@ public interface SemanticNode {
      * @param types A list of strings representing the types of entities to return
      * @return List of RedactionEntities of any provided type
      */
-    default List<RedactionEntity> getEntitiesOfType(List<String> types) {
+    default List<TextEntity> getEntitiesOfType(List<String> types) {
 
-        return getEntities().stream().filter(redactionEntity -> redactionEntity.isAnyType(types)).toList();
+        return getEntities().stream()
+                .filter(redactionEntity -> redactionEntity.isAnyType(types))
+                .toList();
     }
 
 
@@ -274,7 +293,8 @@ public interface SemanticNode {
      */
     default boolean containsStrings(List<String> strings) {
 
-        return strings.stream().allMatch(this::containsString);
+        return strings.stream()
+                .allMatch(this::containsString);
     }
 
 
@@ -298,7 +318,8 @@ public interface SemanticNode {
      */
     default boolean containsAnyString(List<String> strings) {
 
-        return strings.stream().anyMatch(this::containsString);
+        return strings.stream()
+                .anyMatch(this::containsString);
     }
 
 
@@ -310,7 +331,8 @@ public interface SemanticNode {
      */
     default boolean containsAnyStringIgnoreCase(List<String> strings) {
 
-        return strings.stream().anyMatch(this::containsStringIgnoreCase);
+        return strings.stream()
+                .anyMatch(this::containsStringIgnoreCase);
     }
 
 
@@ -318,23 +340,40 @@ public interface SemanticNode {
      * This function is used during insertion of EntityNodes into the graph, it checks if the boundary of the RedactionEntity intersects or even contains the RedactionEntity.
      * It sets the fields accordingly and recursively calls this function on all its children.
      *
-     * @param redactionEntity RedactionEntity, which is being inserted into the graph
+     * @param textEntity RedactionEntity, which is being inserted into the graph
      */
-    default void addThisToEntityIfIntersects(RedactionEntity redactionEntity) {
+    default void addThisToEntityIfIntersects(TextEntity textEntity) {
 
         TextBlock textBlock = getTextBlock();
-        if (textBlock.getBoundary().intersects(redactionEntity.getBoundary())) {
-            if (textBlock.containsBoundary(redactionEntity.getBoundary())) {
-                redactionEntity.setDeepestFullyContainingNode(this);
+        if (textBlock.getTextRange().intersects(textEntity.getTextRange())) {
+            if (textBlock.containsBoundary(textEntity.getTextRange())) {
+                textEntity.setDeepestFullyContainingNode(this);
             }
 
-            redactionEntity.addIntersectingNode(this);
-            streamChildren().filter(semanticNode -> semanticNode.getBoundary().intersects(redactionEntity.getBoundary()))
-                    .forEach(node -> node.addThisToEntityIfIntersects(redactionEntity));
+            textEntity.addIntersectingNode(this);
+            streamChildren().filter(semanticNode -> semanticNode.getBoundary().intersects(textEntity.getTextRange()))
+                    .forEach(node -> node.addThisToEntityIfIntersects(textEntity));
         }
     }
 
 
+    /**
+     * returns the set of layoutengines.
+     *
+     * @return set of layoutengines.
+     */
+    Set<LayoutEngine> getEngines();
+
+
+    /**
+     * adds a layoutengine to the set.
+     */
+    default void addEngine(LayoutEngine engine) {
+
+        getEngines().add(engine);
+    }
+
+
     /**
      * Streams all children located directly underneath this node in the DocumentTree.
      *
@@ -364,7 +403,8 @@ public interface SemanticNode {
      */
     default Stream<SemanticNode> streamAllSubNodes() {
 
-        return getDocumentTree().allSubEntriesInOrder(getTreeId()).map(DocumentTree.Entry::getNode);
+        return getDocumentTree().allSubEntriesInOrder(getTreeId())
+                .map(DocumentTree.Entry::getNode);
     }
 
 
@@ -375,7 +415,9 @@ public interface SemanticNode {
      */
     default Stream<SemanticNode> streamAllSubNodesOfType(NodeType nodeType) {
 
-        return getDocumentTree().allSubEntriesInOrder(getTreeId()).filter(entry -> entry.getType().equals(nodeType)).map(DocumentTree.Entry::getNode);
+        return getDocumentTree().allSubEntriesInOrder(getTreeId())
+                .filter(entry -> entry.getType().equals(nodeType))
+                .map(DocumentTree.Entry::getNode);
     }
 
 
@@ -384,9 +426,9 @@ public interface SemanticNode {
      *
      * @return Boundary of this Node's TextBlock
      */
-    default Boundary getBoundary() {
+    default TextRange getBoundary() {
 
-        return getTextBlock().getBoundary();
+        return getTextBlock().getTextRange();
     }
 
 
@@ -398,12 +440,11 @@ public interface SemanticNode {
      */
     default Map<Page, Rectangle2D> getBBox() {
 
-        Map<Page, Rectangle2D> bBoxPerPage = new HashMap<>();
         if (isLeaf()) {
-            return getBBoxFromLeafTextBlock(bBoxPerPage);
+            return getBBoxFromLeafTextBlock();
         }
 
-        return getBBoxFromChildren(bBoxPerPage);
+        return getBBoxFromChildren();
     }
 
 
@@ -426,28 +467,55 @@ public interface SemanticNode {
 
     /**
      * TODO: this produces unwanted results for sections spanning multiple columns.
+     * Computes the Union of the bounding boxes of all children recursively.
      *
-     * @param bBoxPerPage initial empty BoundingBox
      * @return The union of the BoundingBoxes of all children
      */
-    private Map<Page, Rectangle2D> getBBoxFromChildren(Map<Page, Rectangle2D> bBoxPerPage) {
+    private Map<Page, Rectangle2D> getBBoxFromChildren() {
 
-        return streamChildren().map(SemanticNode::getBBox).reduce((map1, map2) -> {
-            map1.forEach((page, rectangle) -> map2.merge(page, rectangle, (rect1, rect2) -> rect1.createUnion(rect2).getBounds2D()));
-            return map2;
-        }).orElse(bBoxPerPage);
+        List<Map<Page, Rectangle2D>> childrenBBoxes = streamChildren().filter(child -> !isFullPageImage(child))
+                .map(SemanticNode::getBBox)
+                .toList();
+        return BBoxMergingUtility.mergeBBoxes(childrenBBoxes);
+    }
+
+
+    private static boolean isFullPageImage(SemanticNode child) {
+
+        if (!child.getType().equals(NodeType.IMAGE)) {
+            return false;
+        }
+        return ((Image) child).isFullPageImage();
     }
 
 
     /**
-     * @param bBoxPerPage initial empty BoundingBox
      * @return The union of all BoundingBoxes of the TextBlock of this node
      */
-    private Map<Page, Rectangle2D> getBBoxFromLeafTextBlock(Map<Page, Rectangle2D> bBoxPerPage) {
+    private Map<Page, Rectangle2D> getBBoxFromLeafTextBlock() {
 
-        Map<Page, List<AtomicTextBlock>> atomicTextBlockPerPage = getTextBlock().getAtomicTextBlocks().stream().collect(Collectors.groupingBy(AtomicTextBlock::getPage));
+        Map<Page, Rectangle2D> bBoxPerPage = new HashMap<>();
+        Map<Page, List<AtomicTextBlock>> atomicTextBlockPerPage = getTextBlock().getAtomicTextBlocks()
+                .stream()
+                .collect(Collectors.groupingBy(AtomicTextBlock::getPage));
         atomicTextBlockPerPage.forEach((page, atbs) -> bBoxPerPage.put(page, RectangleTransformations.bBoxUnionAtomicTextBlock(atbs)));
         return bBoxPerPage;
     }
 
+
+    void accept(NodeVisitor visitor);
+
+
+    /**
+     * Checks wether this SemanticNode appears on a single page only, and if that page is the provided one.
+     *
+     * @param page the page to check
+     * @return true, when SemanticNode is on a single page only and the page is the provided page. Otherwise, false.
+     */
+    default boolean onlyOnPage(Page page) {
+
+        Set<Page> pages = getPages();
+        return pages.size() == 1 && pages.contains(page);
+    }
+
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/graph/nodes/SuperSection.java

@@ -0,0 +1,47 @@
+package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
+
+import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
+
+import lombok.AccessLevel;
+import lombok.AllArgsConstructor;
+import lombok.Data;
+import lombok.EqualsAndHashCode;
+import lombok.experimental.FieldDefaults;
+import lombok.experimental.SuperBuilder;
+
+@Data
+@SuperBuilder
+@AllArgsConstructor
+@FieldDefaults(level = AccessLevel.PRIVATE)
+@EqualsAndHashCode(callSuper = true)
+public class SuperSection extends AbstractSemanticNode {
+
+    @Override
+    public NodeType getType() {
+
+        return NodeType.SUPER_SECTION;
+    }
+
+
+    public Headline getHeadline() {
+
+        return streamChildrenOfType(NodeType.HEADLINE).map(node -> (Headline) node)
+                .findFirst().orElseGet(() -> getParent().getHeadline());
+    }
+
+
+    @Override
+    public void accept(NodeVisitor visitor) {
+
+        visitor.visit(this);
+    }
+
+
+    @Override
+    public String toString() {
+
+        return getTreeId() + ": " + NodeType.SUPER_SECTION + ": " + this.getTextBlock().buildSummary();
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/graph/nodes/Table.java

@@ -2,17 +2,21 @@ package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
 
 import static java.lang.String.format;
 
+import java.awt.geom.Rectangle2D;
 import java.util.Collection;
 import java.util.HashSet;
 import java.util.List;
 import java.util.Locale;
+import java.util.Map;
 import java.util.Set;
 import java.util.stream.IntStream;
 import java.util.stream.Stream;
 
+import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.TextEntity;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlockCollector;
 
@@ -29,18 +33,21 @@ import lombok.experimental.FieldDefaults;
 @FieldDefaults(level = AccessLevel.PRIVATE)
 public class Table implements SemanticNode {
 
+    @Builder.Default
+    Set<LayoutEngine> engines = new HashSet<>(Set.of(LayoutEngine.ALGORITHM));
     List<Integer> treeId;
     DocumentTree documentTree;
 
     int numberOfRows;
     int numberOfCols;
-
     TextBlock textBlock;
 
     @Builder.Default
     @EqualsAndHashCode.Exclude
-    Set<RedactionEntity> entities = new HashSet<>();
+    Set<TextEntity> entities = new HashSet<>();
 
+    @EqualsAndHashCode.Exclude
+    Map<Page, Rectangle2D> bBoxCache;
 
     /**
      * Streams all entities in this table, that appear in a row, which contains any of the provided strings.
@@ -48,10 +55,9 @@ public class Table implements SemanticNode {
      * @param strings Strings to check whether a row contains them
      * @return Stream of all entities in this table, that appear in a row, which contains any of the provided strings
      */
-    public Stream<RedactionEntity> streamEntitiesWhereRowContainsStringsIgnoreCase(List<String> strings) {
+    public Stream<TextEntity> streamEntitiesWhereRowContainsStringsIgnoreCase(List<String> strings) {
 
-        return IntStream.range(0, numberOfRows)
-                .boxed()
+        return IntStream.range(0, numberOfRows).boxed()
                 .filter(row -> rowContainsStringsIgnoreCase(row, strings))
                 .flatMap(this::streamRow)
                 .map(TableCell::getEntities)
@@ -68,8 +74,11 @@ public class Table implements SemanticNode {
      */
     public boolean rowContainsStringsIgnoreCase(Integer row, List<String> strings) {
 
-        String rowText = streamRow(row).map(TableCell::getTextBlock).collect(new TextBlockCollector()).getSearchText().toLowerCase(Locale.ROOT);
-        return strings.stream().map(String::toLowerCase).allMatch(rowText::contains);
+        String rowText = streamRow(row).map(TableCell::getTextBlock)
+                .collect(new TextBlockCollector()).getSearchText().toLowerCase(Locale.ROOT);
+        return strings.stream()
+                .map(String::toLowerCase)
+                .allMatch(rowText::contains);
     }
 
 
@@ -80,11 +89,15 @@ public class Table implements SemanticNode {
      * @param value  the string which the table cell should contain
      * @return a stream of all entities, which appear in a row where at least one cell has the provided header and the provided value.
      */
-    public Stream<RedactionEntity> streamEntitiesWhereRowHasHeaderAndValue(String header, String value) {
+    public Stream<TextEntity> streamEntitiesWhereRowHasHeaderAndValue(String header, String value) {
 
-        List<Integer> vertebrateStudyCols = streamHeaders().filter(headerNode -> headerNode.containsString(header)).map(TableCell::getCol).toList();
+        List<Integer> vertebrateStudyCols = streamHeaders().filter(headerNode -> headerNode.containsString(header))
+                .map(TableCell::getCol)
+                .toList();
         return streamTableCells().filter(tableCellNode -> vertebrateStudyCols.stream()
-                .anyMatch(vertebrateStudyCol -> getCell(tableCellNode.getRow(), vertebrateStudyCol).containsString(value))).map(TableCell::getEntities).flatMap(Collection::stream);
+                        .anyMatch(vertebrateStudyCol -> getCell(tableCellNode.getRow(), vertebrateStudyCol).containsString(value)))
+                .map(TableCell::getEntities)
+                .flatMap(Collection::stream);
     }
 
 
@@ -95,11 +108,15 @@ public class Table implements SemanticNode {
      * @param values the strings which the table cell should contain
      * @return a stream of all entities, which appear in a row where at least one cell has the provided header and any provided value.
      */
-    public Stream<RedactionEntity> streamEntitiesWhereRowHasHeaderAndAnyValue(String header, List<String> values) {
+    public Stream<TextEntity> streamEntitiesWhereRowHasHeaderAndAnyValue(String header, List<String> values) {
 
-        List<Integer> colsWithHeader = streamHeaders().filter(headerNode -> headerNode.containsString(header)).map(TableCell::getCol).toList();
+        List<Integer> colsWithHeader = streamHeaders().filter(headerNode -> headerNode.containsString(header))
+                .map(TableCell::getCol)
+                .toList();
         return streamTableCells().filter(tableCellNode -> colsWithHeader.stream()
-                .anyMatch(colWithHeader -> getCell(tableCellNode.getRow(), colWithHeader).containsAnyString(values))).map(TableCell::getEntities).flatMap(Collection::stream);
+                        .anyMatch(colWithHeader -> getCell(tableCellNode.getRow(), colWithHeader).containsAnyString(values)))
+                .map(TableCell::getEntities)
+                .flatMap(Collection::stream);
     }
 
 
@@ -109,14 +126,17 @@ public class Table implements SemanticNode {
      * @param types type strings to check whether a row contains an entity like them
      * @return Stream of all entities in this table, that appear in a row, which contains at least one entity with any of the provided types.
      */
-    public Stream<RedactionEntity> streamEntitiesWhereRowContainsEntitiesOfType(List<String> types) {
+    public Stream<TextEntity> streamEntitiesWhereRowContainsEntitiesOfType(List<String> types) {
 
-        List<Integer> rowsWithEntityOfType = IntStream.range(0, numberOfRows)
-                .boxed()
-                .filter(rowNumber -> streamEntityTypesInRow(rowNumber).anyMatch(existingType -> types.stream().anyMatch(typeToCheck -> typeToCheck.equals(existingType))))
+        List<Integer> rowsWithEntityOfType = IntStream.range(0, numberOfRows).boxed()
+                .filter(rowNumber -> streamEntityTypesInRow(rowNumber).anyMatch(existingType -> types.stream()
+                        .anyMatch(typeToCheck -> typeToCheck.equals(existingType))))
                 .toList();
 
-        return rowsWithEntityOfType.stream().flatMap(this::streamRow).map(TableCell::getEntities).flatMap(Collection::stream);
+        return rowsWithEntityOfType.stream()
+                .flatMap(this::streamRow)
+                .map(TableCell::getEntities)
+                .flatMap(Collection::stream);
     }
 
 
@@ -126,20 +146,26 @@ public class Table implements SemanticNode {
      * @param types type strings to check whether a row doesn't contain an entity like it
      * @return Stream of all entities in this table, that appear in a row, which does not contain any entity with any of the provided types.
      */
-    public Stream<RedactionEntity> streamEntitiesWhereRowContainsNoEntitiesOfType(List<String> types) {
+    public Stream<TextEntity> streamEntitiesWhereRowContainsNoEntitiesOfType(List<String> types) {
 
-        List<Integer> rowsWithNoEntityOfType = IntStream.range(0, numberOfRows)
-                .boxed()
-                .filter(rowNumber -> streamEntityTypesInRow(rowNumber).noneMatch(existingType -> types.stream().anyMatch(typeToCheck -> typeToCheck.equals(existingType))))
+        List<Integer> rowsWithNoEntityOfType = IntStream.range(0, numberOfRows).boxed()
+                .filter(rowNumber -> streamEntityTypesInRow(rowNumber).noneMatch(existingType -> types.stream()
+                        .anyMatch(typeToCheck -> typeToCheck.equals(existingType))))
                 .toList();
 
-        return rowsWithNoEntityOfType.stream().flatMap(this::streamRow).map(TableCell::getEntities).flatMap(Collection::stream);
+        return rowsWithNoEntityOfType.stream()
+                .flatMap(this::streamRow)
+                .map(TableCell::getEntities)
+                .flatMap(Collection::stream);
     }
 
 
     private Stream<String> streamEntityTypesInRow(Integer rowNumber) {
 
-        return streamRow(rowNumber).map(TableCell::getEntities).flatMap(Collection::stream).map(RedactionEntity::getType).distinct();
+        return streamRow(rowNumber).map(TableCell::getEntities)
+                .flatMap(Collection::stream)
+                .map(TextEntity::getType)
+                .distinct();
     }
 
 
@@ -156,7 +182,8 @@ public class Table implements SemanticNode {
             throw new IllegalArgumentException(format("row %d, col %d is out of bounds for number of rows of %d and number of cols %d", row, col, numberOfRows, numberOfCols));
         }
         int idx = row * numberOfCols + col;
-        return (TableCell) documentTree.getEntryById(treeId).getChildren().get(idx).getNode();
+        return (TableCell) documentTree.getEntryById(treeId).getChildren()
+                .get(idx).getNode();
     }
 
 
@@ -193,7 +220,8 @@ public class Table implements SemanticNode {
      */
     public Stream<TableCell> streamCol(int col) {
 
-        return IntStream.range(0, numberOfRows).boxed().map(row -> getCell(row, col));
+        return IntStream.range(0, numberOfRows).boxed()
+                .map(row -> getCell(row, col));
     }
 
