Merge branch 'RED-8932' into 'main'

RED-8932 Fixed not merged headline with identifier

See merge request fforesight/layout-parser!141

.gitignore

@@ -42,3 +42,5 @@ gradlew.bat
 gradlew
 gradle.properties
 gradle/
+.DS_Store
+.DS_Store/

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 +1,6 @@
 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"

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

@@ -1,5 +1,7 @@
 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;
@@ -12,7 +14,7 @@ import lombok.experimental.FieldDefaults;
 @AllArgsConstructor
 @FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
 @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 {
+public class DocumentData implements Serializable {
 
     @Schema(description = "Contains information about the document's pages.")
     DocumentPage[] documentPages;

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

@@ -1,5 +1,7 @@
 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;
@@ -14,7 +16,7 @@ import lombok.experimental.FieldDefaults;
 @AllArgsConstructor
 @FieldDefaults(level = AccessLevel.PRIVATE)
 @Schema(description = "Object containing information about the document's pages.")
-public class DocumentPage {
+public class DocumentPage implements Serializable {
 
     @Schema(description = "The page number, starting with 1.")
     int number;

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

@@ -1,5 +1,7 @@
 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;
@@ -14,7 +16,7 @@ import lombok.experimental.FieldDefaults;
 @AllArgsConstructor
 @FieldDefaults(level = AccessLevel.PRIVATE)
 @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 {
+public class DocumentPositionData implements Serializable {
 
     @Schema(description = "Identifier of the text block.")
     Long id;

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

@@ -1,9 +1,11 @@
 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;
@@ -20,14 +22,13 @@ import lombok.experimental.FieldDefaults;
 @AllArgsConstructor
 @FieldDefaults(level = AccessLevel.PRIVATE)
 @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 {
+public class DocumentStructure implements Serializable {
 
     @Schema(description = "The root EntryData represents the Document.")
     EntryData root;
 
-
     @Schema(description = "Object containing the extra field names, a table has in its properties field.")
-    public static class TableProperties {
+    public static class TableProperties implements Serializable {
 
         public static final String NUMBER_OF_ROWS = "numberOfRows";
         public static final String NUMBER_OF_COLS = "numberOfCols";
@@ -35,7 +36,7 @@ public class DocumentStructure {
     }
 
     @Schema(description = "Object containing the extra field names, an Image has in its properties field.")
-    public static class ImageProperties {
+    public static class ImageProperties implements Serializable {
 
         public static final String TRANSPARENT = "transparent";
         public static final String IMAGE_TYPE = "imageType";
@@ -45,7 +46,7 @@ public class DocumentStructure {
     }
 
     @Schema(description = "Object containing the extra field names, a table cell has in its properties field.")
-    public static class TableCellProperties {
+    public static class TableCellProperties implements Serializable {
 
         public static final String B_BOX = "bBox";
         public static final String ROW = "row";
@@ -54,8 +55,16 @@ 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();
@@ -100,7 +109,7 @@ public class DocumentStructure {
     @AllArgsConstructor
     @FieldDefaults(level = AccessLevel.PRIVATE)
     @Schema(description = "Object containing information of a SemanticNode and also structuring the layout with children.")
-    public static class EntryData {
+    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;
@@ -114,6 +123,8 @@ public class DocumentStructure {
         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

@@ -1,5 +1,7 @@
 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;
@@ -14,7 +16,7 @@ import lombok.experimental.FieldDefaults;
 @AllArgsConstructor
 @FieldDefaults(level = AccessLevel.PRIVATE)
 @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 {
+public class DocumentTextData implements Serializable {
 
     @Schema(description = "Identifier of the text block.")
     Long id;

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

@@ -0,0 +1,6 @@
+package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
+
+public enum LayoutEngine {
+    ALGORITHM,
+    AI
+}

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

@@ -1,8 +1,9 @@
 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,
     HEADLINE,

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

@@ -3,6 +3,5 @@ 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";
 }

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

@@ -18,11 +18,15 @@ public record LayoutParsingRequest(
 
         @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.")//

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

@@ -2,6 +2,9 @@ package com.knecon.fforesight.service.layoutparser.internal.api.queue;
 
 public enum LayoutParsingType {
     REDACT_MANAGER,
-    TAAS,
-    DOCUMINE
+    REDACT_MANAGER_OLD,
+    REDACT_MANAGER_PARAGRAPH_DEBUG,
+    DOCUMINE,
+    CLARIFYND,
+    CLARIFYND_PARAGRAPH_DEBUG
 }

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

@@ -1,6 +1,6 @@
 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"
@@ -10,12 +10,13 @@ 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-v1:2.144.0") {
         exclude("org.springframework.boot", "spring-boot-starter-security")
         exclude("org.springframework.boot", "spring-boot-starter-validation")
     }
-    implementation("com.knecon.fforesight:tenant-commons:0.19.0")
+    implementation("com.knecon.fforesight:tenant-commons:0.21.0")
     implementation("com.iqser.red.commons:storage-commons:2.45.0")
 
     implementation("org.apache.pdfbox:pdfbox:${pdfBoxVersion}")
@@ -23,4 +24,6 @@ dependencies {
     implementation("com.fasterxml.jackson.module:jackson-module-afterburner:${jacksonVersion}")
     implementation("com.fasterxml.jackson.datatype:jackson-datatype-jsr310:${jacksonVersion}")
     implementation("org.springframework.boot:spring-boot-starter-web:3.1.3")
+    implementation("org.jgrapht:jgrapht-core:1.5.2")
+    implementation("org.tinspin:tinspin-indexes:2.1.3")
 }

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

@@ -27,31 +27,36 @@ import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBl
 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.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.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.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.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 io.micrometer.observation.Observation;
@@ -74,18 +79,20 @@ public class LayoutParsingPipeline {
     CvTableParsingAdapter cvTableParsingAdapter;
     LayoutParsingStorageService layoutParsingStorageService;
     SectionsBuilderService sectionsBuilderService;
-    TaasClassificationService taasClassificationService;
     RedactManagerClassificationService redactManagerClassificationService;
     DocuMineClassificationService docuMineClassificationService;
     SimplifiedSectionTextService simplifiedSectionTextService;
     BodyTextFrameService bodyTextFrameService;
     RulingCleaningService rulingCleaningService;
     TableExtractionService tableExtractionService;
-    TaasBlockificationService taasBlockificationService;
     DocuMineBlockificationService docuMineBlockificationService;
     RedactManagerBlockificationService redactManagerBlockificationService;
-    ViewerDocumentService viewerDocumentService;
+    DocstrumBlockificationService docstrumBlockificationService;
+    LayoutGridService layoutGridService;
     ObservationRegistry observationRegistry;
+    VisualLayoutParsingAdapter visualLayoutParsingAdapter;
+    ClarifyndClassificationService clarifyndClassificationService;
+    GraphicExtractorService graphicExtractorService;
 
 
     public LayoutParsingFinishedEvent parseLayoutAndSaveFilesToStorage(LayoutParsingRequest layoutParsingRequest) throws IOException {
@@ -94,7 +101,12 @@ public class LayoutParsingPipeline {
         log.info("Starting layout parsing for {}", layoutParsingRequest.identifier());
 
         File originFile = layoutParsingStorageService.getOriginFile(layoutParsingRequest.originFileStorageId());
-        File viewerDocumentFile = File.createTempFile("viewer_document", ".pdf");
+        File viewerDocumentFile = layoutParsingStorageService.getViewerDocFile(layoutParsingRequest.viewerDocumentStorageId()).orElse(originFile);
+
+        VisualLayoutParsingResponse visualLayoutParsingResponse = new VisualLayoutParsingResponse();
+        if (layoutParsingRequest.visualLayoutParsingFileId().isPresent()) {
+            visualLayoutParsingResponse = layoutParsingStorageService.getVisualLayoutParsingFile(layoutParsingRequest.visualLayoutParsingFileId().get());
+        }
 
         ImageServiceResponse imageServiceResponse = new ImageServiceResponse();
         if (layoutParsingRequest.imagesFileStorageId().isPresent()) {
@@ -110,26 +122,33 @@ public class LayoutParsingPipeline {
                 originFile,
                 imageServiceResponse,
                 tableServiceResponse,
-                layoutParsingRequest.identifier().toString());
+                visualLayoutParsingResponse,
+                layoutParsingRequest.identifier());
+
         log.info("Building document graph for {}", layoutParsingRequest.identifier());
 