 
@@ -205,7 +233,8 @@ public class Table implements SemanticNode {
      */
     public Stream<TableCell> streamRow(int row) {
 
-        return IntStream.range(0, numberOfCols).boxed().map(col -> getCell(row, col));
+        return IntStream.range(0, numberOfCols).boxed()
+                .map(col -> getCell(row, col));
     }
 
 
@@ -229,7 +258,8 @@ public class Table implements SemanticNode {
      */
     public Stream<TableCell> streamHeadersForCell(int row, int col) {
 
-        return Stream.concat(streamRow(row), streamCol(col)).filter(TableCell::isHeader);
+        return Stream.concat(streamRow(row), streamCol(col))
+                .filter(TableCell::isHeader);
     }
 
 
@@ -275,12 +305,12 @@ public class Table implements SemanticNode {
      * Finds all entities of the provided type, which appear in the same row that the provided entity appears in.
      *
      * @param type            the type of entities to search for
-     * @param redactionEntity the entity, which appears in the row to search
+     * @param textEntity the entity, which appears in the row to search
      * @return List of all entities of the provided type, which appear in the same row that the provided entity appears in.
      */
-    public List<RedactionEntity> getEntitiesOfTypeInSameRow(String type, RedactionEntity redactionEntity) {
+    public List<TextEntity> getEntitiesOfTypeInSameRow(String type, TextEntity textEntity) {
 
-        return redactionEntity.getIntersectingNodes()
+        return textEntity.getIntersectingNodes()
                 .stream()
                 .filter(node -> node instanceof TableCell)
                 .map(node -> (TableCell) node)
@@ -302,7 +332,7 @@ public class Table implements SemanticNode {
     public TextBlock getTextBlock() {
 
         if (textBlock == null) {
-            textBlock = streamAllSubNodes().filter(SemanticNode::isLeaf).map(SemanticNode::getLeafTextBlock).collect(new TextBlockCollector());
+            textBlock = SemanticNode.super.getTextBlock();
         }
         return textBlock;
     }
@@ -314,4 +344,20 @@ public class Table implements SemanticNode {
         return treeId.toString() + ": " + NodeType.TABLE + ": #cols: " + numberOfCols + ", #rows: " + numberOfRows + ", " + this.getTextBlock().buildSummary();
     }
 
+
+    @Override
+    public Map<Page, Rectangle2D> getBBox() {
+
+        if (bBoxCache == null) {
+            bBoxCache = SemanticNode.super.getBBox();
+        }
+        return bBoxCache;
+    }
+
+
+    @Override
+    public void accept(NodeVisitor visitor) {
+        visitor.visit(this);
+    }
+
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/graph/nodes/TableCell.java

@@ -2,31 +2,27 @@ package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
 
 import java.awt.geom.Rectangle2D;
 import java.util.HashMap;
-import java.util.HashSet;
-import java.util.List;
 import java.util.Map;
-import java.util.Set;
 
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlockCollector;
 
 import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
-import lombok.Builder;
 import lombok.Data;
 import lombok.EqualsAndHashCode;
 import lombok.experimental.FieldDefaults;
+import lombok.experimental.SuperBuilder;
 
 @Data
-@Builder
+@SuperBuilder
 @AllArgsConstructor
+@EqualsAndHashCode(callSuper = true)
 @FieldDefaults(level = AccessLevel.PRIVATE)
-public class TableCell implements GenericSemanticNode {
+public class TableCell extends AbstractSemanticNode {
 
-    List<Integer> treeId;
     int row;
     int col;
     boolean header;
@@ -37,13 +33,6 @@ public class TableCell implements GenericSemanticNode {
 
     TextBlock textBlock;
 
-    @EqualsAndHashCode.Exclude
-    DocumentTree documentTree;
-
-    @Builder.Default
-    @EqualsAndHashCode.Exclude
-    Set<RedactionEntity> entities = new HashSet<>();
-
 
     @Override
     public Map<Page, Rectangle2D> getBBox() {
@@ -54,6 +43,13 @@ public class TableCell implements GenericSemanticNode {
     }
 
 
+    @Override
+    public void accept(NodeVisitor visitor) {
+
+        visitor.visit(this);
+    }
+
+
     @Override
     public NodeType getType() {
 
@@ -76,16 +72,24 @@ public class TableCell implements GenericSemanticNode {
         }
 
         if (textBlock == null) {
-            textBlock = streamAllSubNodes().filter(SemanticNode::isLeaf).map(SemanticNode::getLeafTextBlock).collect(new TextBlockCollector());
+            textBlock = buildTextBlock();
         }
         return textBlock;
     }
 
 
+    private TextBlock buildTextBlock() {
+
+        return streamAllSubNodes().filter(SemanticNode::isLeaf)
+                .map(SemanticNode::getLeafTextBlock)
+                .collect(new TextBlockCollector());
+    }
+
+
     @Override
     public String toString() {
 
-        return treeId + ": " + NodeType.TABLE_CELL + ": " + this.getTextBlock().buildSummary();
+        return getTreeId() + ": " + NodeType.TABLE_CELL + ": " + this.getTextBlock().buildSummary();
     }
 
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/graph/textblock/AtomicTextBlock.java

@@ -10,10 +10,12 @@ import java.util.Collections;
 import java.util.HashMap;
 import java.util.List;
 import java.util.Map;
+import java.util.Set;
+import java.util.stream.Collectors;
 
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentPositionData;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentTextData;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.Boundary;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.TextRange;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Page;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.SemanticNode;
 import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
@@ -36,14 +38,14 @@ public class AtomicTextBlock implements TextBlock {
     Page page;
 
     //string coordinates
-    Boundary boundary;
+    TextRange textRange;
     String searchText;
     @Builder.Default
     List<Integer> lineBreaks = new ArrayList<>();
     @Builder.Default
-    List<Boundary> boldTextBoundaries = new ArrayList<>();
+    List<TextRange> boldTextBoundaries = new ArrayList<>();
     @Builder.Default
-    List<Boundary> italicTextBoundaries = new ArrayList<>();
+    List<TextRange> italicTextBoundaries = new ArrayList<>();
     String orientation;
     int textDirection;
 
@@ -64,10 +66,44 @@ public class AtomicTextBlock implements TextBlock {
     }
 
 
+    @Override
+    public String subSequenceWithLineBreaks(TextRange stringTextRange) {
+
+        if (stringTextRange.length() == 0 || !getTextRange().contains(stringTextRange)) {
+            return "";
+        }
+
+        Set<Integer> lbInBoundary = lineBreaks.stream()
+                .map(i -> i + this.textRange.start())
+                .filter(stringTextRange::contains)
+                .collect(Collectors.toSet());
+        if (stringTextRange.end() == getTextRange().end()) {
+            lbInBoundary.add(getTextRange().end());
+        }
+        StringBuilder sb = new StringBuilder();
+        for (int i = stringTextRange.start(); i < stringTextRange.end(); i++) {
+            char character = this.charAt(i);
+            if (lbInBoundary.contains(i + 1)) {
+                // always plus one, due to the linebreaks being an exclusive end index
+                if (!Character.isWhitespace(character)) {
+                    lbInBoundary.remove(i + 1);
+                    lbInBoundary.add(i + 2);
+                    sb.append(character);
+                    continue;
+                }
+                sb.append("\n");
+            } else {
+                sb.append(character);
+            }
+        }
+        return sb.toString();
+    }
+
+
     public static AtomicTextBlock fromSearchTextWithTextPosition(String searchText,
                                                                  List<Integer> lineBreaks,
-                                                                 List<Boundary> boldTextBoundaries,
-                                                                 List<Boundary> italicTextBoundaries,
+                                                                 List<TextRange> boldTextBoundaries,
+                                                                 List<TextRange> italicTextBoundaries,
                                                                  List<Rectangle2D> positions,
                                                                  List<Integer> stringIdxToPositionIdx,
                                                                  long idx,
@@ -89,7 +125,7 @@ public class AtomicTextBlock implements TextBlock {
                 .italicTextBoundaries(italicTextBoundaries)
                 .positions(positions)
                 .stringIdxToPositionIdx(stringIdxToPositionIdx)
-                .boundary(new Boundary(offset, offset + searchText.length()))
+                .textRange(new TextRange(offset, offset + searchText.length()))
                 .textDirection(textDirection)
                 .orientation(orientation)
                 .build();
@@ -100,7 +136,7 @@ public class AtomicTextBlock implements TextBlock {
 
         return AtomicTextBlock.builder()
                 .id(textBlockIdx)
-                .boundary(new Boundary(stringOffset, stringOffset))
+                .textRange(new TextRange(stringOffset, stringOffset))
                 .searchText("")
                 .page(page)
                 .numberOnPage(numberOnPage)
@@ -109,19 +145,18 @@ public class AtomicTextBlock implements TextBlock {
     }
 
 
-    public static AtomicTextBlock fromAtomicTextBlockData(DocumentTextData documentTextData,
-                                                          DocumentPositionData documentPositionData,
-                                                          SemanticNode parent,
-                                                          Page page) {
+    public static AtomicTextBlock fromAtomicTextBlockData(DocumentTextData documentTextData, DocumentPositionData documentPositionData, SemanticNode parent, Page page) {
 
         return AtomicTextBlock.builder()
                 .id(documentTextData.getId())
                 .numberOnPage(documentTextData.getNumberOnPage())
                 .page(page)
-                .boundary(new Boundary(documentTextData.getStart(), documentTextData.getEnd()))
+                .textRange(new TextRange(documentTextData.getStart(), documentTextData.getEnd()))
                 .searchText(documentTextData.getSearchText())
-                .lineBreaks(Arrays.stream(documentTextData.getLineBreaks()).boxed().toList())
-                .stringIdxToPositionIdx(Arrays.stream(documentPositionData.getStringIdxToPositionIdx()).boxed().toList())
+                .lineBreaks(Arrays.stream(documentTextData.getLineBreaks()).boxed()
+                                    .toList())
+                .stringIdxToPositionIdx(Arrays.stream(documentPositionData.getStringIdxToPositionIdx()).boxed()
+                                                .toList())
                 .positions(toRectangle2DList(documentPositionData.getPositions()))
                 .parent(parent)
                 .build();
@@ -130,7 +165,9 @@ public class AtomicTextBlock implements TextBlock {
 
     private static List<Rectangle2D> toRectangle2DList(float[][] positions) {
 
-        return Arrays.stream(positions).map(floatArr -> (Rectangle2D) new Rectangle2D.Float(floatArr[0], floatArr[1], floatArr[2], floatArr[3])).toList();
+        return Arrays.stream(positions)
+                .map(floatArr -> (Rectangle2D) new Rectangle2D.Float(floatArr[0], floatArr[1], floatArr[2], floatArr[3]))
+                .toList();
     }
 
 
@@ -140,11 +177,14 @@ public class AtomicTextBlock implements TextBlock {
             throw new IndexOutOfBoundsException(format("line %d out of range for AtomicTextBlock with %d lines", lineNumber, numberOfLines()));
         }
         if (lineNumber == 0) {
-            return subSequence(boundary.start(), lineBreaks.get(0) + boundary.start());
+            if (lineBreaks.isEmpty()) {
+                return searchText;
+            }
+            return subSequence(textRange.start(), lineBreaks.get(0) + textRange.start());
         } else if (lineNumber == numberOfLines() - 1) {
-            return subSequence(lineBreaks.get(lineBreaks.size() - 1) + boundary.start(), boundary.end());
+            return subSequence(lineBreaks.get(lineBreaks.size() - 1) + textRange.start(), textRange.end());
         }
-        return subSequence(lineBreaks.get(lineNumber - 1) + boundary.start(), lineBreaks.get(lineNumber) + boundary.start());
+        return subSequence(lineBreaks.get(lineNumber - 1) + textRange.start(), lineBreaks.get(lineNumber) + textRange.start());
     }
 
 
@@ -159,9 +199,9 @@ public class AtomicTextBlock implements TextBlock {
     public int getNextLinebreak(int fromIndex) {
 
         return lineBreaks.stream()//
-                .filter(linebreak -> linebreak > fromIndex - boundary.start()) //
-                .findFirst() //
-                .orElse(searchText.length()) + boundary.start();
+                       .filter(linebreak -> linebreak > fromIndex - textRange.start()) //
+                       .findFirst() //
+                       .orElse(searchText.length()) + textRange.start();
     }
 
 
@@ -169,43 +209,43 @@ public class AtomicTextBlock implements TextBlock {
     public int getPreviousLinebreak(int fromIndex) {
 
         return lineBreaks.stream()//
-                .filter(linebreak -> linebreak <= fromIndex - boundary.start())//
-                .reduce((a, b) -> b)//
-                .orElse(0) + boundary.start();
+                       .filter(linebreak -> linebreak <= fromIndex - textRange.start())//
+                       .reduce((a, b) -> b)//
+                       .orElse(0) + textRange.start();
     }
 
 
     @Override
     public Rectangle2D getPosition(int stringIdx) {
 
-        return positions.get(stringIdxToPositionIdx.get(stringIdx - boundary.start()));
+        return positions.get(stringIdxToPositionIdx.get(stringIdx - textRange.start()));
     }
 
 
     @Override
-    public List<Rectangle2D> getPositions(Boundary stringBoundary) {
+    public List<Rectangle2D> getPositions(TextRange stringTextRange) {
 
-        if (!containsBoundary(stringBoundary)) {
-            throw new IndexOutOfBoundsException(format("%s is out of bounds for %s", stringBoundary, this.boundary));
+        if (!containsBoundary(stringTextRange)) {
+            throw new IndexOutOfBoundsException(format("%s is out of bounds for %s", stringTextRange, this.textRange));
         }
-        if (stringBoundary.length() == 0) {
+        if (stringTextRange.length() == 0) {
             return Collections.emptyList();
         }
 
-        int startPositionIdx = stringIdxToPositionIdx.get(stringBoundary.start() - this.boundary.start());
+        int startPositionIdx = stringIdxToPositionIdx.get(stringTextRange.start() - this.textRange.start());
 
-        if (stringBoundary.end() == this.boundary.end()) {
+        if (stringTextRange.end() == this.textRange.end()) {
             return positions.subList(startPositionIdx, positions.size());
         }
 
-        return positions.subList(startPositionIdx, stringIdxToPositionIdx.get(stringBoundary.end() - this.boundary.start()));
+        return positions.subList(startPositionIdx, stringIdxToPositionIdx.get(stringTextRange.end() - this.textRange.start()));
 
     }
 
 
-    public Map<Page, List<Rectangle2D>> getPositionsPerPage(Boundary stringBoundary) {
+    public Map<Page, List<Rectangle2D>> getPositionsPerPage(TextRange stringTextRange) {
 
-        List<Rectangle2D> rectanglesPerLine = stringBoundary.split(getAllLineBreaksInBoundary(stringBoundary))
+        List<Rectangle2D> rectanglesPerLine = stringTextRange.split(getAllLineBreaksInBoundary(stringTextRange))
                 .stream()
                 .map(this::getPositions)
                 .map(RectangleTransformations::rectangleBBoxWithGaps)
@@ -217,9 +257,12 @@ public class AtomicTextBlock implements TextBlock {
     }
 
 
-    private List<Integer> getAllLineBreaksInBoundary(Boundary boundary) {
+    protected List<Integer> getAllLineBreaksInBoundary(TextRange textRange) {
 
-        return getLineBreaks().stream().map(linebreak -> linebreak + this.boundary.start()).filter(boundary::contains).toList();
+        return getLineBreaks().stream()
+                .map(linebreak -> linebreak + this.textRange.start())
+                .filter(textRange::contains)
+                .toList();
     }
 
 

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/graph/textblock/ConcatenatedTextBlock.java

@@ -11,7 +11,7 @@ import java.util.List;
 import java.util.Map;
 import java.util.stream.Stream;
 
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.Boundary;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.TextRange;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Page;
 
 import lombok.AccessLevel;
@@ -24,7 +24,7 @@ public class ConcatenatedTextBlock implements TextBlock {
 
     List<AtomicTextBlock> atomicTextBlocks;
     String searchText;
-    Boundary boundary;
+    TextRange textRange;
 
 
     public static ConcatenatedTextBlock empty() {
@@ -37,27 +37,30 @@ public class ConcatenatedTextBlock implements TextBlock {
 
         this.atomicTextBlocks = new LinkedList<>();
         if (atomicTextBlocks.isEmpty()) {
-            boundary = new Boundary(-1, -1);
+            textRange = new TextRange(-1, -1);
             return;
         }
         var firstTextBlock = atomicTextBlocks.get(0);
         this.atomicTextBlocks.add(firstTextBlock);
-        boundary = new Boundary(firstTextBlock.getBoundary().start(), firstTextBlock.getBoundary().end());
+        textRange = new TextRange(firstTextBlock.getTextRange().start(), firstTextBlock.getTextRange().end());
 
-        atomicTextBlocks.subList(1, atomicTextBlocks.size()).forEach(this::concat);
+        atomicTextBlocks.subList(1, atomicTextBlocks.size())
+                .forEach(this::concat);
     }
 
 
     public ConcatenatedTextBlock concat(TextBlock textBlock) {
 
+        int start = textBlock.getTextRange().start();
+        int end = textBlock.getTextRange().end();
         if (this.atomicTextBlocks.isEmpty()) {
-            boundary.setStart(textBlock.getBoundary().start());
-            boundary.setEnd(textBlock.getBoundary().end());
-        } else if (boundary.end() != textBlock.getBoundary().start()) {
-            throw new UnsupportedOperationException(format("Can only concat consecutive TextBlocks, trying to concat %s and %s", boundary, textBlock.getBoundary()));
+            textRange.setStart(start);
+            textRange.setEnd(end);
+        } else if (textRange.end() != start) {
+            throw new UnsupportedOperationException(format("Can only concat consecutive TextBlocks, trying to concat %s and %s", textRange, textBlock.getTextRange()));
         }
         this.atomicTextBlocks.addAll(textBlock.getAtomicTextBlocks());
-        boundary.setEnd(textBlock.getBoundary().end());
+        textRange.setEnd(end);
         this.searchText = null;
         return this;
     }
@@ -65,13 +68,18 @@ public class ConcatenatedTextBlock implements TextBlock {
 
     private AtomicTextBlock getAtomicTextBlockByStringIndex(int stringIdx) {
 
-        return atomicTextBlocks.stream().filter(textBlock -> textBlock.getBoundary().contains(stringIdx)).findAny().orElseThrow(IndexOutOfBoundsException::new);
+        return atomicTextBlocks.stream()
+                .filter(textBlock -> textBlock.getTextRange().containsExclusive(stringIdx))
+                .findAny()
+                .orElseThrow(IndexOutOfBoundsException::new);
     }
 
 
-    private List<AtomicTextBlock> getAllAtomicTextBlocksPartiallyInStringBoundary(Boundary boundary) {
+    private List<AtomicTextBlock> getAllAtomicTextBlocksPartiallyInStringBoundary(TextRange textRange) {
 
-        return atomicTextBlocks.stream().filter(tb -> tb.getBoundary().intersects(boundary)).toList();
+        return atomicTextBlocks.stream()
+                .filter(tb -> tb.getTextRange().intersects(textRange))
+                .toList();
     }
 
 
@@ -90,7 +98,9 @@ public class ConcatenatedTextBlock implements TextBlock {
     @Override
     public int numberOfLines() {
 
-        return atomicTextBlocks.stream().map(AtomicTextBlock::getLineBreaks).mapToInt(List::size).sum();
+        return atomicTextBlocks.stream()
+                .map(AtomicTextBlock::getLineBreaks)
+                .mapToInt(List::size).sum();
     }
 
 
@@ -111,7 +121,10 @@ public class ConcatenatedTextBlock implements TextBlock {
     @Override
     public List<Integer> getLineBreaks() {
 
-        return getAtomicTextBlocks().stream().flatMap(atomicTextBlock -> atomicTextBlock.getLineBreaks().stream()).toList();
+        return getAtomicTextBlocks().stream()
+                .flatMap(atomicTextBlock -> atomicTextBlock.getLineBreaks()
+                        .stream())
+                .toList();
     }
 
 
@@ -123,47 +136,48 @@ public class ConcatenatedTextBlock implements TextBlock {
 
 
     @Override
-    public List<Rectangle2D> getPositions(Boundary stringBoundary) {
+    public List<Rectangle2D> getPositions(TextRange stringTextRange) {
 
-        List<AtomicTextBlock> textBlocks = getAllAtomicTextBlocksPartiallyInStringBoundary(stringBoundary);
+        List<AtomicTextBlock> textBlocks = getAllAtomicTextBlocksPartiallyInStringBoundary(stringTextRange);
 
         if (textBlocks.size() == 1) {
-            return textBlocks.get(0).getPositions(stringBoundary);
+            return textBlocks.get(0).getPositions(stringTextRange);
         }
 
         AtomicTextBlock firstTextBlock = textBlocks.get(0);
-        List<Rectangle2D> positions = new LinkedList<>(firstTextBlock.getPositions(new Boundary(stringBoundary.start(), firstTextBlock.getBoundary().end())));
+        List<Rectangle2D> positions = new LinkedList<>(firstTextBlock.getPositions(new TextRange(stringTextRange.start(), firstTextBlock.getTextRange().end())));
 
         for (AtomicTextBlock textBlock : textBlocks.subList(1, textBlocks.size() - 1)) {
             positions.addAll(textBlock.getPositions());
         }
 
         var lastTextBlock = textBlocks.get(textBlocks.size() - 1);
-        positions.addAll(lastTextBlock.getPositions(new Boundary(lastTextBlock.getBoundary().start(), stringBoundary.end())));
+        positions.addAll(lastTextBlock.getPositions(new TextRange(lastTextBlock.getTextRange().start(), stringTextRange.end())));
 
         return positions;
     }
 
 
     @Override
-    public Map<Page, List<Rectangle2D>> getPositionsPerPage(Boundary stringBoundary) {
+    public Map<Page, List<Rectangle2D>> getPositionsPerPage(TextRange stringTextRange) {
 
-        List<AtomicTextBlock> textBlocks = getAllAtomicTextBlocksPartiallyInStringBoundary(stringBoundary);
+        List<AtomicTextBlock> textBlocks = getAllAtomicTextBlocksPartiallyInStringBoundary(stringTextRange);
 
         if (textBlocks.size() == 1) {
-            return textBlocks.get(0).getPositionsPerPage(stringBoundary);
+            return textBlocks.get(0).getPositionsPerPage(stringTextRange);
         }
 
         AtomicTextBlock firstTextBlock = textBlocks.get(0);
-        Map<Page, List<Rectangle2D>> rectanglesPerLinePerPage = firstTextBlock.getPositionsPerPage(new Boundary(stringBoundary.start(), firstTextBlock.getBoundary().end()));
+        Map<Page, List<Rectangle2D>> rectanglesPerLinePerPage = firstTextBlock.getPositionsPerPage(new TextRange(stringTextRange.start(), firstTextBlock.getTextRange().end()));
 
         for (AtomicTextBlock textBlock : textBlocks.subList(1, textBlocks.size() - 1)) {
-            rectanglesPerLinePerPage = mergeEntityPositionsWithSamePageNode(rectanglesPerLinePerPage, textBlock.getPositionsPerPage(textBlock.getBoundary()));
+            rectanglesPerLinePerPage = mergeEntityPositionsWithSamePageNode(rectanglesPerLinePerPage, textBlock.getPositionsPerPage(textBlock.getTextRange()));
         }
 
         AtomicTextBlock lastTextBlock = textBlocks.get(textBlocks.size() - 1);
         rectanglesPerLinePerPage = mergeEntityPositionsWithSamePageNode(rectanglesPerLinePerPage,
-                lastTextBlock.getPositionsPerPage(new Boundary(lastTextBlock.getBoundary().start(), stringBoundary.end())));
+                                                                        lastTextBlock.getPositionsPerPage(new TextRange(lastTextBlock.getTextRange().start(),
+                                                                                                                        stringTextRange.end())));
 
         return rectanglesPerLinePerPage;
     }
@@ -172,11 +186,42 @@ public class ConcatenatedTextBlock implements TextBlock {
     private Map<Page, List<Rectangle2D>> mergeEntityPositionsWithSamePageNode(Map<Page, List<Rectangle2D>> map1, Map<Page, List<Rectangle2D>> map2) {
 
         Map<Page, List<Rectangle2D>> mergedMap = new HashMap<>(map1);
-        map2.forEach((pageNode, rectangles) -> mergedMap.merge(pageNode, rectangles, (l1, l2) -> Stream.concat(l1.stream(), l2.stream()).toList()));
+        map2.forEach((pageNode, rectangles) -> mergedMap.merge(pageNode,
+                                                               rectangles,
+                                                               (l1, l2) -> Stream.concat(l1.stream(), l2.stream())
+                                                                       .toList()));
         return mergedMap;
     }
 
 
+    @Override
+    public String subSequenceWithLineBreaks(TextRange stringTextRange) {
+
+        if (stringTextRange.length() == 0 || !getTextRange().contains(stringTextRange)) {
+            return "";
+        }
+
+        List<AtomicTextBlock> textBlocks = getAllAtomicTextBlocksPartiallyInStringBoundary(stringTextRange);
+
+        if (textBlocks.size() == 1) {
+            return textBlocks.get(0).subSequenceWithLineBreaks(stringTextRange);
+        }
+
+        StringBuilder sb = new StringBuilder();
+        AtomicTextBlock firstTextBlock = textBlocks.get(0);
+        sb.append(firstTextBlock.subSequenceWithLineBreaks(new TextRange(stringTextRange.start(), firstTextBlock.getTextRange().end())));
+
+        for (AtomicTextBlock textBlock : textBlocks.subList(1, textBlocks.size() - 1)) {
+            sb.append(textBlock.searchTextWithLineBreaks());
+        }
+
+        var lastTextBlock = textBlocks.get(textBlocks.size() - 1);
+        sb.append(lastTextBlock.subSequenceWithLineBreaks(new TextRange(lastTextBlock.getTextRange().start(), stringTextRange.end())));
+
+        return sb.toString();
+    }
+
+
     @Override
     public String toString() {
 
@@ -185,16 +230,22 @@ public class ConcatenatedTextBlock implements TextBlock {
 
 
     @Override
-    public List<Boundary> getBoldTextBoundaries() {
+    public List<TextRange> getBoldTextBoundaries() {
 
-        return getAtomicTextBlocks().stream().map(AtomicTextBlock::getBoldTextBoundaries).flatMap(Collection::stream).toList();
+        return getAtomicTextBlocks().stream()
+                .map(AtomicTextBlock::getBoldTextBoundaries)
+                .flatMap(Collection::stream)
+                .toList();
     }
 
 
     @Override
-    public List<Boundary> getItalicTextBoundaries() {
+    public List<TextRange> getItalicTextBoundaries() {
 
-        return getAtomicTextBlocks().stream().map(AtomicTextBlock::getItalicTextBoundaries).flatMap(Collection::stream).toList();
+        return getAtomicTextBlocks().stream()
+                .map(AtomicTextBlock::getItalicTextBoundaries)
+                .flatMap(Collection::stream)
+                .toList();
     }
 
 

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/graph/textblock/TextBlock.java

@@ -10,7 +10,7 @@ import java.util.Map;
 import java.util.Set;
 import java.util.stream.Collectors;
 
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.Boundary;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.TextRange;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Page;
 
 public interface TextBlock extends CharSequence {
@@ -21,10 +21,10 @@ public interface TextBlock extends CharSequence {
     List<AtomicTextBlock> getAtomicTextBlocks();
 
 
-    List<Boundary> getBoldTextBoundaries();
+    List<TextRange> getBoldTextBoundaries();
 
 
-    List<Boundary> getItalicTextBoundaries();
+    List<TextRange> getItalicTextBoundaries();
 
 
     String getOrientation();
@@ -33,7 +33,7 @@ public interface TextBlock extends CharSequence {
     int getTextDirection();
 
 
-    Boundary getBoundary();
+    TextRange getTextRange();
 
 
     int getNextLinebreak(int fromIndex);
@@ -48,31 +48,41 @@ public interface TextBlock extends CharSequence {
     Rectangle2D getPosition(int stringIdx);
 
 
-    List<Rectangle2D> getPositions(Boundary stringBoundary);
+    List<Rectangle2D> getPositions(TextRange stringTextRange);
 
 
-    Map<Page, List<Rectangle2D>> getPositionsPerPage(Boundary stringBoundary);
+    Map<Page, List<Rectangle2D>> getPositionsPerPage(TextRange stringTextRange);
 
 
     int numberOfLines();
 
 
+    String subSequenceWithLineBreaks(TextRange stringTextRange);
+
+
+    default String searchTextWithLineBreaks() {
+
+        return subSequenceWithLineBreaks(getTextRange());
+    }
+
     default int indexOf(String searchTerm) {
 
-        return indexOf(searchTerm, getBoundary().start());
+        return indexOf(searchTerm, getTextRange().start());
     }
 
 
     default Set<Page> getPages() {
 
-        return getAtomicTextBlocks().stream().map(AtomicTextBlock::getPage).collect(Collectors.toUnmodifiableSet());
+        return getAtomicTextBlocks().stream()
+                .map(AtomicTextBlock::getPage)
+                .collect(Collectors.toUnmodifiableSet());
     }
 
 
-    default Set<Page> getPages(Boundary boundary) {
+    default Set<Page> getPages(TextRange textRange) {
 
         return getAtomicTextBlocks().stream()
-                .filter(atomicTextBlock -> atomicTextBlock.getBoundary().intersects(boundary))
+                .filter(atomicTextBlock -> atomicTextBlock.getTextRange().intersects(textRange))
                 .map(AtomicTextBlock::getPage)
                 .collect(Collectors.toUnmodifiableSet());
     }
@@ -80,38 +90,38 @@ public interface TextBlock extends CharSequence {
 
     default int indexOf(String searchTerm, int startOffset) {
 
-        int start = getSearchText().indexOf(searchTerm, startOffset - getBoundary().start());
+        int start = getSearchText().indexOf(searchTerm, startOffset - getTextRange().start());
         if (start == -1) {
             return -1;
         }
-        return start + getBoundary().start();
+        return start + getTextRange().start();
     }
 
 
     default CharSequence getFirstLine() {
 
-        return subSequence(getBoundary().start(), getNextLinebreak(getBoundary().start()));
+        return subSequence(getTextRange().start(), getNextLinebreak(getTextRange().start()));
     }
 
 
-    default boolean containsBoundary(Boundary boundary) {
+    default boolean containsBoundary(TextRange textRange) {
 
-        if (boundary.end() < boundary.start()) {
-            throw new IllegalArgumentException(format("Invalid %s, StartIndex must be smaller than EndIndex", boundary));
+        if (textRange.end() < textRange.start()) {
+            throw new IllegalArgumentException(format("Invalid %s, StartIndex must be smaller than EndIndex", textRange));
         }
-        return getBoundary().contains(boundary);
+        return getTextRange().contains(textRange);
     }
 
 
     default boolean containsIndex(int stringIndex) {
 
-        return getBoundary().contains(stringIndex);
+        return getTextRange().containsExclusive(stringIndex);
     }
 
 
-    default CharSequence subSequence(Boundary boundary) {
+    default CharSequence subSequence(TextRange textRange) {
 
-        return subSequence(boundary.start(), boundary.end());
+        return subSequence(textRange.start(), textRange.end());
     }
 
 
@@ -128,21 +138,21 @@ public interface TextBlock extends CharSequence {
     @Override
     default CharSequence subSequence(int start, int end) {
 
-        return getSearchText().substring(start - getBoundary().start(), end - getBoundary().start());
+        return getSearchText().substring(start - getTextRange().start(), end - getTextRange().start());
     }
 
 
     @Override
     default int length() {
 
-        return getBoundary().length();
+        return getTextRange().length();
     }
 
 
     @Override
     default char charAt(int index) {
 
-        return getSearchText().charAt(index - getBoundary().start());
+        return getSearchText().charAt(index - getTextRange().start());
     }
 
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/image/ClassifiedImage.java

@@ -4,11 +4,13 @@ import java.awt.geom.Rectangle2D;
 
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.ImageType;
 
+import lombok.AllArgsConstructor;
 import lombok.Data;
 import lombok.NonNull;
 import lombok.RequiredArgsConstructor;
 
 @Data
+@AllArgsConstructor
 @RequiredArgsConstructor
 public class ClassifiedImage {
 
@@ -16,10 +18,20 @@ public class ClassifiedImage {
     private Rectangle2D position;
     @NonNull
     private ImageType imageType;
+    private boolean sourceByAi;
     private boolean isAppendedToSection;
-    @NonNull
     private boolean hasTransparency;
-    @NonNull
     private int page;
+    private String representation;
+
+
+    public ClassifiedImage(@NonNull Rectangle2D position, @NonNull ImageType imageType, boolean hasTransparency, int page, String representation) {
+
+        this.position = position;
+        this.imageType = imageType;
+        this.hasTransparency = hasTransparency;
+        this.page = page;
+        this.representation = representation;
+    }
 
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/outline/OutlineExtractorService.java