-        Document documentGraph = observeBuildDocumentGraph(classificationDocument);
-
-        log.info("Storing resulting files for {}", layoutParsingRequest.identifier());
-        layoutParsingStorageService.storeDocumentData(layoutParsingRequest, DocumentDataMapper.toDocumentData(documentGraph));
-        layoutParsingStorageService.storeSimplifiedText(layoutParsingRequest, simplifiedSectionTextService.toSimplifiedText(documentGraph));
+        Document documentGraph = observeBuildDocumentGraph(layoutParsingRequest.layoutParsingType(), classificationDocument);
 
         log.info("Creating viewer document for {}", layoutParsingRequest.identifier());
-        viewerDocumentService.createViewerDocument(originFile, documentGraph, viewerDocumentFile, false);
+
+        layoutGridService.addLayoutGrid(viewerDocumentFile, documentGraph, viewerDocumentFile, false);
+
+        log.info("Storing resulting files for {}", layoutParsingRequest.identifier());
+
+        layoutParsingStorageService.storeDocumentData(layoutParsingRequest, DocumentDataMapper.toDocumentData(documentGraph));
+        layoutParsingStorageService.storeSimplifiedText(layoutParsingRequest, simplifiedSectionTextService.toSimplifiedText(documentGraph));
         layoutParsingStorageService.storeViewerDocument(layoutParsingRequest, viewerDocumentFile);
-        if (layoutParsingRequest.layoutParsingType().equals(LayoutParsingType.TAAS)) {
+
+        if (layoutParsingRequest.layoutParsingType().equals(LayoutParsingType.CLARIFYND)) {
             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();
-        viewerDocumentFile.delete();
 
         return LayoutParsingFinishedEvent.builder()
                 .identifier(layoutParsingRequest.identifier())
@@ -160,13 +179,13 @@ public class LayoutParsingPipeline {
     }
 
 
-    private Document observeBuildDocumentGraph(ClassificationDocument classificationDocument) {
+    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(classificationDocument));
-        });
+        Observation.createNotStarted("LayoutParsingPipeline", observationRegistry)
+                .contextualName("build-document-graph")
+                .observe(() -> documentReference.set(DocumentGraphFactory.buildDocumentGraph(layoutParsingType, classificationDocument)));
 
         return documentReference.get();
     }
@@ -192,13 +211,14 @@ public class LayoutParsingPipeline {
                                               File originFile,
                                               ImageServiceResponse imageServiceResponse,
                                               TableServiceResponse tableServiceResponse,
-                                              String identifier) {
+                                              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<>();
 
@@ -237,11 +257,28 @@ public class LayoutParsingPipeline {
             PDRectangle cropbox = pdPage.getCropBox();
             CleanRulings cleanRulings = rulingCleaningService.getCleanRulings(pdfTableCells.get(pageNumber), stripper.getRulings());
 
+            List<Cell> emptyTableCells = TableExtractionService.findCells(cleanRulings.getHorizontal(), cleanRulings.getVertical());
+
+            var graphics = graphicExtractorService.extractPathElementGraphics(originDocument,
+                    pdPage,
+                    pageNumber,
+                    cleanRulings,
+                    stripper.getTextPositionSequences(),
+                    emptyTableCells,
+                    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 REDACT_MANAGER_OLD -> redactManagerBlockificationService.blockify(stripper.getTextPositionSequences(), emptyTableCells);
                 case DOCUMINE -> docuMineBlockificationService.blockify(stripper.getTextPositionSequences(), cleanRulings.getHorizontal(), cleanRulings.getVertical());
+                case REDACT_MANAGER, REDACT_MANAGER_PARAGRAPH_DEBUG -> docstrumBlockificationService.blockify(stripper.getTextPositionSequences(), emptyTableCells, true);
+                case CLARIFYND, CLARIFYND_PARAGRAPH_DEBUG -> docstrumBlockificationService.blockify(stripper.getTextPositionSequences(), emptyTableCells, false);
             };
+
             classificationPage.setCleanRulings(cleanRulings);
             classificationPage.setRotation(rotation);
             classificationPage.setLandscape(isLandscape);
@@ -258,7 +295,21 @@ public class LayoutParsingPipeline {
                 imageServiceResponseAdapter.findOcr(classificationPage);
             }
 
-            tableExtractionService.extractTables(cleanRulings, classificationPage);
+            if (signatures.containsKey(pageNumber)) {
+                if (classificationPage.getImages() == null || classificationPage.getImages().size() == 0) {
+                    classificationPage.setImages(signatures.get(pageNumber));
+                } else {
+                    classificationPage.getImages().addAll(signatures.get(pageNumber));
+                }
+            }
+
+            tableExtractionService.extractTables(emptyTableCells, classificationPage);
+
+            if (layoutParsingType == LayoutParsingType.REDACT_MANAGER) {
+                docstrumBlockificationService.combineBlocks(classificationPage);
+            } else if (layoutParsingType == LayoutParsingType.CLARIFYND) {
+                docstrumBlockificationService.mergeIntersectingBlocks(classificationPage.getTextBlocks(), 0, 6.5f);
+            }
 
             buildPageStatistics(classificationPage);
             increaseDocumentStatistics(classificationPage, classificationDocument);
@@ -272,14 +323,22 @@ public class LayoutParsingPipeline {
         bodyTextFrameService.setBodyTextFrames(classificationDocument, layoutParsingType);
         log.info("Classify TextBlocks for {}", identifier);
         switch (layoutParsingType) {
-            case TAAS -> taasClassificationService.classifyDocument(classificationDocument);
+            case REDACT_MANAGER, REDACT_MANAGER_PARAGRAPH_DEBUG, REDACT_MANAGER_OLD, CLARIFYND_PARAGRAPH_DEBUG ->
+                    redactManagerClassificationService.classifyDocument(classificationDocument);
             case DOCUMINE -> docuMineClassificationService.classifyDocument(classificationDocument);
-            case REDACT_MANAGER -> redactManagerClassificationService.classifyDocument(classificationDocument);
+            case CLARIFYND -> clarifyndClassificationService.classifyDocument(classificationDocument);
         }
 
         log.info("Building Sections for {}", identifier);
-        sectionsBuilderService.buildSections(classificationDocument);
-        sectionsBuilderService.addImagesToSections(classificationDocument);
+
+        switch (layoutParsingType) {
+            case CLARIFYND_PARAGRAPH_DEBUG, REDACT_MANAGER_PARAGRAPH_DEBUG -> sectionsBuilderService.buildParagraphDebugSections(classificationDocument);
+            default -> {
+                sectionsBuilderService.buildSections(classificationDocument);
+                sectionsBuilderService.addImagesToSections(classificationDocument);
+            }
+        }
+
         return classificationDocument;
     }
 
@@ -313,9 +372,7 @@ public class LayoutParsingPipeline {
 
     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 +1,24 @@
 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.ViewerDocumentService;
+
+import io.micrometer.observation.ObservationRegistry;
+
 @Configuration
 @ComponentScan
 public class LayoutParsingServiceProcessorConfiguration {
 
+
+    @Bean
+    @Autowired
+    public ViewerDocumentService viewerDocumentService(ObservationRegistry registry) {
+
+        return new ViewerDocumentService(registry);
+    }
+
 }

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

@@ -8,9 +8,8 @@ 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.pdfbox.Loader;
-import org.apache.pdfbox.pdmodel.PDDocument;
 import org.springframework.stereotype.Service;
 
 import com.fasterxml.jackson.databind.ObjectMapper;
@@ -21,6 +20,7 @@ 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.tenantcommons.TenantContext;
 
 import io.micrometer.observation.annotation.Observed;
@@ -36,13 +36,6 @@ public class LayoutParsingStorageService {
     private final StorageService storageService;
     private final ObjectMapper objectMapper;
 
-
-    public PDDocument getOriginDocument(String storageId) throws IOException {
-
-        return Loader.loadPDF(getOriginFile(storageId));
-    }
-
-
     @Observed(name = "LayoutParsingStorageService", contextualName = "get-origin-file")
     public File getOriginFile(String storageId) throws IOException {
 