@@ -0,0 +1,229 @@
+package com.knecon.fforesight.service.layoutparser.processor.model.outline;
+
+import java.awt.geom.AffineTransform;
+import java.awt.geom.Point2D;
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.List;
+import java.util.Locale;
+import java.util.Optional;
+
+import org.apache.pdfbox.cos.COSArray;
+import org.apache.pdfbox.cos.COSBase;
+import org.apache.pdfbox.cos.COSDictionary;
+import org.apache.pdfbox.cos.COSName;
+import org.apache.pdfbox.cos.COSString;
+import org.apache.pdfbox.pdmodel.PDDestinationNameTreeNode;
+import org.apache.pdfbox.pdmodel.PDDocument;
+import org.apache.pdfbox.pdmodel.PDDocumentNameDictionary;
+import org.apache.pdfbox.pdmodel.PDPage;
+import org.apache.pdfbox.pdmodel.interactive.action.PDAction;
+import org.apache.pdfbox.pdmodel.interactive.documentnavigation.destination.PDDestination;
+import org.apache.pdfbox.pdmodel.interactive.documentnavigation.destination.PDPageFitHeightDestination;
+import org.apache.pdfbox.pdmodel.interactive.documentnavigation.destination.PDPageFitRectangleDestination;
+import org.apache.pdfbox.pdmodel.interactive.documentnavigation.destination.PDPageFitWidthDestination;
+import org.apache.pdfbox.pdmodel.interactive.documentnavigation.destination.PDPageXYZDestination;
+import org.apache.pdfbox.pdmodel.interactive.documentnavigation.outline.PDDocumentOutline;
+import org.apache.pdfbox.pdmodel.interactive.documentnavigation.outline.PDOutlineItem;
+import org.springframework.stereotype.Service;
+
+import com.knecon.fforesight.service.layoutparser.processor.utils.CoordinateTransforms;
+import com.knecon.fforesight.service.layoutparser.processor.utils.PageInformation;
+
+import lombok.SneakyThrows;
+import lombok.extern.slf4j.Slf4j;
+
+@Service
+@Slf4j
+public class OutlineExtractorService {
+
+    private static final String PDDESTINATION_TYPE_FIT = "Fit";
+    private static final String PDDESTINATION_TYPE_FIT_B = "FitB";
+    private static final String PDDESTINATION_TYPE_FIT_H = "FitH";
+    private static final String PDDESTINATION_TYPE_FIT_V = "FitV";
+    private static final String PDDESTINATION_TYPE_FIT_R = "FitR";
+    private static final String PDDESTINATION_TYPE_FIT_BH = "FitBH";
+    private static final String PDDESTINATION_TYPE_FIT_BV = "FitBV";
+    private static final String PDDESTINATION_TYPE_XYZ = "XYZ";
+
+
+    @SneakyThrows
+    public OutlineObjectTree getOutlineObjectTree(PDDocument document) {
+
+        PDDocumentOutline documentOutline = document.getDocumentCatalog().getDocumentOutline();
+
+        List<OutlineObjectTreeNode> rootNodes = new ArrayList<>();
+        if (documentOutline != null) {
+            for (PDOutlineItem child : documentOutline.children()) {
+                Optional<OutlineObjectTreeNode> outlineObjectWithChildren = createOutlineObjectWithChildren(child, document, 1);
+                outlineObjectWithChildren.ifPresent(rootNodes::add);
+            }
+        }
+
+        return new OutlineObjectTree(rootNodes);
+    }
+
+
+    @SneakyThrows
+    private Optional<OutlineObjectTreeNode> createOutlineObjectWithChildren(PDOutlineItem item, PDDocument document, int depth) {
+
+        Optional<OutlineObjectTreeNode> outlineObject = createOutlineObject(item, document, depth);
+        if (outlineObject.isPresent()) {
+            for (var child : item.children()) {
+                Optional<OutlineObjectTreeNode> outlineObjectWithChildren = createOutlineObjectWithChildren(child, document, depth + 1);
+                outlineObjectWithChildren.ifPresent(outlineObjectTreeNode -> outlineObject.get().addChild(outlineObjectTreeNode));
+            }
+        }
+
+        return outlineObject;
+    }
+
+
+    // if the structure elements are processed beforehand, another case can be handled here as well:
+    // outline objects can reference structure elements (see pdf documentation)
+    @SneakyThrows
+    private Optional<OutlineObjectTreeNode> createOutlineObject(PDOutlineItem item, PDDocument document, int depth) {
+
+        String title = item.getTitle();
+
+        PDPage page;
+        try {
+            // Can throw: "Error: can't convert to Destination COSArray" for some OCR'd PDFs
+            page = item.findDestinationPage(document);
+            if (page == null) {
+                return Optional.empty();
+            }
+        } catch (IOException e) {
+            log.info(String.format("Error occurred during position resolution for outline item with title %s: " + e, title));
+            return Optional.empty();
+        }
+
+        int pageNumber = document.getPages().indexOf(page) + 1;
+        AffineTransform userSpaceToPageCoords = CoordinateTransforms.calculateInitialUserSpaceCoordsToPageCoords(PageInformation.fromPDPage(pageNumber, page));
+
+        Optional<Point2D> outlinePosition = Optional.empty();
+
+        try {
+            PDDocumentNameDictionary names = document.getDocumentCatalog().getNames();
+            PDDestinationNameTreeNode destinations = null;
+            if (names != null) {
+                destinations = names.getDests();
+            }
+
+            PDDestination destination = item.getDestination();
+            if (destination != null) {
+                outlinePosition = getLocationFromCOSBase(destinations, destination.getCOSObject());
+            }
+
+            if (outlinePosition.isEmpty()) {
+
+                PDAction action = item.getAction();
+                if (action != null) {
+                    outlinePosition = extractOutlineLocationGoTo(destinations, action.getCOSObject());
+                }
+
+            }
+
+        } catch (Exception e) {
+            log.info(String.format("Error occurred during position resolution for outline item on page %s with title %s: " + e, pageNumber, title));
+        }
+
+        return Optional.of(new OutlineObjectTreeNode(new OutlineObject(title,
+                                                                       pageNumber,
+                                                                       transformPointToPageCoords(outlinePosition, userSpaceToPageCoords), depth)));
+    }
+
+
+    private static Point2D transformPointToPageCoords(Optional<Point2D> outlinePosition, AffineTransform userSpaceToPageCoords) {
+
+        return outlinePosition.map(point -> userSpaceToPageCoords.transform(point, null)).orElse(null);
+    }
+
+
+    @SneakyThrows
+    private static Optional<Point2D> extractOutlineLocationGoTo(PDDestinationNameTreeNode destinations, COSDictionary cosDictionary) {
+
+        if (isGoToAction(cosDictionary)) {
+            COSBase cosBase = cosDictionary.getItem(COSName.D);
+            return getLocationFromCOSBase(destinations, cosBase);
+        }
+
+        return Optional.empty();
+    }
+
+
+    private static Optional<Point2D> getLocationFromCOSBase(PDDestinationNameTreeNode destinations, COSBase cosBase) throws IOException {
+
+        if (cosBase != null) {
+            if (cosBase instanceof COSArray cosArray) {
+                return getLocationFromCosArray(cosArray);
+            }
+
+            if (cosBase instanceof COSString cosString) {
+                String destinationName = cosString.getString();
+                COSArray cosArray = destinations.getValue(destinationName).getCOSObject();
+                return getLocationFromCosArray(cosArray);
+            }
+
+        }
+        return Optional.empty();
+    }
+
+
+    private static Optional<Point2D> getLocationFromCosArray(COSArray cosArray) {
+
+        boolean located = false;
+        float x = 0;
+        float y = 0;
+
+        try {
+
+            PDDestination destination = PDDestination.create(cosArray);
+            COSName type = (COSName) cosArray.getObject(1);
+            String typeString = type.getName();
+
+            switch (typeString) {
+                case PDDESTINATION_TYPE_FIT_V:
+                case PDDESTINATION_TYPE_FIT_BV:
+                    PDPageFitHeightDestination fitHeightDestination = (PDPageFitHeightDestination) destination;
+                    x = fitHeightDestination.getLeft();
+                    located = true;
+                    break;
+                case PDDESTINATION_TYPE_FIT_R:
+                    PDPageFitRectangleDestination fitRectangleDestination = (PDPageFitRectangleDestination) destination;
+                    x = fitRectangleDestination.getLeft();
+                    y = fitRectangleDestination.getTop();
+                    located = true;
+                    break;
+                case PDDESTINATION_TYPE_FIT_H:
+                case PDDESTINATION_TYPE_FIT_BH:
+                    PDPageFitWidthDestination fitWidthDestination = (PDPageFitWidthDestination) destination;
+                    y = fitWidthDestination.getTop();
+                    located = true;
+                    break;
+                case PDDESTINATION_TYPE_XYZ:
+                    PDPageXYZDestination xyzDestination = (PDPageXYZDestination) destination;
+                    x = xyzDestination.getLeft();
+                    y = xyzDestination.getTop();
+                    located = true;
+                    break;
+                case PDDESTINATION_TYPE_FIT:
+                case PDDESTINATION_TYPE_FIT_B:
+                default:
+            }
+
+        } catch (IOException e) {
+            throw new RuntimeException(e);
+        }
+
+        return located ? Optional.of(new Point2D.Float(x, y)) : Optional.empty();
+
+    }
+
+
+    private static boolean isGoToAction(COSDictionary cosDictionary) {
+
+        return cosDictionary.getNameAsString("S").toLowerCase(Locale.ROOT).equals("goto");
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/outline/OutlineObject.java

@@ -0,0 +1,77 @@
+package com.knecon.fforesight.service.layoutparser.processor.model.outline;
+
+import java.awt.geom.Point2D;
+import java.util.Optional;
+
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.BoundingBox;
+
+import lombok.Getter;
+import lombok.Setter;
+
+public class OutlineObject {
+
+    @Getter
+    private final String title;
+    @Getter
+    private final int pageNumber;
+    @Getter
+    private final int treeDepth;
+
+    private Point2D point; // java coordinates, (0, 0) is always top left
+
+    @Getter
+    @Setter
+    private boolean found;
+
+
+    public OutlineObject(String title, int pageNumber, Point2D point2D, int depth) {
+
+        this.title = title;
+        this.pageNumber = pageNumber;
+        this.treeDepth = depth;
+        this.point = point2D;
+    }
+
+
+    @Override
+    public String toString() {
+
+        return "OutlineObject{" + "title='" + title + '\'' + '}';
+    }
+
+
+    public Optional<Point2D> getPoint() {
+
+        return Optional.ofNullable(point);
+    }
+
+
+    public boolean isAbove(BoundingBox boundingBox) {
+
+        if (point == null) {
+            return true;
+        }
+        return point.getY() <= boundingBox.getMaxY();
+    }
+
+
+    public double distance(BoundingBox boundingBox) {
+
+        if (point == null) {
+            return 0;
+        }
+        if (boundingBox.getBBox().contains(point)) {
+            return 0;
+        }
+        double deltaX = Math.min(Math.abs(boundingBox.getMinX() - point.getX()), Math.abs(boundingBox.getMaxX() - point.getX()));
+        double deltaY = Math.min(Math.abs(boundingBox.getMinY() - point.getY()), Math.abs(boundingBox.getMaxY() - point.getY()));
+        return Math.sqrt(deltaX * deltaX + deltaY * deltaY);
+    }
+
+
+    public void resetPoint() {
+
+        this.point = null;
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/outline/OutlineObjectTree.java

@@ -0,0 +1,66 @@
+package com.knecon.fforesight.service.layoutparser.processor.model.outline;
+
+import java.util.ArrayList;
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+
+import lombok.Data;
+import lombok.RequiredArgsConstructor;
+
+@Data
+@RequiredArgsConstructor
+public class OutlineObjectTree {
+
+    private List<OutlineObjectTreeNode> rootNodes = new ArrayList<>();
+
+    private Map<Integer, List<OutlineObject>> outlineObjectsPerPage = new HashMap<>();
+
+
+    public OutlineObjectTree(List<OutlineObjectTreeNode> rootNodes) {
+
+        this.rootNodes = rootNodes;
+        flattenNodesAndGroupByPage(rootNodes);
+    }
+
+
+    private void flattenNodesAndGroupByPage(List<OutlineObjectTreeNode> outlineObjectTreeNodes) {
+
+        for (OutlineObjectTreeNode node : outlineObjectTreeNodes) {
+            int pageNumber = node.getOutlineObject().getPageNumber();
+            if (!this.outlineObjectsPerPage.containsKey(pageNumber)) {
+                outlineObjectsPerPage.put(pageNumber, new ArrayList<>());
+            }
+            outlineObjectsPerPage.get(pageNumber).add(node.getOutlineObject());
+
+            if (!node.getChildren().isEmpty()) {
+                flattenNodesAndGroupByPage(node.getChildren());
+            }
+        }
+    }
+
+    @Override
+    public String toString() {
+
+        StringBuilder sb = new StringBuilder();
+        sb.append("OutlineObjectTree(\n");
+        for (OutlineObjectTreeNode node : rootNodes) {
+            buildString(node, sb, 1);
+        }
+        sb.append(")");
+        return sb.toString();
+    }
+
+    private void buildString(OutlineObjectTreeNode node, StringBuilder sb, int depth) {
+
+        for (int i = 0; i < depth; i++) {
+            sb.append("  ");
+        }
+        sb.append(node.getOutlineObject().getTitle()).append("\n");
+
+        for (OutlineObjectTreeNode child : node.getChildren()) {
+            buildString(child, sb, depth + 1);
+        }
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/outline/OutlineObjectTreeNode.java

@@ -0,0 +1,34 @@
+package com.knecon.fforesight.service.layoutparser.processor.model.outline;
+
+import java.util.ArrayList;
+import java.util.List;
+
+import lombok.Data;
+
+@Data
+public class OutlineObjectTreeNode {
+
+    private OutlineObject outlineObject;
+
+    private List<OutlineObjectTreeNode> children = new ArrayList<>();
+
+
+    public OutlineObjectTreeNode(OutlineObject outlineObject) {
+
+        this.outlineObject = outlineObject;
+    }
+
+
+    public void addChild(OutlineObjectTreeNode outlineObject) {
+
+        children.add(outlineObject);
+    }
+
+
+    @Override
+    public String toString() {
+
+        return "OutlineObjectTreeNode{" + "outlineObject=" + outlineObject + '}';
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/outline/OutlineValidationService.java

@@ -0,0 +1,63 @@
+package com.knecon.fforesight.service.layoutparser.processor.model.outline;
+
+import static com.knecon.fforesight.service.layoutparser.processor.model.PageBlockType.getHeadlineNumber;
+
+import java.util.ArrayList;
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+import java.util.TreeSet;
+
+import org.springframework.stereotype.Service;
+
+import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
+
+import io.micrometer.observation.annotation.Observed;
+import lombok.extern.slf4j.Slf4j;
+
+@Service
+@Slf4j
+public class OutlineValidationService {
+
+    @Observed(name = "OutlineValidationService", contextualName = "create-toc")
+    public TableOfContents createToC(List<TextPageBlock> headlines) {
+
+        List<TableOfContentItem> mainSections = new ArrayList<>();
+        Map<Integer, TableOfContentItem> lastItemsPerDepth = new HashMap<>();
+        TableOfContentItem last = null;
+        TreeSet<Integer> depths = new TreeSet<>();
+
+        for (TextPageBlock current : headlines) {
+            int currentDepth = getHeadlineNumber(current.getClassification());
+            Integer parentDepth = depths.floor(currentDepth - 1);
+
+            var tocItem = new TableOfContentItem(current);
+
+            if (parentDepth == null) {
+                mainSections.add(tocItem);
+                lastItemsPerDepth = new HashMap<>();
+                depths = new TreeSet<>();
+
+            } else {
+                assert last != null;
+                int lastDepth = getHeadlineNumber(last.getHeadline().getClassification());
+
+                if (lastDepth < parentDepth) {
+                    parentDepth = lastDepth;
+                } else if (lastDepth == currentDepth && last.getParent() != null) {
+                    parentDepth = getHeadlineNumber(last.getParent().getHeadline().getClassification());
+                }
+
+                TableOfContentItem parent = lastItemsPerDepth.get(parentDepth);
+                parent.addChild(tocItem);
+            }
+
+            last = tocItem;
+            lastItemsPerDepth.put(currentDepth, tocItem);
+            depths.add(currentDepth);
+        }
+
+        return new TableOfContents(mainSections);
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/outline/TOCEnrichmentService.java

@@ -0,0 +1,252 @@
+package com.knecon.fforesight.service.layoutparser.processor.model.outline;
+
+import java.util.ArrayList;
+import java.util.Collection;
+import java.util.Collections;
+import java.util.Iterator;
+import java.util.List;
+
+import org.springframework.stereotype.Service;
+
+import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
+import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationDocument;
+import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationFooter;
+import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationHeader;
+import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
+import com.knecon.fforesight.service.layoutparser.processor.model.PageBlockType;
+import com.knecon.fforesight.service.layoutparser.processor.model.image.ClassifiedImage;
+import com.knecon.fforesight.service.layoutparser.processor.model.table.Cell;
+import com.knecon.fforesight.service.layoutparser.processor.model.table.TablePageBlock;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
+
+import lombok.extern.slf4j.Slf4j;
+
+@Slf4j
+@Service
+public class TOCEnrichmentService {
+
+    public void assignSectionBlocksAndImages(ClassificationDocument document) {
+
+        TableOfContents toc = document.getTableOfContents();
+        Iterator<TableOfContentItem> iterator = toc.iterator();
+        TableOfContentItem currentTOCItem = null;
+        if (iterator.hasNext()) {
+            currentTOCItem = iterator.next();
+        }
+        List<AbstractPageBlock> startBlocks = new ArrayList<>();
+        List<ClassifiedImage> startImages = new ArrayList<>();
+        TableOfContentItem currentSection = null;
+        boolean foundFirstHeadline = false;
+
+        List<ClassificationHeader> headers = new ArrayList<>();
+        List<ClassificationFooter> footers = new ArrayList<>();
+        TablePageBlock previousTable = null;
+        List<TableOfContentItem> lastFoundTOCItems = new ArrayList<>();
+
+        for (ClassificationPage page : document.getPages()) {
+            List<TableOfContentItem> currentPageTOCItems = new ArrayList<>();
+            List<TextPageBlock> header = new ArrayList<>();
+            List<TextPageBlock> footer = new ArrayList<>();
+            for (AbstractPageBlock current : page.getTextBlocks()) {
+
+                if (current.getClassification() == null) {
+                    continue;
+                }
+
+                current.setPage(page.getPageNumber());
+
+                if (current.getClassification().equals(PageBlockType.HEADER)) {
+                    header.add((TextPageBlock) current);
+                    continue;
+                }
+
+                if (current.getClassification().equals(PageBlockType.FOOTER)) {
+                    footer.add((TextPageBlock) current);
+                    continue;
+                }
+
+                if (current instanceof TablePageBlock table) {
+                    if (previousTable != null) {
+                        mergeTableMetadata(table, previousTable);
+                    }
+                    previousTable = table;
+                }
+
+                if (current instanceof TextPageBlock && currentTOCItem != null && currentTOCItem.getHeadline().getText().equals(current.getText())) {
+                    if (!foundFirstHeadline) {
+                        foundFirstHeadline = true;
+                    }
+                    currentSection = currentTOCItem;
+                    currentTOCItem.getSectionBlocks().add(current);
+                    currentPageTOCItems.add(currentTOCItem);
+
+                    if (iterator.hasNext()) {
+                        currentTOCItem = iterator.next();
+                    }
+                } else if (!foundFirstHeadline) {
+                    startBlocks.add(current);
+                } else {
+                    currentSection.getSectionBlocks().add(current);
+                }
+            }
+
+            if (!currentPageTOCItems.isEmpty()) {
+                lastFoundTOCItems = currentPageTOCItems;
+            }
+
+            for (ClassifiedImage image : page.getImages()) {
+
+                Double xMin = null;
+                Double yMin = null;
+                Double xMax = null;
+                Double yMax = null;
+
+                for (TableOfContentItem tocItem : lastFoundTOCItems) {
+                    var headline = tocItem.getHeadline();
+
+                    if (headline.getPage() != page.getPageNumber()) {
+                        continue;
+                    }
+
+                    if (headline.getMinX() < headline.getMaxX()) {
+                        if (xMin == null || headline.getMinX() < xMin) {
+                            xMin = headline.getMinX();
+                        }
+                        if (xMax == null || headline.getMaxX() > xMax) {
+                            xMax = headline.getMaxX();
+                        }
+                    } else {
+                        if (xMin == null || headline.getMaxX() < xMin) {
+                            xMin = headline.getMaxX();
+                        }
+                        if (xMax == null || headline.getMinX() > xMax) {
+                            xMax = headline.getMinX();
+                        }
+                    }
+
+                    if (headline.getMinY() < headline.getMaxY()) {
+                        if (yMin == null || headline.getMinY() < yMin) {
+                            yMin = headline.getMinY();
+                        }
+                        if (yMax == null || headline.getMaxY() > yMax) {
+                            yMax = headline.getMaxY();
+                        }
+                    } else {
+                        if (yMin == null || headline.getMaxY() < yMin) {
+                            yMin = headline.getMaxY();
+                        }
+                        if (yMax == null || headline.getMinY() > yMax) {
+                            yMax = headline.getMinY();
+                        }
+                    }
+
+                    log.debug("Image position x: {}, y: {}", image.getPosition().getX(), image.getPosition().getY());
+                    log.debug("Headline position xMin: {}, xMax: {}, yMin: {}, yMax: {}", xMin, xMax, yMin, yMax);
+
+                    if (image.getPosition().getX() >= xMin && image.getPosition().getX() <= xMax && image.getPosition().getY() >= yMin && image.getPosition().getY() <= yMax) {
+                        tocItem.getImages().add(image);
+                        image.setAppendedToSection(true);
+                        break;
+                    }
+                }
+                if (!image.isAppendedToSection()) {
+                    log.debug("Image uses last found section");
+                    if (!lastFoundTOCItems.isEmpty()) {
+                        lastFoundTOCItems.get(lastFoundTOCItems.size() - 1).getImages().add(image);
+                    } else {
+                        startImages.add(image);
+                    }
+                    image.setAppendedToSection(true);
+                }
+            }
+
+            if (!header.isEmpty()) {
+                headers.add(new ClassificationHeader(header));
+            }
+            if (!footer.isEmpty()) {
+                footers.add(new ClassificationFooter(footer));
+            }
+        }
+
+        if (!startBlocks.isEmpty() || !startImages.isEmpty()) {
+            TableOfContentItem unassigned = new TableOfContentItem(null);
+            unassigned.setSectionBlocks(startBlocks);
+            unassigned.setImages(startImages);
+            document.getTableOfContents().getMainSections().add(0, unassigned);
+        }
+        document.setHeaders(headers);
+        document.setFooters(footers);
+    }
+
+
+    private void mergeTableMetadata(TablePageBlock currentTable, TablePageBlock previousTable) {
+
+        // Distribute header information for subsequent tables
+        if (previousTable != null && hasInvalidHeaderInformation(currentTable) && hasValidHeaderInformation(previousTable)) {
+            List<Cell> previousTableNonHeaderRow = getRowWithNonHeaderCells(previousTable);
+            List<Cell> tableNonHeaderRow = getRowWithNonHeaderCells(currentTable);
+            // Allow merging of tables if header row is separated from first logical non-header row
+            if (previousTableNonHeaderRow.isEmpty() && previousTable.getRowCount() == 1 && previousTable.getRows().get(0).size() == tableNonHeaderRow.size()) {
+                previousTableNonHeaderRow = previousTable.getRows().get(0)
+                        .stream()
+                        .map(cell -> {
+                            Cell fakeCell = Cell.copy(cell);
+                            fakeCell.setHeaderCells(Collections.singletonList(cell));
+                            return fakeCell;
+                        })
+                        .toList();
+            }
+            if (previousTableNonHeaderRow.size() == tableNonHeaderRow.size()) {
+                for (int i = currentTable.getRowCount() - 1; i >= 0; i--) { // Non header rows are most likely at bottom of table
+                    List<Cell> row = currentTable.getRows().get(i);
+                    if (row.size() == tableNonHeaderRow.size() && row.stream()
+                            .allMatch(cell -> cell.getHeaderCells().isEmpty())) {
+                        for (int j = 0; j < row.size(); j++) {
+                            row.get(j).setHeaderCells(previousTableNonHeaderRow.get(j).getHeaderCells());
+                        }
+                    }
+                }
+            }
+        }
+    }
+
+
+    private boolean hasValidHeaderInformation(TablePageBlock table) {
+
+        return !hasInvalidHeaderInformation(table);
+    }
+
+
+    private boolean hasInvalidHeaderInformation(TablePageBlock table) {
+
+        return table.getRows()
+                .stream()
+                .flatMap(Collection::stream)
+                .allMatch(cell -> cell.getHeaderCells().isEmpty());
+    }
+
+
+    private List<Cell> getRowWithNonHeaderCells(TablePageBlock table) {
+
+        for (int i = table.getRowCount() - 1; i >= 0; i--) { // Non header rows are most likely at bottom of table
+            List<Cell> row = table.getRows().get(i);
+            if (row.size() == 1) {
+                continue;
+            }
+            boolean allNonHeader = true;
+            for (Cell cell : row) {
+                if (cell.isHeaderCell()) {
+                    allNonHeader = false;
+                    break;
+                }
+            }
+            if (allNonHeader) {
+                return row;
+            }
+        }
+
+        return Collections.emptyList();
+
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/outline/TableOfContentItem.java

@@ -0,0 +1,109 @@
+package com.knecon.fforesight.service.layoutparser.processor.model.outline;
+
+import java.util.ArrayList;
+import java.util.List;
+import java.util.stream.Collectors;
+
+import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.GenericSemanticNode;
+import com.knecon.fforesight.service.layoutparser.processor.model.image.ClassifiedImage;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
+
+import lombok.Data;
+import lombok.EqualsAndHashCode;
+
+@Data
+@EqualsAndHashCode(onlyExplicitlyIncluded = true)
+public class TableOfContentItem {
+
+    @EqualsAndHashCode.Include
+    private TextPageBlock headline;
+    private List<TableOfContentItem> children = new ArrayList<>();
+    private TableOfContentItem parent;
+
+    private List<AbstractPageBlock> sectionBlocks = new ArrayList<>();
+    private List<ClassifiedImage> images = new ArrayList<>();
+
+    private GenericSemanticNode section;
+
+
+    public TableOfContentItem(TextPageBlock headline) {
+
+        this.headline = headline;
+    }
+
+
+    public void addChild(TableOfContentItem tableOfContentItem) {
+
+        children.add(tableOfContentItem);
+        tableOfContentItem.setParent(this);
+    }
+
+
+    public TableOfContentItem getSiblingBefore() {
+
+        if (parent != null) {
+            int index = parent.getChildren().indexOf(this);
+            if (index > 0) {
+                return parent.getChildren().get(index - 1);
+            }
+        }
+        return null;
+    }
+
+
+    public TableOfContentItem getSiblingAfter() {
+
+        if (parent != null) {
+            int index = parent.getChildren().indexOf(this);
+            if (index >= 0 && index < parent.getChildren().size() - 1) {
+                return parent.getChildren().get(index + 1);
+            }
+        }
+        return null;
+    }
+
+
+    public boolean contains(TextPageBlock block) {
+
+        if (headline.equals(block)) {
+            return true;
+        }
+        for (TableOfContentItem child : children) {
+            if (child.contains(block)) {
+                return true;
+            }
+        }
+        return false;
+    }
+
+
+    public boolean contains(TableOfContentItem tocItem) {
+
+        if (this.equals(tocItem)) {
+            return true;
+        }
+        for (TableOfContentItem child : children) {
+            if (child.contains(tocItem)) {
+                return true;
+            }
+        }
+        return false;
+    }
+
+
+    public List<AbstractPageBlock> getNonEmptySectionBlocks() {
+
+        return sectionBlocks.stream()
+                .filter(pageBlock -> !pageBlock.isEmpty())
+                .collect(Collectors.toList());
+    }
+
+
+    @Override
+    public String toString() {
+
+        return "OutlineObjectTreeNode{" + "textPageBlock=" + headline + '}';
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/outline/TableOfContents.java

@@ -0,0 +1,136 @@
+package com.knecon.fforesight.service.layoutparser.processor.model.outline;
+
+import java.util.ArrayList;
+import java.util.Iterator;
+import java.util.List;
+import java.util.Stack;
+
+import org.springframework.lang.NonNull;
+
+import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
+
+import lombok.Data;
+import lombok.RequiredArgsConstructor;
+
+@Data
+@RequiredArgsConstructor
+public class TableOfContents implements Iterable<TableOfContentItem> {
+
+    private List<TableOfContentItem> mainSections = new ArrayList<>();
+
+
+    public TableOfContents(List<TableOfContentItem> mainSections) {
+
+        this.mainSections = mainSections;
+    }
+
+
+    public List<TextPageBlock> getAllTextPageBlocks() {
+
+        List<TextPageBlock> allTextPageBlocks = new ArrayList<>();
+        for (TableOfContentItem item : mainSections) {
+            collectTextPageBlocks(item, allTextPageBlocks);
+        }
+        return allTextPageBlocks;
+    }
+
+
+    private void collectTextPageBlocks(TableOfContentItem item, List<TextPageBlock> textPageBlocks) {
+
+        textPageBlocks.add(item.getHeadline());
+        for (TableOfContentItem child : item.getChildren()) {
+            collectTextPageBlocks(child, textPageBlocks);
+        }
+    }
+
+
+    public List<TableOfContentItem> getAllTableOfContentItems() {
+
+        List<TableOfContentItem> allItems = new ArrayList<>();
+        for (TableOfContentItem item : mainSections) {
+            collectTableOfContentItems(item, allItems);
+        }
+        return allItems;
+    }
+
+
+    private void collectTableOfContentItems(TableOfContentItem item, List<TableOfContentItem> allItems) {
+
+        allItems.add(item);
+        for (TableOfContentItem child : item.getChildren()) {
+            collectTableOfContentItems(child, allItems);
+        }
+    }
+
+
+    private boolean containsBlock(TextPageBlock block) {
+
+        for (TableOfContentItem existingItem : this.getMainSections()) {
+            if (existingItem.getHeadline().equals(block) || existingItem.contains(block)) {
+                return true;
+            }
+        }
+        return false;
+    }
+
+
+    private boolean containsItem(TableOfContentItem tocItem) {
+
+        for (TableOfContentItem existingItem : this.getMainSections()) {
+            if (existingItem.equals(tocItem) || existingItem.contains(tocItem)) {
+                return true;
+            }
+        }
+        return false;
+    }
+
+
+    @Override
+    public @NonNull Iterator<TableOfContentItem> iterator() {
+
+        return new TableOfContentItemIterator(mainSections);
+    }
+
+
+    private static class TableOfContentItemIterator implements Iterator<TableOfContentItem> {
+
+        private final Stack<Iterator<TableOfContentItem>> stack = new Stack<>();
+
+
+        TableOfContentItemIterator(List<TableOfContentItem> mainSections) {
+
+            stack.push(mainSections.iterator());
+        }
+
+
+        @Override
+        public boolean hasNext() {
+
+            ensureStackTopIsCurrent();
+            return !stack.isEmpty() && stack.peek().hasNext();
+        }
+
+
+        @Override
+        public TableOfContentItem next() {
+
+            ensureStackTopIsCurrent();
+            TableOfContentItem currentItem = stack.peek().next();
+            if (currentItem.getChildren() != null && !currentItem.getChildren().isEmpty()) {
+                stack.push(currentItem.getChildren()
+                                   .iterator());
+            }
+            return currentItem;
+        }
+
+
+        private void ensureStackTopIsCurrent() {
+
+            while (!stack.isEmpty() && !stack.peek().hasNext()) {
+                stack.pop();
+            }
+        }
+
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/table/Cell.java

@@ -1,10 +1,13 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.table;
 
+import java.awt.geom.AffineTransform;
 import java.awt.geom.Point2D;
+import java.awt.geom.Rectangle2D;
 import java.util.ArrayList;
 import java.util.Iterator;
 import java.util.List;
 
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.BoundingBox;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
 import com.knecon.fforesight.service.layoutparser.processor.utils.TextNormalizationUtilities;
@@ -17,7 +20,7 @@ import lombok.NoArgsConstructor;
 @Data
 @EqualsAndHashCode(callSuper = true)
 @NoArgsConstructor
-public class Cell extends Rectangle {
+public class Cell extends BoundingBox {
 
     private List<TextPageBlock> textBlocks = new ArrayList<>();
 
@@ -32,7 +35,24 @@ public class Cell extends Rectangle {
 
     public Cell(Point2D topLeft, Point2D bottomRight) {
 
-        super((float) topLeft.getY(), (float) topLeft.getX(), (float) (bottomRight.getX() - topLeft.getX()), (float) (bottomRight.getY() - topLeft.getY()));
+        this.bBoxPdf = new Rectangle2D.Double(topLeft.getX(), topLeft.getY(), (bottomRight.getX() - topLeft.getX()), (bottomRight.getY() - topLeft.getY()));
+        this.bBox = bBoxPdf;
+    }
+
+
+    public Cell(Rectangle2D bBoxInitialUserSpace, AffineTransform initialUserSpaceToJava) {
+
+        this.bBoxPdf = bBoxInitialUserSpace;
+        this.bBox = initialUserSpaceToJava.createTransformedShape(bBoxInitialUserSpace).getBounds2D();
+    }
+
+
+    public static Cell copy(Cell cell) {
+
+        Cell copy = new Cell();
+        copy.bBoxPdf = cell.bBoxPdf;
+        copy.bBox = cell.bBox;
+        return copy;
     }
 
 
@@ -67,7 +87,7 @@ public class Cell extends Rectangle {
 
         }
 
-        return TextNormalizationUtilities.removeHyphenLineBreaks(sb.toString()).replaceAll("\n", " ").replaceAll(" {2}", " ");
+        return TextNormalizationUtilities.cleanString(sb.toString());
     }
 
 

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/table/CleanRulings.java

@@ -1,15 +1,206 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.table;
 
+import java.awt.geom.Line2D;
+import java.awt.geom.Point2D;
+import java.awt.geom.Rectangle2D;
+import java.util.ArrayList;
+import java.util.Collection;
+import java.util.Collections;
+import java.util.Comparator;
+import java.util.LinkedList;
 import java.util.List;
+import java.util.stream.Stream;
 
-import lombok.Builder;
-import lombok.Data;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.BoundingBox;
 
-@Data
-@Builder
+import lombok.Getter;
+
+@Getter
 public class CleanRulings {
 
-    List<Ruling> horizontal;
-    List<Ruling> vertical;
+    List<Ruling> horizontals; // unmodifiable sorted by Y list
+    List<Ruling> verticals; // unmodifiable sorted by X list
+
+
+    public CleanRulings(List<Ruling> horizontals, List<Ruling> verticals) {
+
+        this.horizontals = horizontals.stream()
+                .peek(Ruling::assertHorizontal)
+                .sorted(Comparator.comparing(Line2D.Float::getY1))
+                .toList();
+        this.verticals = verticals.stream()
+                .peek(Ruling::assertVertical)
+                .sorted(Comparator.comparing(Line2D.Float::getX1))
+                .toList();
+    }
+
+
+    public CleanRulings getTableLines() {
+
+        return new CleanRulings(horizontals.stream()
+                                        .filter(ruling -> ruling.getClassification().equals(Ruling.Classification.TABLE_LINE))
+                                        .toList(),
+                                verticals.stream()
+                                        .filter(ruling -> ruling.getClassification().equals(Ruling.Classification.TABLE_LINE))
+                                        .toList());
+    }
+
+
+    public CleanRulings withoutTextRulings() {
+
+        return new CleanRulings(horizontals.stream()
+                                        .filter(ruling -> !(ruling.getClassification().equals(Ruling.Classification.UNDERLINE) || ruling.getClassification()
+                                                .equals(Ruling.Classification.STRIKETROUGH)))
+                                        .toList(),
+                                verticals.stream()
+                                        .filter(ruling -> !(ruling.getClassification().equals(Ruling.Classification.UNDERLINE) || ruling.getClassification()
+                                                .equals(Ruling.Classification.STRIKETROUGH)))
+                                        .toList());
+    }
+
+
+    public List<Ruling> buildAll() {
+
+        ArrayList<Ruling> rulings = new ArrayList<>(horizontals.size() + verticals.size());
+        rulings.addAll(horizontals);
+        rulings.addAll(verticals);
+        return rulings;
+    }
+
+
+    public boolean lineBetween(BoundingBox a, BoundingBox b) {
+
+        return lineBetween(a.getBBoxPdf(), b.getBBoxPdf());
+    }
+
+
+    public boolean lineBetween(Rectangle2D a, Rectangle2D b) {
+
+        return lineBetween(new Point2D.Double(a.getCenterX(), a.getCenterY()), new Point2D.Double(b.getCenterX(), b.getCenterY()));
+    }
+
+
+    public boolean lineBetween(Point2D p1, Point2D p2) {
+
+        Ruling ruling = new Ruling(p1, p2);
+
+        if (ruling.isHorizontal()) {
+            return getVerticalsInXInterval(ruling.x1, ruling.x2).stream()
+                    .anyMatch(vertical -> vertical.intersectsLine(ruling));
+
+        }
+
+        if (ruling.isVertical()) {
+            return getHorizontalsInYInterval(ruling.y1, ruling.y2).stream()
+                    .anyMatch(horizontal -> horizontal.intersectsLine(ruling));
+
+        }
+
+        return Stream.of(getVerticalsInXInterval(ruling.x1, ruling.x2), getHorizontalsInYInterval(ruling.y1, ruling.y2))
+                .flatMap(Collection::stream)
+                .anyMatch(other -> other.intersectsLine(ruling));
+    }
+
+
+    public List<Ruling> getHorizontalsInYInterval(float y1, float y2) {
+
+        float startY = Math.min(y1, y2);
+        float endY = Math.max(y1, y2);
+
+        if (horizontals.isEmpty() || Float.isNaN(startY) || Float.isNaN(endY)) {
+            return Collections.emptyList();
+        }
+
+        int firstGreaterThanIdx = findFirstHorizontalRulingIdxAbove(startY);
+
+        if (firstGreaterThanIdx == -1) {
+            return Collections.emptyList();
+        }
+
+        List<Ruling> result = new LinkedList<>();
+        for (int i = firstGreaterThanIdx; i < horizontals.size(); i++) {
+            Ruling horizontal = horizontals.get(i);
+            if (horizontal.y1 > endY) {
+                break;
+            }
+            result.add(horizontal);
+        }
+        return result;
+    }
+
+
+    private int findFirstHorizontalRulingIdxAbove(float y) {
+
+        int low = 0;
+        int high = horizontals.size() - 1;
+
+        while (low <= high) {
+            int mid = low + (high - low) / 2;
+            Line2D.Float midLine = horizontals.get(mid);
+            float midY = midLine.y1;
+
+            if (midY == y) {
+                return mid;
+            } else if (midY > y) {
+                high = mid - 1;
+            } else {
+                low = mid + 1;
+            }
+        }
+
+        // Return the index of the first element greater than y or -1 if not found
+        return horizontals.size() > low && horizontals.get(low).y1 > y ? low : -1;
+    }
+
+
+    public List<Ruling> getVerticalsInXInterval(float x1, float x2) {
+
+        float startX = Math.min(x1, x2);
+        float endX = Math.max(x1, x2);
+
+        if (verticals.isEmpty() || Float.isNaN(startX) || Float.isNaN(endX)) {
+            return Collections.emptyList();
+        }
+
+        int firstGreaterThanIdx = findFirstVerticalRulingIdxRightOf(startX);
+
+        if (firstGreaterThanIdx == -1) {
+            return Collections.emptyList();
+        }
+
+        List<Ruling> result = new LinkedList<>();
+        for (int i = firstGreaterThanIdx; i < verticals.size(); i++) {
+            Ruling horizontal = verticals.get(i);
+            if (horizontal.x1 > endX) {
+                break;
+            }
+            result.add(horizontal);
+        }
+        return result;
+    }
+
+
+    private int findFirstVerticalRulingIdxRightOf(float x) {
+
+        int low = 0;
+        int high = verticals.size() - 1;
+
+        while (low <= high) {
+            int mid = low + (high - low) / 2;
+            Line2D.Float midLine = verticals.get(mid);
+            float midX = midLine.x1;
+
+            if (midX == x) {
+                return mid;
+            } else if (midX > x) {
+                high = mid - 1;
+            } else {
+                low = mid + 1;
+            }
+        }
+
+        // Return the index of the first element greater than y or -1 if not found
+        return verticals.size() > low && verticals.get(low).x1 > x ? low : -1;
+    }
 