@@ -52,6 +45,18 @@ public class LayoutParsingStorageService {
     }
 
 
+    @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);
+        return Optional.of(tempFile);
+    }
+
+
     public ImageServiceResponse getImagesFile(String storageId) throws IOException {
 
         try (InputStream inputStream = getObject(storageId)) {
@@ -73,6 +78,14 @@ public class LayoutParsingStorageService {
         }
     }
 
+    public VisualLayoutParsingResponse getVisualLayoutParsingFile(String storageId) throws IOException {
+
+        try (InputStream inputStream = getObject(storageId)) {
+            VisualLayoutParsingResponse visualLayoutParsingResponse = objectMapper.readValue(inputStream, VisualLayoutParsingResponse.class);
+            return visualLayoutParsingResponse;
+        }
+    }
+
 
     @Observed(name = "LayoutParsingStorageService", contextualName = "store-document-data")
     public void storeDocumentData(LayoutParsingRequest layoutParsingRequest, DocumentData documentData) {
@@ -137,7 +150,6 @@ public class LayoutParsingStorageService {
     public void storeViewerDocument(LayoutParsingRequest layoutParsingRequest, File out) {
 
         try (var in = new FileInputStream(out)) {
-
             storageService.storeObject(TenantContext.getTenantId(), layoutParsingRequest.viewerDocumentStorageId(), in);
         }
     }

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

@@ -0,0 +1,59 @@
+package com.knecon.fforesight.service.layoutparser.processor.docstrum;
+
+import java.util.ArrayList;
+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.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.text.TextDirection;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
+
+import lombok.RequiredArgsConstructor;
+
+@Service
+@RequiredArgsConstructor
+public class DocstrumSegmentationService {
+
+    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) {
+
+        List<Zone> zones = new ArrayList<>();
+        zones.addAll(computeZones(textPositions, TextDirection.ZERO));
+        zones.addAll(computeZones(textPositions, TextDirection.QUARTER_CIRCLE));
+        zones.addAll(computeZones(textPositions, TextDirection.HALF_CIRCLE));
+        zones.addAll(computeZones(textPositions, TextDirection.THREE_QUARTER_CIRCLE));
+
+        return readingOrderService.resolve(zones, xyOrder);
+    }
+
+
+    private List<Zone> computeZones(List<TextPositionSequence> textPositions, TextDirection direction) {
+
+        var positions = textPositions.stream().filter(t -> t.getDir() == direction).map(TextPositionSequence::getTextPositions).flatMap(List::stream).toList();
+
+        var characters = positions.stream().map(Character::new).collect(Collectors.toList());
+
+        nearestNeighbourService.findNearestNeighbors(characters);
+
+        var characterSpacing = spacingService.computeCharacterSpacing(characters);
+        var lineSpacing = Math.min(spacingService.computeLineSpacing(characters), 20);
+
+        var lines = lineBuilderService.buildLines(characters, characterSpacing, lineSpacing);
+        return zoneBuilderService.buildZones(lines, characterSpacing, lineSpacing);
+    }
+
+}

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

@@ -0,0 +1,25 @@
+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) {
+
+        if (lowerAngle <= upperAngle) {
+            return lowerAngle <= neighbor.getAngle() && neighbor.getAngle() < upperAngle;
+        } else {
+            return lowerAngle <= neighbor.getAngle() || neighbor.getAngle() < upperAngle;
+        }
+    }
+
+}

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

@@ -0,0 +1,54 @@
+package com.knecon.fforesight.service.layoutparser.processor.docstrum.model;
+
+import java.awt.geom.Rectangle2D;
+
+import lombok.Data;
+
+@Data
+public abstract class BoundingBox {
+
+    private Rectangle2D bBox;
+
+
+    public double getX() {
+
+        return bBox.getX();
+    }
+
+
+    public double getY() {
+
+        return bBox.getY();
+    }
+
+
+    public double getWidth() {
+
+        return bBox.getWidth();
+    }
+
+
+    public double getHeight() {
+
+        return bBox.getHeight();
+    }
+
+
+    public double getArea() {
+
+        return (bBox.getHeight() * bBox.getWidth());
+    }
+
+
+    public boolean contains(Rectangle2D contained, double tolerance) {
+
+        return bBox.getX() <= contained.getX() + tolerance && bBox.getY() <= contained.getY() + tolerance && bBox.getX() + bBox.getWidth() >= contained.getX() + contained.getWidth() - tolerance && bBox.getY() + bBox.getHeight() >= contained.getY() + contained.getHeight() - tolerance;
+    }
+
+
+    public boolean intersectsY(BoundingBox other) {
+
+        return this.getBBox().getMinY() <= other.getBBox().getMaxY() && this.getBBox().getMaxY() >= other.getBBox().getMinY();
+    }
+
+}

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.getXDirAdj() + chunk.getWidthDirAdj() / 2;
+        this.y = chunk.getYDirAdj() + chunk.getHeightDir() / 2;
+        this.textPosition = chunk;
+    }
+
+
+    public double getHeight() {
+
+        return textPosition.getHeightDir();
+    }
+
+
+    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.getHeightDir());
+        xs[2] = c * other.x - s * other.y;
+        xs[3] = c * (other.x + other.textPosition.getWidthDirAdj()) - s * (other.y + other.textPosition.getHeightDir());
+        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/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 +1,170 @@
+package com.knecon.fforesight.service.layoutparser.processor.docstrum.model;
+
+import java.awt.geom.Rectangle2D;
+import java.util.ArrayList;
+import java.util.Arrays;
+import java.util.List;
+
+import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
+
+import lombok.Data;
+import lombok.EqualsAndHashCode;
+
+@Data
+@EqualsAndHashCode(onlyExplicitlyIncluded = true)
+public class Line extends BoundingBox {
+
+    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 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();
+    }
+
+
+    public double getAngle() {
+
+        return Math.atan2(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) / Math.sqrt(1);
+    }
+
+
+    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() {
+
+        double minX = Double.POSITIVE_INFINITY;
+        double minY = Double.POSITIVE_INFINITY;
+        double maxX = Double.NEGATIVE_INFINITY;
+        double maxY = Double.NEGATIVE_INFINITY;
+
+        for (Character character : characters) {
+
+            minX = Math.min(minX, character.getTextPosition().getXDirAdj());
+            minY = Math.min(minY, character.getTextPosition().getYDirAdj());
+            maxX = Math.max(maxX, character.getTextPosition().getXDirAdj() + character.getTextPosition().getWidthDirAdj());
+            maxY = Math.max(maxY, character.getTextPosition().getYDirAdj() + character.getTextPosition().getHeightDir());
+
+        }
+
+        this.setBBox(new Rectangle2D.Double(minX, minY, maxX - minX, maxY - minY));
+    }
+
+
+    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/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 +1,51 @@
+package com.knecon.fforesight.service.layoutparser.processor.docstrum.model;
+
+import java.awt.geom.Rectangle2D;
+import java.util.Comparator;
+import java.util.List;
+
+import lombok.Data;
+
+@Data
+public class Zone extends BoundingBox {
+
+    private List<Line> lines;
+
+
+    @SuppressWarnings("PMD.ConstructorCallsOverridableMethod")
+    public Zone(List<Line> lines) {
+
+        lines.sort(Comparator.comparingDouble(Line::getY));
+        this.lines = lines;
+        buildBBox();
+    }
+
+
+    public void buildBBox() {
+
+        double minX = Double.POSITIVE_INFINITY;
+        double minY = Double.POSITIVE_INFINITY;
+        double maxX = Double.NEGATIVE_INFINITY;
+        double maxY = Double.NEGATIVE_INFINITY;
+
+        for (Line line : lines) {
+
+            minX = Math.min(minX, line.getX());
+            minY = Math.min(minY, line.getY());
+            maxX = Math.max(maxX, line.getX() + line.getWidth());
+            maxY = Math.max(maxY, line.getY() + line.getHeight());
+
+        }
+
+        this.setBBox(new Rectangle2D.Double(minX, minY, maxX - minX, maxY - minY));
+    }
+
+
+    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 +1,53 @@
+package com.knecon.fforesight.service.layoutparser.processor.docstrum.service;
+
+import java.util.ArrayList;
+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;
+
+@Service
+public class LineBuilderService {
+
+    private static final double CHARACTER_SPACING_DISTANCE_MULTIPLIER = 3.5;
+    private static final double MAX_VERTICAL_CHARACTER_DISTANCE = 0.67;
+    private static final double ANGLE_TOLERANCE = Math.PI / 6;
+
+
+    public List<Line> buildLines(List<Character> characters, double characterSpacing, double lineSpacing) {
+
+        double maxHorizontalDistance = characterSpacing * CHARACTER_SPACING_DISTANCE_MULTIPLIER;
+        double maxVerticalDistance = lineSpacing * MAX_VERTICAL_CHARACTER_DISTANCE;
+
+        UnionFind<Character> unionFind = new UnionFind<>(new HashSet<>(characters));
+
+        AngleFilter filter = new AngleFilter(-ANGLE_TOLERANCE, ANGLE_TOLERANCE);
+
+        characters.forEach(character -> {
+            character.getNeighbors().forEach(neighbor -> {
+                double x = neighbor.getHorizontalDistance() / maxHorizontalDistance;
+                double y = neighbor.getVerticalDistance() / maxVerticalDistance;
+                if (character.getTextPosition().getDir() == neighbor.getCharacter().getTextPosition().getDir() && filter.matches(neighbor) && Math.pow(x, 2) + Math.pow(y,
+                        2) <= 1) {
+                    unionFind.union(character, neighbor.getCharacter());
+                }
+            });
+        });
+
+        List<Line> lines = new ArrayList<>();
+        unionFind.getGroups().forEach(group -> {
+            List<Character> lineCharacters = new ArrayList<>(group);
+            lineCharacters.sort(Comparator.comparingDouble(Character::getX));
+            lines.add(new Line(lineCharacters, characterSpacing));
+        });
+
+        return lines;
+    }
+
+}

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 +1,165 @@
+package com.knecon.fforesight.service.layoutparser.processor.docstrum.service;
+
+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 org.springframework.stereotype.Service;
+
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.BoundingBox;
+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;
+
+
+    public List<Zone> resolve(List<Zone> zones, boolean xyReadingOrder) {
+
+        if (zones.isEmpty() || zones.size() == 1) {
+            return zones;
+        }
+
+        if (xyReadingOrder) {
+            return resolveSingleColumnReadingOrder(zones);
+        }
+
+        Map<Long, Integer> histogram = new HashMap<>();
+        for (Zone zone : zones) {
+            long minY = Math.round(zone.getBBox().getMinY());
+            long maxY = Math.round(zone.getBBox().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);
+        } else {
+
+            return resolveMultiColumnReadingOder(zones);
+        }
+
+    }
+
+
+    private static List<Zone> resolveSingleColumnReadingOrder(List<Zone> zones) {
+
+        zones.sort(Comparator.comparing(BoundingBox::getY, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD))
+                .thenComparing(BoundingBox::getX, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD)));
+        return zones;
+    }
+
+
+    private List<Zone> resolveMultiColumnReadingOder(List<Zone> zones) {
+
+        // 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) {
+            if (zone.getX() < minX) {
+                minX = zone.getX();
+            }
+            if (zone.getX() + zone.getWidth() > maxX) {
+                maxX = zone.getX() + zone.getWidth();
+            }
+        }
+
+        double midLineXCoordinate = (minX + maxX) / 2;
+
+        List<Zone> leftOf = new ArrayList<>();
+        List<Zone> rightOf = new ArrayList<>();
+        List<Zone> middle = new ArrayList<>();
+        for (Zone zone : zones) {
+            if (zone.getX() < midLineXCoordinate && zone.getX() + zone.getWidth() < midLineXCoordinate) {
+                leftOf.add(zone);
+            } else if (zone.getX() > midLineXCoordinate && zone.getX() + zone.getWidth() > midLineXCoordinate) {
+                rightOf.add(zone);
+            } else {
+                middle.add(zone);
+            }
+        }
+
+        leftOf.sort(Comparator.comparing(BoundingBox::getY, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD))
+                .thenComparing(BoundingBox::getX, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD)));
+
+        rightOf.sort(Comparator.comparing(BoundingBox::getY, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD))
+                .thenComparing(BoundingBox::getX, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD)));
+
+        middle.sort(Comparator.comparing(BoundingBox::getY, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD))
+                .thenComparing(BoundingBox::getX, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD)));
+
+        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();
+            for (int i = 0; i < sortedZones.size(); i++) {
+                if (current.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 +1,152 @@
+package com.knecon.fforesight.service.layoutparser.processor.docstrum.service;
+
+import java.util.ArrayList;
+import java.util.Comparator;
+import java.util.HashSet;
+import java.util.List;
+import java.util.Set;
+
+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;
+
+@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 = -3.0;
+
+    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 int MAX_ZONES = 300;
+
+    private static final double MAX_VERTICAL_MERGE_DISTANCE = 0.5;
+
+
+    public List<Zone> buildZones(List<Line> lines, double characterSpacing, double lineSpacing) {
+
+        double minHorizontalDistance = characterSpacing * MIN_HORIZONTAL_DISTANCE_MULTIPLIER;
+        double maxVerticalDistance = lineSpacing * MAX_VERTICAL_DISTANCE_MULTIPLIER;
+        double minHorizontalMergeDistance = characterSpacing * 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 -> {
+
+                    double scale = Math.min(outerLine.getHeight(), innerLine.getHeight()) / meanHeight;
+                    scale = Math.max(MIN_LINE_SIZE_SCALE, Math.min(scale, MAX_LINE_SIZE_SCALE));
+
+                    if (!unionFind.inSameSet(outerLine, innerLine) && outerLine.angularDifference(innerLine) <= ANGLE_TOLERANCE) {
+
+                        double horizontalDistance = outerLine.horizontalDistance(innerLine) / scale;
+                        double verticalDistance = outerLine.verticalDistance(innerLine) / scale;
+
+                        if (minHorizontalDistance <= horizontalDistance && verticalDistance <= maxVerticalDistance //
+                                || minHorizontalMergeDistance <= horizontalDistance && verticalDistance <= maxVerticalMergeDistance) {
+                            unionFind.union(outerLine, innerLine);
+                        }
+                    }
+                }));
+
+        List<Zone> zones = new ArrayList<>();
+        unionFind.getGroups().forEach(group -> {
+            zones.add(mergeLinesInZone(new ArrayList<>(group), characterSpacing, lineSpacing));
+        });
+
+        if (zones.size() > MAX_ZONES) {
+            List<Line> oneZoneLines = new ArrayList<>();
+            for (Zone zone : zones) {
+                oneZoneLines.addAll(zone.getLines());
+            }
+            return List.of(mergeLinesInZone(oneZoneLines, characterSpacing, lineSpacing));
+        }
+
+        return zones;
+    }
+
+
+    private double calculateMeanHeight(List<Line> lines) {
+
+        double meanHeight = 0.0;
+        double weights = 0.0;
+        for (Line line : lines) {
+            double weight = line.getLength();
+            meanHeight += line.getHeight() * 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) {
+
+                    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);
+    }
+
+}

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

@@ -0,0 +1,15 @@
+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);
+        }
+        long i1 = Math.round(d1 / (precision == 0 ? 1 : precision));
+        long i2 = Math.round(d2 / (precision == 0 ? 1 : precision));
+        return Long.compare(i1, i2);
+    }
+
+}

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

@@ -6,12 +6,14 @@ import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageB
 
 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 Rectangle {
 
     @JsonIgnore
     protected float minX;
@@ -94,7 +96,7 @@ public abstract class AbstractPageBlock {
 
         return this.minX - threshold <= apb.getMaxX() && this.maxX + threshold >= apb.getMinX();
     }
-
+    
 
     public abstract boolean isEmpty();
 

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

@@ -12,6 +12,7 @@ import lombok.Setter;
 
 @Setter
 @EqualsAndHashCode
+@SuppressWarnings("PMD.AvoidFieldNameMatchingMethodName")
 public class Boundary implements Comparable<Boundary> {
 
     private int start;

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

@@ -10,11 +10,11 @@ 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.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 lombok.AccessLevel;
 import lombok.AllArgsConstructor;
@@ -30,6 +30,9 @@ import lombok.experimental.FieldDefaults;
 @FieldDefaults(level = AccessLevel.PRIVATE)
 public class Document implements GenericSemanticNode {
 