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/table/Rectangle.java

@@ -1,218 +0,0 @@
-package com.knecon.fforesight.service.layoutparser.processor.model.table;
-
-import java.awt.geom.Point2D;
-import java.awt.geom.Rectangle2D;
-import java.util.Comparator;
-import java.util.List;
-
-@SuppressWarnings("all")
-public class Rectangle extends Rectangle2D.Float {
-
-    protected static final float VERTICAL_COMPARISON_THRESHOLD = 0.4f;
-    /**
-     * Ill-defined comparator, from when Rectangle was Comparable.
-     * <p>
-     * see https://github.com/tabulapdf/tabula-java/issues/116
-     *
-     * @deprecated with no replacement
-     */
-    @Deprecated
-    public static final Comparator<Rectangle> ILL_DEFINED_ORDER = new Comparator<Rectangle>() {
-        @Override
-        public int compare(Rectangle o1, Rectangle o2) {
-
-            if (o1.equals(o2)) {
-                return 0;
-            }
-            if (o1.verticalOverlap(o2) > VERTICAL_COMPARISON_THRESHOLD) {
-                return o1.isLtrDominant() == -1 && o2.isLtrDominant() == -1 ? -java.lang.Double.compare(o1.getX(), o2.getX()) : java.lang.Double.compare(o1.getX(), o2.getX());
-            } else {
-                return java.lang.Float.compare(o1.getBottom(), o2.getBottom());
-            }
-        }
-    };
-
-
-    public Rectangle() {
-
-        super();
-    }
-
-
-    public Rectangle(float top, float left, float width, float height) {
-
-        super();
-        this.setRect(left, top, width, height);
-    }
-
-
-    /**
-     * @param rectangles
-     * @return minimum bounding box that contains all the rectangles
-     */
-    public static Rectangle boundingBoxOf(List<? extends Rectangle> rectangles) {
-
-        float minx = java.lang.Float.MAX_VALUE;
-        float miny = java.lang.Float.MAX_VALUE;
-        float maxx = java.lang.Float.MIN_VALUE;
-        float maxy = java.lang.Float.MIN_VALUE;
-
-        for (Rectangle r : rectangles) {
-            minx = (float) Math.min(r.getMinX(), minx);
-            miny = (float) Math.min(r.getMinY(), miny);
-            maxx = (float) Math.max(r.getMaxX(), maxx);
-            maxy = (float) Math.max(r.getMaxY(), maxy);
-        }
-        return new Rectangle(miny, minx, maxx - minx, maxy - miny);
-    }
-
-
-    public int compareTo(Rectangle other) {
-
-        return ILL_DEFINED_ORDER.compare(this, other);
-    }
-
-
-    // I'm bad at Java and need this for fancy sorting in
-    // technology.tabula.TextChunk.
-    public int isLtrDominant() {
-
-        return 0;
-    }
-
-
-    public float getArea() {
-
-        return this.width * this.height;
-    }
-
-
-    public float verticalOverlap(Rectangle other) {
-
-        return Math.max(0, Math.min(this.getBottom(), other.getBottom()) - Math.max(this.getTop(), other.getTop()));
-    }
-
-
-    public boolean verticallyOverlaps(Rectangle other) {
-
-        return verticalOverlap(other) > 0;
-    }
-
-
-    public float horizontalOverlap(Rectangle other) {
-
-        return Math.max(0, Math.min(this.getRight(), other.getRight()) - Math.max(this.getLeft(), other.getLeft()));
-    }
-
-
-    public boolean horizontallyOverlaps(Rectangle other) {
-
-        return horizontalOverlap(other) > 0;
-    }
-
-
-    public float verticalOverlapRatio(Rectangle other) {
-
-        float rv = 0, delta = Math.min(this.getBottom() - this.getTop(), other.getBottom() - other.getTop());
-
-        if (other.getTop() <= this.getTop() && this.getTop() <= other.getBottom() && other.getBottom() <= this.getBottom()) {
-            rv = (other.getBottom() - this.getTop()) / delta;
-        } else if (this.getTop() <= other.getTop() && other.getTop() <= this.getBottom() && this.getBottom() <= other.getBottom()) {
-            rv = (this.getBottom() - other.getTop()) / delta;
-        } else if (this.getTop() <= other.getTop() && other.getTop() <= other.getBottom() && other.getBottom() <= this.getBottom()) {
-            rv = (other.getBottom() - other.getTop()) / delta;
-        } else if (other.getTop() <= this.getTop() && this.getTop() <= this.getBottom() && this.getBottom() <= other.getBottom()) {
-            rv = (this.getBottom() - this.getTop()) / delta;
-        }
-
-        return rv;
-
-    }
-
-
-    public float overlapRatio(Rectangle other) {
-
-        double intersectionWidth = Math.max(0, Math.min(this.getRight(), other.getRight()) - Math.max(this.getLeft(), other.getLeft()));
-        double intersectionHeight = Math.max(0, Math.min(this.getBottom(), other.getBottom()) - Math.max(this.getTop(), other.getTop()));
-        double intersectionArea = Math.max(0, intersectionWidth * intersectionHeight);
-        double unionArea = this.getArea() + other.getArea() - intersectionArea;
-
-        return (float) (intersectionArea / unionArea);
-    }
-
-
-    public Rectangle merge(Rectangle other) {
-
-        this.setRect(this.createUnion(other));
-        return this;
-    }
-
-
-    public float getTop() {
-
-        return (float) this.getMinY();
-    }
-
-
-    public void setTop(float top) {
-
-        float deltaHeight = top - this.y;
-        this.setRect(this.x, top, this.width, this.height - deltaHeight);
-    }
-
-
-    public float getRight() {
-
-        return (float) this.getMaxX();
-    }
-
-
-    public void setRight(float right) {
-
-        this.setRect(this.x, this.y, right - this.x, this.height);
-    }
-
-
-    public float getLeft() {
-
-        return (float) this.getMinX();
-    }
-
-
-    public void setLeft(float left) {
-
-        float deltaWidth = left - this.x;
-        this.setRect(left, this.y, this.width - deltaWidth, this.height);
-    }
-
-
-    public float getBottom() {
-
-        return (float) this.getMaxY();
-    }
-
-
-    public void setBottom(float bottom) {
-
-        this.setRect(this.x, this.y, this.width, bottom - this.y);
-    }
-
-
-    public Point2D[] getPoints() {
-
-        return new Point2D[]{new Point2D.Float(this.getLeft(), this.getTop()), new Point2D.Float(this.getRight(), this.getTop()), new Point2D.Float(this.getRight(),
-                this.getBottom()), new Point2D.Float(this.getLeft(), this.getBottom())};
-    }
-
-
-    @Override
-    public String toString() {
-
-        StringBuilder sb = new StringBuilder();
-        String s = super.toString();
-        sb.append(s.substring(0, s.length() - 1));
-        sb.append(String.format(",bottom=%f,right=%f]", this.getBottom(), this.getRight()));
-        return sb.toString();
-    }
-
-}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/table/Ruling.java

@@ -4,28 +4,41 @@ import java.awt.geom.Line2D;
 import java.awt.geom.Point2D;
 import java.awt.geom.Rectangle2D;
 import java.util.ArrayList;
-import java.util.Collections;
-import java.util.Comparator;
 import java.util.Formatter;
 import java.util.List;
-import java.util.Map;
-import java.util.TreeMap;
 
 import com.knecon.fforesight.service.layoutparser.processor.utils.CohenSutherlandClipping;
 import com.knecon.fforesight.service.layoutparser.processor.utils.DoubleComparisons;
 
+import lombok.Getter;
+import lombok.Setter;
 import lombok.extern.slf4j.Slf4j;
 
 @Slf4j
 @SuppressWarnings("all")
 public class Ruling extends Line2D.Float {
 
-    private static int PERPENDICULAR_PIXEL_EXPAND_AMOUNT = 2;
+    public static final int PERPENDICULAR_UNIT_EXPAND_AMOUNT = 2;
+    public static final int COLINEAR_OR_PARALLEL_UNIT_EXPAND_AMOUNT = 2;
+
+    public enum Classification {
+        TABLE_LINE,
+        UNDERLINE,
+        STRIKETROUGH,
+        HEADER_SEPARATOR,
+        FOOTER_SEPARATOR,
+        OTHER
+    }
+
+    @Getter
+    @Setter
+    private Classification classification;
 
 
     public Ruling(Point2D p1, Point2D p2) {
 
         super(p1, p2);
+        this.classification = Classification.OTHER;
     }
 
 
@@ -59,126 +72,32 @@ public class Ruling extends Line2D.Float {
     }
 
 
-    // log(n) implementation of find_intersections
-    // based on http://people.csail.mit.edu/indyk/6.838-old/handouts/lec2.pdf
-    public static Map<Point2D, Ruling[]> findIntersections(List<Ruling> horizontals, List<Ruling> verticals) {
-
-        class SortObject {
-
-            protected SOType type;
-            protected float position;
-            protected Ruling ruling;
-
-
-            public SortObject(SOType type, float position, Ruling ruling) {
-
-                this.type = type;
-                this.position = position;
-                this.ruling = ruling;
-            }
+    public void assertHorizontal() {
 
+        if (isHorizontal()) {
+            return;
         }
-
-        List<SortObject> sos = new ArrayList<>();
-
-        TreeMap<Ruling, Boolean> tree = new TreeMap<>(new Comparator<Ruling>() {
-            @Override
-            public int compare(Ruling o1, Ruling o2) {
-
-                return java.lang.Double.compare(o1.getTop(), o2.getTop());
-            }
-        });
-
-        TreeMap<Point2D, Ruling[]> rv = new TreeMap<>(new Comparator<Point2D>() {
-            @Override
-            public int compare(Point2D o1, Point2D o2) {
-
-                if (o1.getY() > o2.getY()) {
-                    return 1;
-                }
-                if (o1.getY() < o2.getY()) {
-                    return -1;
-                }
-                if (o1.getX() > o2.getX()) {
-                    return 1;
-                }
-                if (o1.getX() < o2.getX()) {
-                    return -1;
-                }
-                return 0;
-            }
-        });
-
-        for (Ruling h : horizontals) {
-            sos.add(new SortObject(SOType.HLEFT, h.getLeft() - PERPENDICULAR_PIXEL_EXPAND_AMOUNT, h));
-            sos.add(new SortObject(SOType.HRIGHT, h.getRight() + PERPENDICULAR_PIXEL_EXPAND_AMOUNT, h));
-        }
-
-        for (Ruling v : verticals) {
-            sos.add(new SortObject(SOType.VERTICAL, v.getLeft(), v));
-        }
-
-        Collections.sort(sos, new Comparator<SortObject>() {
-            @Override
-            public int compare(SortObject a, SortObject b) {
-
-                int rv;
-                if (DoubleComparisons.feq(a.position, b.position)) {
-                    if (a.type == SOType.VERTICAL && b.type == SOType.HLEFT) {
-                        rv = 1;
-                    } else if (a.type == SOType.VERTICAL && b.type == SOType.HRIGHT) {
-                        rv = -1;
-                    } else if (a.type == SOType.HLEFT && b.type == SOType.VERTICAL) {
-                        rv = -1;
-                    } else if (a.type == SOType.HRIGHT && b.type == SOType.VERTICAL) {
-                        rv = 1;
-                    } else {
-                        rv = java.lang.Double.compare(a.position, b.position);
-                    }
-                } else {
-                    return java.lang.Double.compare(a.position, b.position);
-                }
-                return rv;
-            }
-        });
-
-        for (SortObject so : sos) {
-            switch (so.type) {
-                case VERTICAL:
-                    for (Map.Entry<Ruling, Boolean> h : tree.entrySet()) {
-                        try {
-                            Point2D i = h.getKey().intersectionPoint(so.ruling);
-                            if (i == null) {
-                                continue;
-                            }
-                            rv.put(i, new Ruling[]{h.getKey().expand(PERPENDICULAR_PIXEL_EXPAND_AMOUNT), so.ruling.expand(PERPENDICULAR_PIXEL_EXPAND_AMOUNT)});
-                        } catch (UnsupportedOperationException e) {
-                            log.info("Some line are oblique, ignoring...");
-                            continue;
-                        }
-                    }
-                    break;
-                case HRIGHT:
-                    tree.remove(so.ruling);
-                    break;
-                case HLEFT:
-                    tree.put(so.ruling, true);
-                    break;
-            }
-        }
-
-        return rv;
+        throw new IllegalArgumentException("Ruling " + this + " is not horizontal");
 
     }
 
 
-    public boolean vertical() {
+    public void assertVertical() {
+
+        if (isVertical()) {
+            return;
+        }
+        throw new IllegalArgumentException("Ruling " + this + " is not vertical");
+    }
+
+
+    public boolean isVertical() {
 
         return this.length() > 0 && DoubleComparisons.feq(this.x1, this.x2); //diff < ORIENTATION_CHECK_THRESHOLD;
     }
 
 
-    public boolean horizontal() {
+    public boolean isHorizontal() {
 
         return this.length() > 0 && DoubleComparisons.feq(this.y1, this.y2); //diff < ORIENTATION_CHECK_THRESHOLD;
     }
@@ -187,36 +106,36 @@ public class Ruling extends Line2D.Float {
     // these are used to have a single collapse method (in page, currently)
 
 
-    public boolean oblique() {
+    public boolean isOblique() {
 
-        return !(this.vertical() || this.horizontal());
+        return !(this.isVertical() || this.isHorizontal());
     }
 
 
     public float getPosition() {
 
-        if (this.oblique()) {
+        if (this.isOblique()) {
             throw new UnsupportedOperationException();
         }
-        return this.vertical() ? this.getLeft() : this.getTop();
+        return this.isVertical() ? this.getLeft() : this.getTop();
     }
 
 
     public float getStart() {
 
-        if (this.oblique()) {
+        if (this.isOblique()) {
             throw new UnsupportedOperationException();
         }
-        return this.vertical() ? this.getTop() : this.getLeft();
+        return this.isVertical() ? this.getTop() : this.getLeft();
     }
 
 
     public void setStart(float v) {
 
-        if (this.oblique()) {
+        if (this.isOblique()) {
             throw new UnsupportedOperationException();
         }
-        if (this.vertical()) {
+        if (this.isVertical()) {
             this.setTop(v);
         } else {
             this.setLeft(v);
@@ -226,19 +145,19 @@ public class Ruling extends Line2D.Float {
 
     public float getEnd() {
 
-        if (this.oblique()) {
+        if (this.isOblique()) {
             throw new UnsupportedOperationException();
         }
-        return this.vertical() ? this.getBottom() : this.getRight();
+        return this.isVertical() ? this.getBottom() : this.getRight();
     }
 
 
     public void setEnd(float v) {
 
-        if (this.oblique()) {
+        if (this.isOblique()) {
             throw new UnsupportedOperationException();
         }
-        if (this.vertical()) {
+        if (this.isVertical()) {
             this.setBottom(v);
         } else {
             this.setRight(v);
@@ -248,10 +167,10 @@ public class Ruling extends Line2D.Float {
 
     public void setStartEnd(float start, float end) {
 
-        if (this.oblique()) {
+        if (this.isOblique()) {
             throw new UnsupportedOperationException();
         }
-        if (this.vertical()) {
+        if (this.isVertical()) {
             this.setTop(start);
             this.setBottom(end);
         } else {
@@ -263,11 +182,11 @@ public class Ruling extends Line2D.Float {
 
     public boolean perpendicularTo(Ruling other) {
 
-        return this.vertical() == other.horizontal();
+        return this.isVertical() == other.isHorizontal();
     }
 
 
-    public boolean nearlyIntersects(Ruling another, int colinearOrParallelExpandAmount) {
+    public boolean nearlyIntersects(Ruling another) {
 
         if (this.intersectsLine(another)) {
             return true;
@@ -276,9 +195,9 @@ public class Ruling extends Line2D.Float {
         boolean rv = false;
 
         if (this.perpendicularTo(another)) {
-            rv = this.expand(PERPENDICULAR_PIXEL_EXPAND_AMOUNT).intersectsLine(another);
+            rv = this.expand(PERPENDICULAR_UNIT_EXPAND_AMOUNT).intersectsLine(another);
         } else {
-            rv = this.expand(colinearOrParallelExpandAmount).intersectsLine(another.expand(colinearOrParallelExpandAmount));
+            rv = this.expand(COLINEAR_OR_PARALLEL_UNIT_EXPAND_AMOUNT).intersectsLine(another.expand(COLINEAR_OR_PARALLEL_UNIT_EXPAND_AMOUNT));
         }
 
         return rv;
@@ -317,30 +236,6 @@ public class Ruling extends Line2D.Float {
     }
 
 
-    public Point2D intersectionPoint(Ruling other) {
-
-        Ruling this_l = this.expand(PERPENDICULAR_PIXEL_EXPAND_AMOUNT);
-        Ruling other_l = other.expand(PERPENDICULAR_PIXEL_EXPAND_AMOUNT);
-        Ruling horizontal, vertical;
-
-        if (!this_l.intersectsLine(other_l)) {
-            return null;
-        }
-
-        if (this_l.horizontal() && other_l.vertical()) {
-            horizontal = this_l;
-            vertical = other_l;
-        } else if (this_l.vertical() && other_l.horizontal()) {
-            vertical = this_l;
-            horizontal = other_l;
-        } else {
-            log.warn("lines must be orthogonal, vertical and horizontal");
-            return null;
-        }
-        return new Point2D.Float(vertical.getLeft(), horizontal.getTop());
-    }
-
-
     @Override
     public boolean equals(Object other) {
 
@@ -450,16 +345,9 @@ public class Ruling extends Line2D.Float {
 
         final float TOLERANCE = 1;
         return Math.abs(ruling.getX1() - x1) < TOLERANCE &&//
-                Math.abs(ruling.getY1() - y1) < TOLERANCE &&//
-                Math.abs(ruling.getX2() - x2) < TOLERANCE &&//
-                Math.abs(ruling.getY2() - y2) < TOLERANCE;
-    }
-
-
-    private enum SOType {
-        VERTICAL,
-        HRIGHT,
-        HLEFT
+               Math.abs(ruling.getY1() - y1) < TOLERANCE &&//
+               Math.abs(ruling.getX2() - x2) < TOLERANCE &&//
+               Math.abs(ruling.getY2() - y2) < TOLERANCE;
     }
 