+    @Builder.Default
+    Set<LayoutEngine> engines = new HashSet<>(Set.of(LayoutEngine.ALGORITHM));
+
     Set<Page> pages;
     DocumentTree documentTree;
     Integer numberOfPages;
@@ -48,7 +51,7 @@ public class Document implements GenericSemanticNode {
     public TextBlock getTextBlock() {
 
         if (textBlock == null) {
-            textBlock = streamTerminalTextBlocksInOrder().collect(new TextBlockCollector());
+            textBlock = GenericSemanticNode.super.getTextBlock();
         }
         return textBlock;
     }
@@ -56,13 +59,29 @@ public class Document implements GenericSemanticNode {
 
     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());
+    }
+
+
+    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());
     }
 
 
     public Stream<TextBlock> streamTerminalTextBlocksInOrder() {
 
-        return streamAllNodes().filter(SemanticNode::isLeaf).map(SemanticNode::getLeafTextBlock);
+        return streamAllNodes().filter(SemanticNode::isLeaf)
+                .map(SemanticNode::getTextBlock);
     }
 
 
@@ -83,13 +102,16 @@ public class Document implements GenericSemanticNode {
     @Override
     public Headline getHeadline() {
 
-        return streamAllSubNodesOfType(NodeType.HEADLINE).map(node -> (Headline) node).findFirst().orElse(Headline.builder().build());
+        return streamAllSubNodesOfType(NodeType.HEADLINE).map(node -> (Headline) node)
+                .findFirst()
+                .orElse(Headline.builder().build());
     }
 
 
     private Stream<SemanticNode> streamAllNodes() {
 
-        return documentTree.allEntriesInOrder().map(DocumentTree.Entry::getNode);
+        return documentTree.allEntriesInOrder()
+                .map(DocumentTree.Entry::getNode);
     }
 
 

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

@@ -0,0 +1,34 @@
+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(leafTextBlock, unsortedLeafTextBlock).collect(new TextBlockCollector());
+
+    }
+
+
+    @Override
+    public String toString() {
+
+        return super.toString();
+    }
+
+}

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

@@ -6,6 +6,7 @@ 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.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;
@@ -26,6 +27,9 @@ import lombok.experimental.FieldDefaults;
 @FieldDefaults(level = AccessLevel.PRIVATE)
 public class Footer implements GenericSemanticNode {
 
+    @Builder.Default
+    Set<LayoutEngine> engines = new HashSet<>(Set.of(LayoutEngine.ALGORITHM));
+
     List<Integer> treeId;
     TextBlock leafTextBlock;
 

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

@@ -6,6 +6,7 @@ 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.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;
@@ -26,6 +27,8 @@ import lombok.experimental.FieldDefaults;
 @FieldDefaults(level = AccessLevel.PRIVATE)
 public class Header implements GenericSemanticNode {
 
+    @Builder.Default
+    Set<LayoutEngine> engines = new HashSet<>(Set.of(LayoutEngine.ALGORITHM));
     List<Integer> treeId;
     TextBlock leafTextBlock;
 

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

@@ -6,6 +6,7 @@ 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.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;
@@ -26,6 +27,8 @@ import lombok.experimental.FieldDefaults;
 @FieldDefaults(level = AccessLevel.PRIVATE)
 public class Headline implements GenericSemanticNode {
 
+    @Builder.Default
+    Set<LayoutEngine> engines = new HashSet<>(Set.of(LayoutEngine.ALGORITHM));
     List<Integer> treeId;
     TextBlock leafTextBlock;
 

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

@@ -8,11 +8,11 @@ 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.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 lombok.AccessLevel;
 import lombok.AllArgsConstructor;
@@ -29,6 +29,9 @@ import lombok.experimental.FieldDefaults;
 @FieldDefaults(level = AccessLevel.PRIVATE)
 public class Image implements GenericSemanticNode {
 
+    @Builder.Default
+    Set<LayoutEngine> engines = new HashSet<>(Set.of(LayoutEngine.ALGORITHM));
+
     List<Integer> treeId;
     String id;
 
@@ -36,6 +39,8 @@ public class Image implements GenericSemanticNode {
     boolean transparent;
     Rectangle2D position;
 
+    TextBlock leafTextBlock;
+
     boolean redaction;
     boolean ignored;
     @Builder.Default
@@ -66,7 +71,7 @@ public class Image implements GenericSemanticNode {
     @Override
     public TextBlock getTextBlock() {
 
-        return streamAllSubNodes().filter(SemanticNode::isLeaf).map(SemanticNode::getLeafTextBlock).collect(new TextBlockCollector());
+        return leafTextBlock;
     }
 
 
@@ -92,4 +97,11 @@ public class Image implements GenericSemanticNode {
         return bBoxPerPage;
     }
 
+
+    @Override
+    public boolean isLeaf() {
+
+        return true;
+    }
+
 }

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

@@ -6,8 +6,11 @@ public enum ImageType {
     LOGO,
     FORMULA,
     SIGNATURE,
+
+    SIGNATURE_VISUAL,
     OTHER,
-    OCR;
+    OCR,
+    GRAPHIC;
 
 
     public static ImageType fromString(String imageType) {
@@ -17,6 +20,7 @@ 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/Paragraph.java

@@ -6,6 +6,7 @@ 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.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;
@@ -17,13 +18,17 @@ 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)
+@FieldDefaults(level = AccessLevel.PROTECTED)
 public class Paragraph implements GenericSemanticNode {
 
+    @Builder.Default
+    Set<LayoutEngine> engines = new HashSet<>(Set.of(LayoutEngine.ALGORITHM));
+
     List<Integer> treeId;
     TextBlock leafTextBlock;
 

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

@@ -6,11 +6,11 @@ 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.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 lombok.AccessLevel;
 import lombok.AllArgsConstructor;
@@ -27,6 +27,8 @@ import lombok.extern.slf4j.Slf4j;
 @FieldDefaults(level = AccessLevel.PRIVATE)
 public class Section implements GenericSemanticNode {
 
+    @Builder.Default
+    Set<LayoutEngine> engines = new HashSet<>(Set.of(LayoutEngine.ALGORITHM));
     List<Integer> treeId;
 
     TextBlock textBlock;
@@ -50,7 +52,8 @@ public class Section implements GenericSemanticNode {
 
     public boolean hasTables() {
 
-        return streamAllSubNodesOfType(NodeType.TABLE).findAny().isPresent();
+        return streamAllSubNodesOfType(NodeType.TABLE).findAny()
+                .isPresent();
     }
 
 
@@ -58,7 +61,7 @@ public class Section implements GenericSemanticNode {
     public TextBlock getTextBlock() {
 
         if (textBlock == null) {
-            textBlock = streamAllSubNodes().filter(SemanticNode::isLeaf).map(SemanticNode::getLeafTextBlock).collect(new TextBlockCollector());
+            textBlock = GenericSemanticNode.super.getTextBlock();
         }
         return textBlock;
     }

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

@@ -12,6 +12,7 @@ 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;
@@ -19,6 +20,7 @@ import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.E
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
 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.RectangleTransformations;
 
 public interface SemanticNode {
@@ -38,7 +40,10 @@ 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());
+    }
 
 
     /**
@@ -335,6 +340,23 @@ public interface SemanticNode {
     }
 
 
+    /**
+     * 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.
      *
@@ -426,6 +448,7 @@ public interface SemanticNode {
     /**
      * TODO: this produces unwanted results for sections spanning multiple columns.
      * Computes the Union of the bounding boxes of all children recursively.
+     *
      * @return The union of the BoundingBoxes of all children
      */
     private Map<Page, Rectangle2D> getBBoxFromChildren() {

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

@@ -12,6 +12,7 @@ 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;
@@ -31,6 +32,8 @@ 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;
 
@@ -53,8 +56,7 @@ public class Table implements SemanticNode {
      */
     public Stream<RedactionEntity> 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)
@@ -71,8 +73,11 @@ 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);
     }
 
 
@@ -85,9 +90,13 @@ public class Table implements SemanticNode {
      */
     public Stream<RedactionEntity> 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);
     }
 
 
@@ -100,9 +109,13 @@ public class Table implements SemanticNode {
      */
     public Stream<RedactionEntity> 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);
     }
 
 
@@ -114,12 +127,15 @@ public class Table implements SemanticNode {
      */
     public Stream<RedactionEntity> 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);
     }
 
 
@@ -131,18 +147,24 @@ public class Table implements SemanticNode {
      */
     public Stream<RedactionEntity> 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(RedactionEntity::getType)
+                .distinct();
     }
 
 
@@ -159,7 +181,8 @@ 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();
     }
 
 
@@ -196,7 +219,8 @@ 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));
     }
 
 
@@ -208,9 +232,11 @@ 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));
     }
 
+
     /**
      * Streams all TableCells row-wise and filters them with header == true.
      *
@@ -231,7 +257,8 @@ 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);
     }
 
 
@@ -304,7 +331,7 @@ 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;
     }
@@ -315,6 +342,8 @@ public class Table implements SemanticNode {
 
         return treeId.toString() + ": " + NodeType.TABLE + ": #cols: " + numberOfCols + ", #rows: " + numberOfRows + ", " + this.getTextBlock().buildSummary();
     }
+
+
     @Override
     public Map<Page, Rectangle2D> getBBox() {
 
@@ -323,4 +352,5 @@ public class Table implements SemanticNode {
         }
         return bBoxCache;
     }
+
 }

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

@@ -7,6 +7,7 @@ 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.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;
@@ -26,6 +27,8 @@ import lombok.experimental.FieldDefaults;
 @FieldDefaults(level = AccessLevel.PRIVATE)
 public class TableCell implements GenericSemanticNode {
 
+    @Builder.Default
+    Set<LayoutEngine> engines = new HashSet<>(Set.of(LayoutEngine.ALGORITHM));
     List<Integer> treeId;
     int row;
     int col;
@@ -84,14 +87,16 @@ public class TableCell implements GenericSemanticNode {
 
     private TextBlock buildTextBlock() {
 
-        return streamAllSubNodes().filter(SemanticNode::isLeaf).map(SemanticNode::getLeafTextBlock).collect(new TextBlockCollector());
+        return streamAllSubNodes().filter(SemanticNode::isLeaf)
+                .map(SemanticNode::getLeafTextBlock)
+                .collect(new TextBlockCollector());
     }
 
 
     @Override
     public String toString() {
 
-        return treeId + ": " + NodeType.TABLE_CELL + ": " + this.buildTextBlock().buildSummary();
+        return treeId + ": " + NodeType.TABLE_CELL + ": " + this.getTextBlock().buildSummary();
     }
 
 }

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

@@ -4,18 +4,21 @@ 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
 @RequiredArgsConstructor
+@AllArgsConstructor
 public class ClassifiedImage {
 
     @NonNull
     private Rectangle2D position;
     @NonNull
     private ImageType imageType;
+    private boolean sourceByAi;
     private boolean isAppendedToSection;
     @NonNull
     private boolean hasTransparency;

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

@@ -1,6 +1,7 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.table;
 
 import java.awt.geom.Point2D;
+import java.awt.geom.Rectangle2D;
 import java.util.ArrayList;
 import java.util.Iterator;
 import java.util.List;
@@ -36,6 +37,12 @@ public class Cell extends Rectangle {
     }
 
 
+    public Cell(Rectangle2D r) {
+
+        super((float) r.getY(), (float) r.getX(), (float) r.getWidth(), (float) r.getHeight());
+    }
+
+
     public void addTextBlock(TextPageBlock textBlock) {
 
         textBlocks.add(textBlock);

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

@@ -20,7 +20,8 @@ import lombok.extern.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 Ruling(Point2D p1, Point2D p2) {
@@ -110,8 +111,8 @@ public class Ruling extends Line2D.Float {
         });
 
         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));
+            sos.add(new SortObject(SOType.HLEFT, h.getLeft() - PERPENDICULAR_UNIT_EXPAND_AMOUNT, h));
+            sos.add(new SortObject(SOType.HRIGHT, h.getRight() + PERPENDICULAR_UNIT_EXPAND_AMOUNT, h));
         }
 
         for (Ruling v : verticals) {
@@ -151,7 +152,7 @@ public class Ruling extends Line2D.Float {
                             if (i == null) {
                                 continue;
                             }
-                            rv.put(i, new Ruling[]{h.getKey().expand(PERPENDICULAR_PIXEL_EXPAND_AMOUNT), so.ruling.expand(PERPENDICULAR_PIXEL_EXPAND_AMOUNT)});
+                            rv.put(i, new Ruling[]{h.getKey().expand(PERPENDICULAR_UNIT_EXPAND_AMOUNT), so.ruling.expand(PERPENDICULAR_UNIT_EXPAND_AMOUNT)});
                         } catch (UnsupportedOperationException e) {
                             log.info("Some line are oblique, ignoring...");
                             continue;
@@ -267,7 +268,7 @@ public class Ruling extends Line2D.Float {
     }
 
 
-    public boolean nearlyIntersects(Ruling another, int colinearOrParallelExpandAmount) {
+    public boolean nearlyIntersects(Ruling another) {
 
         if (this.intersectsLine(another)) {
             return true;
@@ -276,9 +277,9 @@ 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;
@@ -319,8 +320,8 @@ 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 this_l = this.expand(PERPENDICULAR_UNIT_EXPAND_AMOUNT);
+        Ruling other_l = other.expand(PERPENDICULAR_UNIT_EXPAND_AMOUNT);
         Ruling horizontal, vertical;
 
         if (!this_l.intersectsLine(other_l)) {

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

@@ -1,18 +1,18 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.table;
 
 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.HashSet;
 import java.util.List;
 import java.util.Set;
 import java.util.TreeMap;
-import java.util.stream.Collectors;
 
 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;
@@ -21,7 +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
@@ -30,10 +31,14 @@ 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) {
 
+        this.cells = cells;
         addCells(cells);
         minX = area.getLeft();
         minY = area.getBottom();
@@ -50,6 +55,7 @@ public class TablePageBlock extends AbstractPageBlock {
         return getColCount() == 0 || getRowCount() == 0;
     }
 
+
     public List<List<Cell>> getRows() {
 
         if (rows == null) {
@@ -80,14 +86,17 @@ 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() {
@@ -95,7 +104,7 @@ 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);
@@ -120,7 +129,8 @@ 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);
                     }
@@ -135,7 +145,8 @@ 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);
                 }
             }
@@ -151,7 +162,7 @@ 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);
                     }
@@ -162,7 +173,7 @@ 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);
                     }
@@ -173,7 +184,7 @@ 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);
                     }
@@ -187,17 +198,6 @@ 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()) {
@@ -206,11 +206,12 @@ 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);
             }
         }
 
@@ -221,29 +222,36 @@ 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());
-        });
+        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());
+                });
 
-        var sortedUniqueX = uniqueX.stream().sorted().toList();
-        var sortedUniqueY = uniqueY.stream().sorted().toList();
+        var sortedUniqueX = uniqueX.stream()
+                .sorted()
+                .toList();
+        var sortedUniqueY = uniqueY.stream()
+                .sorted()
+                .toList();
+
+        List<List<Cell>> rowsOfCells = new ArrayList<>();
 
         Float prevY = null;
+
         for (Float y : sortedUniqueY) {
 
             List<Cell> row = new ArrayList<>();
@@ -252,44 +260,87 @@ public class TablePageBlock extends AbstractPageBlock {
             for (Float 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.Float(prevX, prevY), new Point2D.Float(x, y));
 
-                    var intersectionCell = cells.stream().filter(c -> intersects(cell, c)).findFirst();
+                    if (cellFromGridStructure.hasMinimumSize()) {
 
-                    intersectionCell.ifPresent(value -> cell.getTextBlocks().addAll(value.getTextBlocks()));
-                    if (cell.hasMinimumSize()) {
-                        row.add(cell);
+                        cells.stream()
+                                .map(originalCell -> new CellWithIntersection(originalCell, RectangleTransformations.calculateIntersectedArea(cellFromGridStructure, originalCell)))
+                                .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);
 
-        return matrix;
-    }
-
-
-
-    public boolean intersects(Cell cell1, Cell cell2) {
-        if (cell1.getHeight() <= 0 || cell2.getHeight() <= 0) {
-            return false;
+        // 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<>());
         }
-        double x0 = cell1.getX() + 2;
-        double y0 = cell1.getY() + 2;
-        return (cell2.x + cell2.width > x0 &&
-                cell2.y + cell2.height > y0 &&
-                cell2.x < x0 + cell1.getWidth() -2 &&
-                cell2.y < y0 + cell1.getHeight() -2);
+
+        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);
+
+    }
+
 