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/table/TablePageBlock.java

@@ -3,6 +3,7 @@ package com.knecon.fforesight.service.layoutparser.processor.model.table;
 import java.awt.geom.Point2D;
 import java.util.ArrayList;
 import java.util.Collections;
+import java.util.Comparator;
 import java.util.HashSet;
 import java.util.List;
 import java.util.Set;
@@ -11,6 +12,7 @@ import java.util.TreeMap;
 import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.PageBlockType;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
+import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
 
 import lombok.Getter;
 import lombok.Setter;
@@ -19,7 +21,8 @@ import lombok.extern.slf4j.Slf4j;
 @Slf4j
 public class TablePageBlock extends AbstractPageBlock {
 
-    private final TreeMap<CellPosition, Cell> cells = new TreeMap<>();
+    public static final double CELL_AREA_CONTAINED_THRESHOLD = 0.98;
+    private final TreeMap<CellPosition, Cell> cellTreeMap = new TreeMap<>();
 
     private final int rotation;
     @Getter
@@ -28,20 +31,28 @@ public class TablePageBlock extends AbstractPageBlock {
     private int unrotatedRowCount;
     private int unrotatedColCount;
     private List<List<Cell>> rows;
+    @Getter
+    @Setter
+    private List<Cell> cells;
 
 
-    public TablePageBlock(List<Cell> cells, Rectangle area, int rotation) {
+    public TablePageBlock(List<Cell> cells, int rotation) {
 
+        setToBBoxOfComponents(cells);
+        this.cells = cells;
         addCells(cells);
-        minX = area.getLeft();
-        minY = area.getBottom();
-        maxX = area.getRight();
-        maxY = area.getTop();
         classification = PageBlockType.TABLE;
         this.rotation = rotation;
     }
 
 
+    @Override
+    public boolean isEmpty() {
+
+        return getColCount() == 0 || getRowCount() == 0;
+    }
+
+
     public List<List<Cell>> getRows() {
 
         if (rows == null) {
@@ -72,14 +83,17 @@ public class TablePageBlock extends AbstractPageBlock {
 
     public int getColCount() {
 
-        return getRows().stream().mapToInt(List::size).max().orElse(0);
+        return getRows().stream()
+                .mapToInt(List::size)
+                .max()
+                .orElse(0);
 
     }
 
 
     /**
      * Detect header cells (either first row or first column):
-     * Column is marked as header if cell text is bold and row cell text is not bold.
+     * Column is marked as header if originalCell text is bold and row originalCell text is not bold.
      * Defaults to row.
      */
     private void computeHeaders() {
@@ -87,7 +101,7 @@ public class TablePageBlock extends AbstractPageBlock {
         if (rows == null) {
             rows = computeRows();
         }
-        // A bold cell is a header cell as long as every cell to the left/top is bold, too
+        // A bold originalCell is a header originalCell as long as every originalCell to the left/top is bold, too
         // we move from left to right and top to bottom
         for (int rowIndex = 0; rowIndex < rows.size(); rowIndex++) {
             List<Cell> rowCells = rows.get(rowIndex);
@@ -112,7 +126,8 @@ public class TablePageBlock extends AbstractPageBlock {
                 List<Cell> cellsToTheTop = new ArrayList<>();
                 for (int i = 0; i < rowIndex; i++) {
                     try {
-                        cellsToTheTop.add(rows.get(i).get(colIndex));
+                        cellsToTheTop.add(rows.get(i)
+                                                  .get(colIndex));
                     } catch (IndexOutOfBoundsException e) {
                         log.debug("No cell {} in row {}, ignoring.", colIndex, rowIndex);
                     }
@@ -127,7 +142,8 @@ public class TablePageBlock extends AbstractPageBlock {
                 if (lastHeaderCell != null) {
                     cell.getHeaderCells().add(lastHeaderCell);
                 }
-                if (!cell.getTextBlocks().isEmpty() && cell.getTextBlocks().get(0).getMostPopularWordStyle().equals("bold")) {
+                if (!cell.getTextBlocks().isEmpty() && cell.getTextBlocks()
+                        .get(0).getMostPopularWordStyle().equals("bold")) {
                     cell.setHeaderCell(true);
                 }
             }
@@ -143,7 +159,7 @@ public class TablePageBlock extends AbstractPageBlock {
             for (int i = 0; i < unrotatedColCount; i++) { // rows
                 List<Cell> lastRow = new ArrayList<>();
                 for (int j = unrotatedRowCount - 1; j >= 0; j--) { // cols
-                    Cell cell = cells.get(new CellPosition(j, i));
+                    Cell cell = cellTreeMap.get(new CellPosition(j, i));
                     if (cell != null) {
                         lastRow.add(cell);
                     }
@@ -154,7 +170,7 @@ public class TablePageBlock extends AbstractPageBlock {
             for (int i = unrotatedColCount - 1; i >= 0; i--) { // rows
                 List<Cell> lastRow = new ArrayList<>();
                 for (int j = 0; j < unrotatedRowCount; j++) { // cols
-                    Cell cell = cells.get(new CellPosition(j, i));
+                    Cell cell = cellTreeMap.get(new CellPosition(j, i));
                     if (cell != null) {
                         lastRow.add(cell);
                     }
@@ -165,7 +181,7 @@ public class TablePageBlock extends AbstractPageBlock {
             for (int i = 0; i < unrotatedRowCount; i++) {
                 List<Cell> lastRow = new ArrayList<>();
                 for (int j = 0; j < unrotatedColCount; j++) {
-                    Cell cell = cells.get(new CellPosition(i, j)); // JAVA_8 use getOrDefault()
+                    Cell cell = cellTreeMap.get(new CellPosition(i, j)); // JAVA_8 use getOrDefault()
                     if (cell != null) {
                         lastRow.add(cell);
                     }
@@ -179,17 +195,6 @@ public class TablePageBlock extends AbstractPageBlock {
     }
 
 
-    private void add(Cell chunk, int row, int col) {
-
-        unrotatedRowCount = Math.max(unrotatedRowCount, row + 1);
-        unrotatedColCount = Math.max(unrotatedColCount, col + 1);
-
-        CellPosition cp = new CellPosition(row, col);
-        cells.put(cp, chunk);
-
-    }
-
-
     private void addCells(List<Cell> cells) {
 
         if (cells.isEmpty()) {
@@ -198,11 +203,12 @@ public class TablePageBlock extends AbstractPageBlock {
 
         cells.removeIf(cell -> cell.getWidth() < 1.1 || cell.getHeight() < 1.1);
 
-        List<List<Cell>> rowsOfCells = calculateStructure(cells);
+        List<List<Cell>> rowsOfCellsMatrix = calculateTableStructure(cells);
 
-        for (int i = 0; i < rowsOfCells.size(); i++) {
-            for (int j = 0; j < rowsOfCells.get(i).size(); j++) {
-                add(rowsOfCells.get(i).get(j), i, j);
+        for (int i = 0; i < rowsOfCellsMatrix.size(); i++) {
+            for (int j = 0; j < rowsOfCellsMatrix.get(i).size(); j++) {
+                addCellToRowAndCol(rowsOfCellsMatrix.get(i)
+                                           .get(j), i, j);
             }
         }
 
@@ -213,57 +219,125 @@ public class TablePageBlock extends AbstractPageBlock {
      * Calculates the structure of the table. For spanning rows and columns multiple cells with the same values will be inserted.
      *
      * @param cells The found cells
-     * @return TablePageBlock Structure
+     * @return TablePageBlock Structure as a rows of cells matrix
      */
-    private List<List<Cell>> calculateStructure(List<Cell> cells) {
-
-        List<List<Cell>> matrix = new ArrayList<>();
+    private List<List<Cell>> calculateTableStructure(List<Cell> cells) {
 
         if (cells.isEmpty()) {
-            return matrix;
+            return new ArrayList<>();
         }
 
-        Set<Float> uniqueX = new HashSet<>();
-        Set<Float> uniqueY = new HashSet<>();
-        cells.stream().filter(c -> !c.getTextBlocks().isEmpty() || c.getHeight() > 3 && c.getWidth() > 3).forEach(c -> {
-            uniqueX.add(c.getLeft());
-            uniqueX.add(c.getRight());
-            uniqueY.add(c.getBottom());
-            uniqueY.add(c.getTop());
-        });
+        Set<Double> uniqueX = new HashSet<>();
+        Set<Double> uniqueY = new HashSet<>();
+        cells.stream()
+                .filter(c -> !c.getTextBlocks().isEmpty() || c.getHeight() > 3 && c.getWidth() > 3)
+                .forEach(c -> {
+                    uniqueX.add(c.getPdfMinX());
+                    uniqueX.add(c.getPdfMaxX());
+                    uniqueY.add(c.getPdfMinY());
+                    uniqueY.add(c.getPdfMaxY());
+                });
 
-        var sortedUniqueX = uniqueX.stream().sorted().toList();
-        var sortedUniqueY = uniqueY.stream().sorted().toList();
+        var sortedUniqueX = uniqueX.stream()
+                .sorted()
+                .toList();
+        var sortedUniqueY = uniqueY.stream()
+                .sorted()
+                .toList();
 
-        Float prevY = null;
-        for (Float y : sortedUniqueY) {
+        List<List<Cell>> rowsOfCells = new ArrayList<>();
+
+        Double prevY = null;
+
+        for (Double y : sortedUniqueY) {
 
             List<Cell> row = new ArrayList<>();
 
-            Float prevX = null;
-            for (Float x : sortedUniqueX) {
+            Double prevX = null;
+            for (Double x : sortedUniqueX) {
 
                 if (prevY != null && prevX != null) {
-                    var cell = new Cell(new Point2D.Float(prevX, prevY), new Point2D.Float(x, y));
+                    var cellFromGridStructure = new Cell(new Point2D.Double(prevX, prevY), new Point2D.Double(x, y));
 
-                    var intersectionCell = cells.stream().filter(c -> cell.intersects(c) && cell.overlapRatio(c) > 0.1f).findFirst();
-                    intersectionCell.ifPresent(value -> cell.getTextBlocks().addAll(value.getTextBlocks()));
-                    if (cell.hasMinimumSize()) {
-                        row.add(cell);
+                    if (cellFromGridStructure.hasMinimumSize()) {
+
+                        cells.stream()
+                                .map(originalCell -> new CellWithIntersection(originalCell,
+                                                                              RectangleTransformations.calculateIntersectedArea(cellFromGridStructure.getBBoxPdf(),
+                                                                                                                                originalCell.getBBoxPdf())))
+                                .filter(cellWithIntersection -> cellWithIntersection.intersectedArea > 0)
+                                .filter(cellWithIntersection -> cellWithIntersection.originalCell.getArea() > cellWithIntersection.intersectedArea * CELL_AREA_CONTAINED_THRESHOLD)
+                                .max(Comparator.comparing(CellWithIntersection::intersectedArea))
+                                .map(CellWithIntersection::originalCell)
+                                .ifPresent(matchingCell -> cellFromGridStructure.getTextBlocks().addAll(matchingCell.getTextBlocks()));
+
+                        row.add(cellFromGridStructure);
                     }
                 }
                 prevX = x;
             }
 
-            if (prevY != null && prevX != null && !row.isEmpty()) {
-                matrix.add(row);
+            // exclude empty rows and rows where all text blocks are empty
+            if (prevY != null && prevX != null && !row.isEmpty() && !row.stream()
+                    .allMatch(cell -> cell.getTextBlocks().isEmpty())) {
+
+                rowsOfCells.add(row);
             }
             prevY = y;
         }
 
-        Collections.reverse(matrix);
+        Collections.reverse(rowsOfCells);
+
+        // now cells are removed which are part of a column without any text blocks
+        // this is done by first computing the inverse matrix which contains call columns of cells
+        // then the column indices that have to be removed are determined
+        List<List<Cell>> columnsOfCells = new ArrayList<>();
+        int maxRowLength = rowsOfCells.stream()
+                .map(List::size)
+                .max(java.util.Comparator.naturalOrder())
+                .orElse(0);
+        for (int i = 0; i < maxRowLength; i++) {
+            columnsOfCells.add(new ArrayList<>());
+        }
+
+        for (List<Cell> row : rowsOfCells) {
+            for (int j = 0; j < row.size(); j++) {
+                columnsOfCells.get(j).add(row.get(j));
+            }
+        }
+
+        List<Integer> columnIndicesToRemove = new ArrayList<>();
+        int columnIndex = 0;
+        for (List<Cell> col : columnsOfCells) {
+            if (col.stream()
+                    .allMatch(cell -> cell.getTextBlocks().isEmpty())) {
+                columnIndicesToRemove.add(columnIndex);
+            }
+            columnIndex++;
+        }
+        columnIndicesToRemove.sort(Collections.reverseOrder());
+
+        // update all rows so that the values of the empty columns get removed
+        var rowsOfCellsBefore = new ArrayList<>(rowsOfCells);
+        rowsOfCells = new ArrayList<>();
+        for (List<Cell> row : rowsOfCellsBefore) {
+            var updatedRow = new ArrayList<>(row);
+            columnIndicesToRemove.forEach(idxToRemove -> updatedRow.remove(updatedRow.get(idxToRemove)));
+            rowsOfCells.add(updatedRow);
+        }
+
+        return rowsOfCells;
+    }
+
+
+    private void addCellToRowAndCol(Cell cell, int row, int col) {
+
+        unrotatedRowCount = Math.max(unrotatedRowCount, row + 1);
+        unrotatedColCount = Math.max(unrotatedColCount, col + 1);
+
+        CellPosition cp = new CellPosition(row, col);
+        cellTreeMap.put(cp, cell);
 
-        return matrix;
     }
 
 
@@ -290,7 +364,7 @@ public class TablePageBlock extends AbstractPageBlock {
                             if (!first) {
                                 sb.append("\n");
                             }
-                            sb.append('\"').append(textBlock.getText().replaceAll("\"", "\\\"")).append('\"');
+                            sb.append('\"').append(textBlock.getText().replaceAll("\"", "\"")).append('\"');
                             first = false;
                         }
                     }
@@ -337,4 +411,9 @@ public class TablePageBlock extends AbstractPageBlock {
         return sb.toString();
     }
 
+
+    record CellWithIntersection(Cell originalCell, double intersectedArea) {
+
+    }
+
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/text/FontStyle.java

@@ -0,0 +1,9 @@
+package com.knecon.fforesight.service.layoutparser.processor.model.text;
+
+
+public enum FontStyle {
+    REGULAR,
+    BOLD,
+    ITALIC,
+    BOLD_ITALIC;
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/text/RedTextPosition.java

@@ -1,100 +1,92 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.text;
 
+import java.awt.geom.AffineTransform;
+import java.awt.geom.Rectangle2D;
+
 import org.apache.pdfbox.text.TextPosition;
-import org.springframework.beans.BeanUtils;
 
-import com.fasterxml.jackson.annotation.JsonIgnore;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.TextBoundingBox;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.utils.FastAtan2;
 
+import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
-import lombok.Builder;
 import lombok.Data;
+import lombok.EqualsAndHashCode;
 import lombok.NoArgsConstructor;
 import lombok.SneakyThrows;
+import lombok.experimental.FieldDefaults;
+import lombok.experimental.SuperBuilder;
 
 @Data
-@Builder
+@SuperBuilder
 @NoArgsConstructor
 @AllArgsConstructor
-public class RedTextPosition {
+@EqualsAndHashCode(callSuper = true)
+@FieldDefaults(level = AccessLevel.PRIVATE)
+public class RedTextPosition extends TextBoundingBox {
 
-    private String textMatrix;
-    private float[] position;
+    public final static int HEIGHT_PADDING = 2;
 
-    @JsonIgnore
-    private int rotation;
+    String unicode;
 
-    @JsonIgnore
-    private float pageHeight;
+    // estimated using the TextMatrix in radians
+    float exactDir;
 
-    @JsonIgnore
-    private float pageWidth;
+    float widthOfSpace;
 
-    private String unicode;
+    float fontSizeInPt;
 
-    @JsonIgnore
-    private float dir;
-
-    // not used in reanalysis
-    @JsonIgnore
-    private float widthOfSpace;
-
-    // not used in reanalysis
-    @JsonIgnore
-    private float fontSizeInPt;
-
-    // not used in reanalysis
-    @JsonIgnore
-    private String fontName;
+    String fontName;
 
 
     @SneakyThrows
     public static RedTextPosition fromTextPosition(TextPosition textPosition) {
 
         var pos = new RedTextPosition();
-        BeanUtils.copyProperties(textPosition, pos);
-        pos.setFontName(textPosition.getFont().getName());
-
+        pos.setUnicode(textPosition.getUnicode());
+        pos.setWidthOfSpace(textPosition.getWidthOfSpace());
         pos.setFontSizeInPt(textPosition.getFontSizeInPt());
+        pos.setFontName(textPosition.getFont().getName());
+        pos.setExactDir((float) FastAtan2.fastAtan2(textPosition.getTextMatrix().getShearY(), textPosition.getTextMatrix().getScaleX()));
+        pos.setDir(TextDirection.fromDegrees(textPosition.getDir()));
 
-        pos.setTextMatrix(textPosition.getTextMatrix().toString());
+        //TODO: There is a mismatch in the java coords of the text and the rulings,
+        // I guess if we start with the initial user space positions and transform them the same way we do the rulings it would work.
+        pos.setBBox(new Rectangle2D.Float(textPosition.getX(), textPosition.getY(), textPosition.getWidthDirAdj(), textPosition.getHeight()));
 
-        var position = new float[4];
+        float textHeight = textPosition.getHeight() + 2 * HEIGHT_PADDING;
+        Rectangle2D.Float dirAdjPosition = new Rectangle2D.Float(textPosition.getXDirAdj(),
+                                                                 textPosition.getYDirAdj() - textHeight,
+                                                                 textPosition.getWidthDirAdj(),
+                                                                 textHeight + HEIGHT_PADDING);
 
-        position[0] = textPosition.getXDirAdj();
-        position[1] = textPosition.getYDirAdj();
-        position[2] = textPosition.getWidthDirAdj();
-        position[3] = textPosition.getHeightDir();
+        pos.setBBoxDirAdj(dirAdjPosition);
+
+        AffineTransform affineTransform = getRotationMatrix(TextDirection.fromDegrees(textPosition.getDir()), textPosition.getPageWidth(), textPosition.getPageHeight());
+        Rectangle2D bBoxInitialUserSpace = affineTransform.createTransformedShape(dirAdjPosition).getBounds2D();
+
+        pos.setBBoxPdf(bBoxInitialUserSpace); // These are definitely correct
 