     @Override
     public String getText() {
@@ -314,7 +365,7 @@ 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;
                         }
                     }
@@ -328,8 +379,6 @@ public class TablePageBlock extends AbstractPageBlock {
     }
 
 
-
-
     public String getTextAsHtml() {
 
         StringBuilder sb = new StringBuilder();
@@ -363,4 +412,9 @@ 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/RedTextPosition.java

@@ -1,7 +1,6 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.text;
 
 import org.apache.pdfbox.text.TextPosition;
-import org.springframework.beans.BeanUtils;
 
 import com.fasterxml.jackson.annotation.JsonIgnore;
 
@@ -50,10 +49,14 @@ public class RedTextPosition {
     public static RedTextPosition fromTextPosition(TextPosition textPosition) {
 
         var pos = new RedTextPosition();
-        BeanUtils.copyProperties(textPosition, pos);
-        pos.setFontName(textPosition.getFont().getName());
-
+        pos.setRotation(textPosition.getRotation());
+        pos.setPageHeight(textPosition.getPageHeight());
+        pos.setPageWidth(textPosition.getPageWidth());
+        pos.setUnicode(textPosition.getUnicode());
+        pos.setDir(textPosition.getDir());
+        pos.setWidthOfSpace(textPosition.getWidthOfSpace());
         pos.setFontSizeInPt(textPosition.getFontSizeInPt());
+        pos.setFontName(textPosition.getFont().getName());
 
         var position = new float[4];
 

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

@@ -53,6 +53,9 @@ public class TextPageBlock extends AbstractPageBlock {
     @JsonIgnore
     private PageBlockType classification;
 
+    @JsonIgnore
+    private boolean toDuplicate;
+
 
     @JsonIgnore
     public TextDirection getDir() {
@@ -73,7 +76,7 @@ public class TextPageBlock extends AbstractPageBlock {
 
         return sequences.get(0).getPageWidth();
     }
-    
+
 
     public static TextPageBlock merge(List<TextPageBlock> textBlocksToMerge) {
 
@@ -82,6 +85,7 @@ public class TextPageBlock extends AbstractPageBlock {
         return fromTextPositionSequences(sequences);
     }
 
+
     public static TextPageBlock fromTextPositionSequences(List<TextPositionSequence> wordBlockList) {
 
         TextPageBlock textBlock = null;
@@ -133,7 +137,6 @@ public class TextPageBlock extends AbstractPageBlock {
     }
 
 
-
     /**
      * Returns the minX value in pdf coordinate system.
      * Note: This needs to use Pdf Coordinate System where {0,0} rotated with the page rotation.
@@ -362,7 +365,22 @@ public class TextPageBlock extends AbstractPageBlock {
         }
 
         return TextNormalizationUtilities.removeHyphenLineBreaks(sb.toString());
+    }
 
+
+    public int getNumberOfLines() {
+
+        int numberOfLines = 1;
+        TextPositionSequence previous = null;
+        for (TextPositionSequence word : sequences) {
+            if (previous != null) {
+                if (word.getMaxYDirAdj() - previous.getMaxYDirAdj() > word.getTextHeight()) {
+                    numberOfLines++;
+                }
+            }
+            previous = word;
+        }
+        return numberOfLines;
     }
 
 

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

@@ -15,6 +15,7 @@ import com.iqser.red.service.persistence.service.v1.api.shared.model.redactionlo
 import lombok.AllArgsConstructor;
 import lombok.Builder;
 import lombok.Data;
+import lombok.EqualsAndHashCode;
 import lombok.NoArgsConstructor;
 import lombok.SneakyThrows;
 import lombok.extern.slf4j.Slf4j;
@@ -24,12 +25,18 @@ import lombok.extern.slf4j.Slf4j;
 @Builder
 @NoArgsConstructor
 @AllArgsConstructor
+@EqualsAndHashCode(onlyExplicitlyIncluded = true)
 public class TextPositionSequence implements CharSequence {
 
     public static final int HEIGHT_PADDING = 2;
+
+    @EqualsAndHashCode.Include
     private int page;
+
+    @EqualsAndHashCode.Include
     private List<RedTextPosition> textPositions = new ArrayList<>();
 
+    @EqualsAndHashCode.Include
     private TextDirection dir;
     private int rotation;
     private float pageHeight;
@@ -55,6 +62,17 @@ public class TextPositionSequence implements CharSequence {
     }
 
 
+    public TextPositionSequence(List<RedTextPosition> textPositions, int page) {
+
+        this.textPositions = textPositions;
+        this.page = page;
+        this.dir = TextDirection.fromDegrees(textPositions.get(0).getDir());
+        this.rotation = textPositions.get(0).getRotation();
+        this.pageHeight = textPositions.get(0).getPageHeight();
+        this.pageWidth = textPositions.get(0).getPageWidth();
+    }
+
+
     @Override
     public int length() {
 
@@ -190,6 +208,12 @@ public class TextPositionSequence implements CharSequence {
     }
 
 
+    public float getTextHeightNoPadding() {
+
+        return textPositions.get(0).getHeightDir();
+    }
+
+
     public float getTextHeight() {
 
         return textPositions.get(0).getHeightDir() + HEIGHT_PADDING;

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

@@ -1,8 +0,0 @@
-package com.knecon.fforesight.service.layoutparser.processor.model.visualization;
-
-import java.awt.Color;
-import java.awt.geom.Line2D;
-
-public record ColoredLine(Line2D line, Color color) {
-
-}

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

@@ -1,7 +0,0 @@
-package com.knecon.fforesight.service.layoutparser.processor.model.visualization;
-
-import java.awt.geom.Point2D;
-
-public record PlacedText(String text, Point2D lineStart) {
-
-}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/python_api/adapter/ImageServiceResponseAdapter.java

@@ -9,10 +9,10 @@ import java.util.Map;
 
 import org.springframework.stereotype.Service;
 
-import com.knecon.fforesight.service.layoutparser.processor.python_api.model.image.ImageServiceResponse;
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
-import com.knecon.fforesight.service.layoutparser.processor.model.image.ClassifiedImage;
 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.python_api.model.image.ImageServiceResponse;
 
 import lombok.RequiredArgsConstructor;
 
@@ -20,8 +20,7 @@ import lombok.RequiredArgsConstructor;
 @RequiredArgsConstructor
 public class ImageServiceResponseAdapter {
 
-
-    public Map<Integer, List<ClassifiedImage>> buildClassifiedImagesPerPage(ImageServiceResponse imageServiceResponse ) {
+    public Map<Integer, List<ClassifiedImage>> buildClassifiedImagesPerPage(ImageServiceResponse imageServiceResponse) {
 
         Map<Integer, List<ClassifiedImage>> images = new HashMap<>();
         imageServiceResponse.getData().forEach(imageMetadata -> {

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/python_api/adapter/VisualLayoutParsingAdapter.java