-        pos.setPosition(position);
         return pos;
     }
 
 
-    @JsonIgnore
-    public float getXDirAdj() {
+    private static AffineTransform getRotationMatrix(TextDirection textDirection, float pageWidth, float pageHeight) {
 
-        return position[0];
-    }
+        AffineTransform transform = new AffineTransform();
 
-
-    @JsonIgnore
-    public float getYDirAdj() {
-
-        return position[1];
-    }
-
-
-    @JsonIgnore
-    public float getWidthDirAdj() {
-
-        return position[2];
-    }
-
-
-    @JsonIgnore
-    public float getHeightDir() {
-
-        return position[3];
+        if (textDirection == TextDirection.ZERO || textDirection == TextDirection.HALF_CIRCLE) {
+            transform.rotate(textDirection.getRadians(), pageWidth / 2f, pageHeight / 2f);
+            transform.translate(0f, pageHeight);
+        } else if (textDirection == TextDirection.QUARTER_CIRCLE) {
+            transform.rotate(textDirection.getRadians(), pageWidth / 2f, pageWidth / 2f);
+            transform.translate(0f, pageWidth);
+        } else {
+            transform.rotate(textDirection.getRadians(), pageHeight / 2f, pageHeight / 2f);
+            transform.translate(0f, pageWidth);
+        }
+        transform.scale(1., -1.);
+        return transform;
     }
 
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/text/SearchableText.java

@@ -2,6 +2,7 @@ package com.knecon.fforesight.service.layoutparser.processor.model.text;
 
 import java.util.ArrayList;
 import java.util.List;
+
 import com.knecon.fforesight.service.layoutparser.processor.utils.TextNormalizationUtilities;
 
 import lombok.Getter;
@@ -38,11 +39,7 @@ public class SearchableText {
             sb.append(word);
             sb.append(' ');
         }
-        String text = sb.toString();
-        text = TextNormalizationUtilities.removeHyphenLineBreaks(text);
-        text = TextNormalizationUtilities.removeLineBreaks(text);
-        text = TextNormalizationUtilities.removeRepeatingWhitespaces(text);
-        return text;
+        return TextNormalizationUtilities.cleanString(sb.toString());
     }
 
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/text/TextDirection.java

@@ -44,4 +44,15 @@ public enum TextDirection {
 
         throw new IllegalArgumentException(String.format("A value of %f is not supported by TextDirection", degrees));
     }
+
+
+    public int getRotation() {
+
+        return switch (this) {
+            case ZERO -> 0;
+            case QUARTER_CIRCLE -> 1;
+            case HALF_CIRCLE -> 2;
+            case THREE_QUARTER_CIRCLE -> 3;
+        };
+    }
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/text/TextPageBlock.java

@@ -1,16 +1,15 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.text;
 
-import static java.util.stream.Collectors.toSet;
-
+import java.awt.geom.Rectangle2D;
 import java.util.ArrayList;
-import java.util.Comparator;
+import java.util.Collections;
 import java.util.List;
 
 import com.fasterxml.jackson.annotation.JsonIgnore;
 import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.FloatFrequencyCounter;
 import com.knecon.fforesight.service.layoutparser.processor.model.PageBlockType;
-import com.knecon.fforesight.service.layoutparser.processor.utils.DoubleComparisons;
+import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
 import com.knecon.fforesight.service.layoutparser.processor.utils.TextNormalizationUtilities;
 
 import lombok.AllArgsConstructor;
@@ -29,61 +28,97 @@ public class TextPageBlock extends AbstractPageBlock {
     @Builder.Default
     private List<TextPositionSequence> sequences = new ArrayList<>();
 
-    @JsonIgnore
-    private int rotation;
+    private Rectangle2D bBoxDirAdj;
 
-    @JsonIgnore
     private String mostPopularWordFont;
 
-    @JsonIgnore
     private String mostPopularWordStyle;
 
-    @JsonIgnore
-    private float mostPopularWordFontSize;
+    private double mostPopularWordFontSize;
 
-    @JsonIgnore
-    private float mostPopularWordHeight;
+    private double mostPopularWordHeight;
 
-    @JsonIgnore
-    private float mostPopularWordSpaceWidth;
+    private double mostPopularWordSpaceWidth;
 
-    @JsonIgnore
-    private float highestFontSize;
+    private boolean underlined;
+
+    private double highestFontSize;
 
-    @JsonIgnore
     private PageBlockType classification;
 
+    private boolean toDuplicate;
+
+    private String text;
+    private boolean changed;
+
+
+    public TextPageBlock(List<TextPositionSequence> sequences) {
+
+        this.sequences = new ArrayList<>(sequences);
+        if (!sequences.isEmpty()) {
+            calculateFrequencyCounters();
+        }
+        calculateBBox();
+    }
+
+
+    public List<TextPositionSequence> getSequences() {
+
+        return Collections.unmodifiableList(sequences);
+    }
+
 
-    @JsonIgnore
     public TextDirection getDir() {
 
         return sequences.get(0).getDir();
     }
 
 
-    @JsonIgnore
-    private float getPageHeight() {
+    private void calculateBBox() {
 
-        return sequences.get(0).getPageHeight();
+        if (sequences == null) {
+            this.bBox = new Rectangle2D.Double();
+            this.bBoxPdf = new Rectangle2D.Double();
+            this.bBoxDirAdj = new Rectangle2D.Double();
+            return;
+        }
+        this.bBoxDirAdj = sequences.stream()
+                .map(TextPositionSequence::getBBoxDirAdj)
+                .collect(RectangleTransformations.collectBBox());
+
+        setToBBoxOfComponents(sequences);
     }
 
 
-    @JsonIgnore
-    private float getPageWidth() {
+    public void recalculateBBox() {
 
-        return sequences.get(0).getPageWidth();
+        calculateBBox();
     }
 
+
     public static TextPageBlock merge(List<TextPageBlock> textBlocksToMerge) {
 
-        List<TextPositionSequence> sequences = textBlocksToMerge.stream().map(TextPageBlock::getSequences).flatMap(java.util.Collection::stream).toList();
+        if (textBlocksToMerge.isEmpty()) {
+            throw new IllegalArgumentException("Need to provide at least one TextPageBlock.");
+        }
+        if (textBlocksToMerge.stream()
+                    .map(AbstractPageBlock::getPage)
+                    .distinct()
+                    .count() != 1) {
+            throw new IllegalArgumentException("Cannot merge textBlocks on different pages.");
+        }
+
+        List<TextPositionSequence> sequences = textBlocksToMerge.stream()
+                .map(TextPageBlock::getSequences)
+                .flatMap(java.util.Collection::stream)
+                .toList();
         sequences = new ArrayList<>(sequences);
-        return fromTextPositionSequences(sequences);
+
+        return new TextPageBlock(sequences);
     }
 
-    public static TextPageBlock fromTextPositionSequences(List<TextPositionSequence> wordBlockList) {
 
-        TextPageBlock textBlock = null;
+    private void calculateFrequencyCounters() {
 
         FloatFrequencyCounter lineHeightFrequencyCounter = new FloatFrequencyCounter();
         FloatFrequencyCounter fontSizeFrequencyCounter = new FloatFrequencyCounter();
@@ -91,7 +126,7 @@ public class TextPageBlock extends AbstractPageBlock {
         StringFrequencyCounter fontFrequencyCounter = new StringFrequencyCounter();
         StringFrequencyCounter styleFrequencyCounter = new StringFrequencyCounter();
 
-        for (TextPositionSequence wordBlock : wordBlockList) {
+        for (TextPositionSequence wordBlock : sequences) {
 
             lineHeightFrequencyCounter.add(wordBlock.getTextHeight());
             fontSizeFrequencyCounter.add(wordBlock.getFontSize());
@@ -99,154 +134,17 @@ public class TextPageBlock extends AbstractPageBlock {
             fontFrequencyCounter.add(wordBlock.getFont());
             styleFrequencyCounter.add(wordBlock.getFontStyle());
 
-            if (textBlock == null) {
-                textBlock = new TextPageBlock(wordBlock.getMinXDirAdj(),
-                        wordBlock.getMaxXDirAdj(),
-                        wordBlock.getMinYDirAdj(),
-                        wordBlock.getMaxYDirAdj(),
-                        wordBlockList,
-                        wordBlock.getRotation());
-            } else {
-                TextPageBlock spatialEntity = textBlock.union(wordBlock);
-                textBlock.resize(spatialEntity.getMinX(), spatialEntity.getMinY(), spatialEntity.getWidth(), spatialEntity.getHeight());
-            }
         }
 
-        if (textBlock != null) {
-            textBlock.setMostPopularWordFont(fontFrequencyCounter.getMostPopular());
-            textBlock.setMostPopularWordStyle(styleFrequencyCounter.getMostPopular());
-            textBlock.setMostPopularWordFontSize(fontSizeFrequencyCounter.getMostPopular());
-            textBlock.setMostPopularWordHeight(lineHeightFrequencyCounter.getMostPopular());
-            textBlock.setMostPopularWordSpaceWidth(spaceFrequencyCounter.getMostPopular());
-            textBlock.setHighestFontSize(fontSizeFrequencyCounter.getHighest());
-        }
+        setMostPopularWordFont(fontFrequencyCounter.getMostPopular());
+        setMostPopularWordStyle(styleFrequencyCounter.getMostPopular());
+        setMostPopularWordFontSize(fontSizeFrequencyCounter.getMostPopular());
+        setMostPopularWordHeight(lineHeightFrequencyCounter.getMostPopular());
+        setMostPopularWordSpaceWidth(spaceFrequencyCounter.getMostPopular());
+        setHighestFontSize(fontSizeFrequencyCounter.getHighest());
 
-        if (textBlock != null && textBlock.getSequences() != null && textBlock.getSequences()
-                .stream()
-                .map(t -> DoubleComparisons.round(t.getMinYDirAdj(), 3))
-                .collect(toSet())
-                .size() == 1) {
-            textBlock.getSequences().sort(Comparator.comparing(TextPositionSequence::getMinXDirAdj));
-        }
-        return textBlock;
-    }
-
-
-
-    /**
-     * Returns the minX value in pdf coordinate system.
-     * Note: This needs to use Pdf Coordinate System where {0,0} rotated with the page rotation.
-     * 0 -> LowerLeft
-     * 90 -> UpperLeft
-     * 180 -> UpperRight
-     * 270 -> LowerRight
-     *
-     * @return the minX value in pdf coordinate system
-     */
-    @JsonIgnore
-    public float getPdfMinX() {
-
-        if (getDir().getDegrees() == 90) {
-            return minY;
-        } else if (getDir().getDegrees() == 180) {
-            return getPageWidth() - maxX;
-
-        } else if (getDir().getDegrees() == 270) {
-
-            return getPageWidth() - maxY;
-        } else {
-            return minX;
-        }
-    }
-
-
-    /**
-     * Returns the maxX value in pdf coordinate system.
-     * Note: This needs to use Pdf Coordinate System where {0,0} rotated with the page rotation.
-     * 0 -> LowerLeft
-     * 90 -> UpperLeft
-     * 180 -> UpperRight
-     * 270 -> LowerRight
-     *
-     * @return the maxX value in pdf coordinate system
-     */
-    @JsonIgnore
-    public float getPdfMaxX() {
-
-        if (getDir().getDegrees() == 90) {
-            return maxY;
-        } else if (getDir().getDegrees() == 180) {
-            return getPageWidth() - minX;
-        } else if (getDir().getDegrees() == 270) {
-            return getPageWidth() - minY;
-
-        } else {
-            return maxX;
-        }
-    }
-
-
-    /**
-     * Returns the minY value in pdf coordinate system.
-     * Note: This needs to use Pdf Coordinate System where {0,0} rotated with the page rotation.
-     * 0 -> LowerLeft
-     * 90 -> UpperLeft
-     * 180 -> UpperRight
-     * 270 -> LowerRight
-     *
-     * @return the minY value in pdf coordinate system
-     */
-    @JsonIgnore
-    public float getPdfMinY() {
-
-        if (getDir().getDegrees() == 90) {
-            return minX;
-        } else if (getDir().getDegrees() == 180) {
-            return maxY;
-
-        } else if (getDir().getDegrees() == 270) {
-            return getPageHeight() - maxX;
-
-        } else {
-            return getPageHeight() - maxY;
-        }
-    }
-
-
-    /**
-     * Returns the maxY value in pdf coordinate system.
-     * Note: This needs to use Pdf Coordinate System where {0,0} rotated with the page rotation.
-     * 0 -> LowerLeft
-     * 90 -> UpperLeft
-     * 180 -> UpperRight
-     * 270 -> LowerRight
-     *
-     * @return the maxY value in pdf coordinate system
-     */
-    @JsonIgnore
-    public float getPdfMaxY() {
-
-        if (getDir().getDegrees() == 90) {
-            return maxX;
-        } else if (getDir().getDegrees() == 180) {
-
-            return minY;
-        } else if (getDir().getDegrees() == 270) {
-            return getPageHeight() - minX;
-        } else {
-            return getPageHeight() - minY;
-        }
-    }
-
-
-    public TextPageBlock(float minX, float maxX, float minY, float maxY, List<TextPositionSequence> sequences, int rotation) {
-
-        this.minX = minX;
-        this.maxX = maxX;
-        this.minY = minY;
-        this.maxY = maxY;
-        this.sequences = sequences;
-        this.rotation = rotation;
+        setUnderlined(sequences.stream()
+                              .allMatch(TextPositionSequence::isUnderline));
     }
 
 
@@ -254,6 +152,8 @@ public class TextPageBlock extends AbstractPageBlock {
 
         TextPageBlock union = this.copy();
         union.add(r);
+        calculateFrequencyCounters();
+        calculateBBox();
         return union;
     }
 
@@ -261,83 +161,50 @@ public class TextPageBlock extends AbstractPageBlock {
     public TextPageBlock union(TextPageBlock r) {
 
         TextPageBlock union = this.copy();
-        union.add(r);
+        union.addAll(r.getSequences());
+        calculateFrequencyCounters();
+        calculateBBox();
         return union;
     }
 
 
-    public void add(TextPageBlock r) {
+    public void add(TextPageBlock textPageBlock) {
 
-        if (r.getMinX() < minX) {
-            minX = r.getMinX();
-        }
-        if (r.getMaxX() > maxX) {
-            maxX = r.getMaxX();
-        }
-        if (r.getMinY() < minY) {
-            minY = r.getMinY();
-        }
-        if (r.getMaxY() > maxY) {
-            maxY = r.getMaxY();
-        }
-        sequences.addAll(r.getSequences());
+        changed = true;
+        sequences.addAll(textPageBlock.getSequences());
+        calculateFrequencyCounters();
+        calculateBBox();
     }
 
 
-    public void add(TextPositionSequence r) {
+    public void add(TextPositionSequence textPositionSequence) {
 
-        if (r.getMinXDirAdj() < minX) {
-            minX = r.getMinXDirAdj();
-        }
-        if (r.getMaxXDirAdj() > maxX) {
-            maxX = r.getMaxXDirAdj();
-        }
-        if (r.getMinYDirAdj() < minY) {
-            minY = r.getMinYDirAdj();
-        }
-        if (r.getMaxYDirAdj() > maxY) {
-            maxY = r.getMaxYDirAdj();
-        }
+        changed = true;
+        sequences.add(textPositionSequence);
+        calculateFrequencyCounters();
+        calculateBBox();
+    }
+
+
+    public void addAll(List<TextPositionSequence> textPositionSequences) {
+
+        changed = true;
+        sequences.addAll(textPositionSequences);
+        calculateFrequencyCounters();
+        calculateBBox();
     }
 
 
     public TextPageBlock copy() {
 
-        return new TextPageBlock(minX, maxX, minY, maxY, sequences, rotation);
-    }
-
-
-    public void resize(float x1, float y1, float width, float height) {
-
-        set(x1, y1, x1 + width, y1 + height);
-    }
-
-
-    public void set(float x1, float y1, float x2, float y2) {
-
-        this.minX = Math.min(x1, x2);
-        this.maxX = Math.max(x1, x2);
-        this.minY = Math.min(y1, y2);
-        this.maxY = Math.max(y1, y2);
+        return new TextPageBlock(new ArrayList<>(sequences));
     }
 
 
     @Override
     public String toString() {
 
-        StringBuilder builder = new StringBuilder();
-
-        for (int i = 0; i < sequences.size(); i++) {
-            String sequenceAsString = sequences.get(i).toString();
-            // Fix for missing Whitespace. This is recognized in getSequences method. See PDFTextStripper Line 1730.
-            if (i != 0 && sequences.get(i - 1).charAt(sequences.get(i - 1).length() - 1) != ' ' && sequenceAsString.charAt(0) != ' ') {
-                builder.append(' ');
-            }
-            builder.append(sequenceAsString);
-        }
-
-        return builder.toString();
-
+        return getText();
     }
 
 
@@ -345,23 +212,51 @@ public class TextPageBlock extends AbstractPageBlock {
     @JsonIgnore
     public String getText() {
 
-        StringBuilder sb = new StringBuilder();
+        if (text == null || changed) {
 
+            StringBuilder sb = new StringBuilder();
+
+            TextPositionSequence previous = null;
+            for (TextPositionSequence word : sequences) {
+                if (previous != null) {
+                    if (Math.abs(previous.getMaxYDirAdj() - word.getMaxYDirAdj()) > word.getTextHeight()) {
+                        sb.append('\n');
+                    } else {
+                        sb.append(' ');
+                    }
+                }
+                sb.append(word.toString());
+                previous = word;
+            }
+
+            text = TextNormalizationUtilities.removeHyphenLinebreaks(sb.toString());
+            changed = false;
+        }
+
+        return text;
+    }
+
+
+    public int getNumberOfLines() {
+
+        int numberOfLines = 1;
         TextPositionSequence previous = null;
         for (TextPositionSequence word : sequences) {
             if (previous != null) {
-                if (Math.abs(previous.getMaxYDirAdj() - word.getMaxYDirAdj()) > word.getTextHeight()) {
-                    sb.append('\n');
-                } else {
-                    sb.append(' ');
+                if (word.getMaxYDirAdj() - previous.getMaxYDirAdj() > word.getTextHeight()) {
+                    numberOfLines++;
                 }
             }
-            sb.append(word.toString());
             previous = word;
         }
+        return numberOfLines;
+    }
 
-        return TextNormalizationUtilities.removeHyphenLineBreaks(sb.toString());
 
+    @Override
+    public boolean isEmpty() {
+
+        return sequences.isEmpty();
     }
 
 }