@@ -0,0 +1,92 @@
+package com.knecon.fforesight.service.layoutparser.processor.python_api.adapter;
+
+import java.awt.geom.Rectangle2D;
+import java.util.ArrayList;
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+
+import org.springframework.stereotype.Service;
+
+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.python_api.model.table.VisualLayoutParsingBox;
+import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.VisualLayoutParsingResponse;
+import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.VisualLayoutParsingResult;
+
+import lombok.RequiredArgsConstructor;
+import lombok.extern.slf4j.Slf4j;
+
+@Service
+@RequiredArgsConstructor
+@Slf4j
+public class VisualLayoutParsingAdapter {
+
+    private static String SIGNATURES = "signature";
+
+
+    public Map<Integer, List<VisualLayoutParsingResult>> buildExtractedTablesPerPage(VisualLayoutParsingResponse visualLayoutParsingResponse) {
+
+        Map<Integer, List<VisualLayoutParsingResult>> tableCells = new HashMap<>();
+        visualLayoutParsingResponse.getData()
+                .forEach(tableData -> tableCells.computeIfAbsent(tableData.getPage_idx(), tableCell -> new ArrayList<>()).addAll(convertTableCells(tableData.getBoxes())));
+
+        return tableCells;
+
+    }
+
+
+    public Map<Integer, List<ClassifiedImage>> buildExtractedSignaturesPerPage(VisualLayoutParsingResponse visualLayoutParsingResponse) {
+
+        Map<Integer, List<ClassifiedImage>> signatures = new HashMap<>();
+        if (visualLayoutParsingResponse.getData() != null) {
+            visualLayoutParsingResponse.getData()
+                    .forEach(tableData -> signatures.computeIfAbsent(tableData.getPage_idx() + 1, tableCell -> new ArrayList<>())
+                            .addAll(convertSignatures(tableData.getPage_idx()+1, tableData.getBoxes())));
+        }
+        return signatures;
+
+    }
+
+
+    public List<VisualLayoutParsingResult> convertTableCells(List<VisualLayoutParsingBox> tableObjects) {
+
+        List<VisualLayoutParsingResult> parsedTableCells = new ArrayList<>();
+
+        tableObjects.stream().forEach(t -> {
+            VisualLayoutParsingResult result = new VisualLayoutParsingResult();
+            result.setX0(t.getBox().getX1());
+            result.setX1(t.getBox().getX2());
+            result.setY0(t.getBox().getY1());
+            result.setY1(t.getBox().getY2());
+            result.setWidth(result.getX1() - result.getX0());
+            result.setHeight(result.getY1() - result.getY0());
+            result.setLabel(t.getLabel());
+            parsedTableCells.add(result);
+        });
+
+        return parsedTableCells;
+
+    }
+
+
+    public List<ClassifiedImage> convertSignatures(int pageNumber, List<VisualLayoutParsingBox> tableObjects) {
+
+        List<ClassifiedImage> signatures = new ArrayList<>();
+
+        tableObjects.stream().forEach(t -> {
+            if (t.getLabel().equals(SIGNATURES)) {
+                ClassifiedImage signature = new ClassifiedImage(new Rectangle2D.Float(t.getBox().getX1(),
+                        t.getBox().getY1(),
+                        t.getBox().getX2() - t.getBox().getX1(),
+                        t.getBox().getY2() - t.getBox().getY1()), ImageType.SIGNATURE, true, false, false, pageNumber);
+
+                signatures.add(signature);
+            }
+        });
+
+        return signatures;
+
+    }
+
+}

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

@@ -0,0 +1,18 @@
+package com.knecon.fforesight.service.layoutparser.processor.python_api.model.table;
+
+import lombok.AllArgsConstructor;
+import lombok.Builder;
+import lombok.Data;
+import lombok.NoArgsConstructor;
+
+@Data
+@Builder
+@NoArgsConstructor
+@AllArgsConstructor
+public class VisualLayoutParsingBox {
+
+    private VisualLayoutParsingBoxValue box;
+    private String label;
+    private float probability;
+
+}

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

@@ -0,0 +1,19 @@
+package com.knecon.fforesight.service.layoutparser.processor.python_api.model.table;
+
+import lombok.AllArgsConstructor;
+import lombok.Builder;
+import lombok.Data;
+import lombok.NoArgsConstructor;
+
+@Data
+@Builder
+@NoArgsConstructor
+@AllArgsConstructor
+public class VisualLayoutParsingBoxValue {
+
+    private float x1;
+    private float y1;
+    private float x2;
+    private float y2;
+
+}

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

@@ -0,0 +1,21 @@
+package com.knecon.fforesight.service.layoutparser.processor.python_api.model.table;
+
+import java.util.List;
+
+
+import lombok.AllArgsConstructor;
+import lombok.Builder;
+import lombok.Data;
+import lombok.NoArgsConstructor;
+
+@Data
+@Builder
+@NoArgsConstructor
+@AllArgsConstructor
+public class VisualLayoutParsingData {
+
+    private int page_idx;
+
+    private List<VisualLayoutParsingBox> boxes;
+
+}

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

@@ -0,0 +1,23 @@
+package com.knecon.fforesight.service.layoutparser.processor.python_api.model.table;
+
+import java.util.List;
+
+import lombok.AllArgsConstructor;
+import lombok.Builder;
+import lombok.Data;
+import lombok.NoArgsConstructor;
+
+@Data
+@Builder
+@NoArgsConstructor
+@AllArgsConstructor
+public class VisualLayoutParsingResponse {
+
+    private String dossierId;
+    private String fileId;
+    private String targetFileExtension;
+    private String responseFileExtension;
+    private String X_TENANT_ID;
+    private List<VisualLayoutParsingData> data;
+
+}

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

@@ -0,0 +1,22 @@
+package com.knecon.fforesight.service.layoutparser.processor.python_api.model.table;
+
+import lombok.AllArgsConstructor;
+import lombok.Builder;
+import lombok.Data;
+import lombok.NoArgsConstructor;
+
+@Data
+@Builder
+@NoArgsConstructor
+@AllArgsConstructor
+public class VisualLayoutParsingResult {
+
+    private float x0;
+    private float y0;
+    private float x1;
+    private float y1;
+    private float width;
+    private float height;
+    private String label;
+
+}

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

@@ -25,6 +25,7 @@ public class BodyTextFrameService {
     private static final float RULING_HEIGHT_THRESHOLD = 0.15f; // multiplied with page height. Header/Footer Rulings must be within that border of the page.
     private static final float RULING_WIDTH_THRESHOLD = 0.75f; // multiplied with page width. Header/Footer Rulings must be at least that wide.
 
+
     public void setBodyTextFrames(ClassificationDocument classificationDocument, LayoutParsingType layoutParsingType) {
 
         Rectangle bodyTextFrame = calculateBodyTextFrame(classificationDocument.getPages(), classificationDocument.getFontSizeCounter(), false, layoutParsingType);
@@ -132,12 +133,7 @@ public class BodyTextFrameService {
                                                boolean landscape,
                                                LayoutParsingType layoutParsingType) {
 
-        float approximateHeaderLineCount;
-        if (layoutParsingType.equals(LayoutParsingType.TAAS)) {
-            approximateHeaderLineCount = 3.3f;
-        } else {
-            approximateHeaderLineCount = 2.9f;
-        }
+        float approximateHeaderLineCount = 2.9f;
 
         BodyTextFrameExpansionsRectangle expansionsRectangle = new BodyTextFrameExpansionsRectangle();
 
@@ -155,8 +151,9 @@ public class BodyTextFrameService {
                         continue;
                     }
 
-                    if (MarkedContentUtils.intersects(textBlock, page.getMarkedContentBboxPerType(), MarkedContentUtils.HEADER)
-                            || MarkedContentUtils.intersects(textBlock, page.getMarkedContentBboxPerType(), MarkedContentUtils.FOOTER)) {
+                    if (MarkedContentUtils.intersects(textBlock, page.getMarkedContentBboxPerType(), MarkedContentUtils.HEADER) || MarkedContentUtils.intersects(textBlock,
+                            page.getMarkedContentBboxPerType(),
+                            MarkedContentUtils.FOOTER)) {
                         continue;
                     }
 

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

@@ -1,21 +1,21 @@
 package com.knecon.fforesight.service.layoutparser.processor.services;
 
-import java.awt.geom.Line2D;
+import static com.knecon.fforesight.service.layoutparser.processor.utils.GeometricComparators.X_FIRST_RULING_COMPARATOR;
+
 import java.awt.geom.Point2D;
 import java.util.ArrayList;
-import java.util.Collection;
-import java.util.Collections;
-import java.util.Comparator;
 import java.util.HashMap;
 import java.util.List;
 import java.util.Map;
+import java.util.stream.Collectors;
 
 import org.springframework.stereotype.Service;
 
 import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRulings;
+import com.knecon.fforesight.service.layoutparser.processor.model.table.Rectangle;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.Ruling;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.TableCells;
-import com.knecon.fforesight.service.layoutparser.processor.utils.DoubleComparisons;
+import com.knecon.fforesight.service.layoutparser.processor.utils.UnionFind;
 
 import lombok.RequiredArgsConstructor;
 import lombok.extern.slf4j.Slf4j;
@@ -25,25 +25,145 @@ import lombok.extern.slf4j.Slf4j;
 @RequiredArgsConstructor
 public class RulingCleaningService {
 
-    private static final float THRESHOLD = 6;
+    private static final float THRESHOLD_X_VERTICAL = 1;
+    private static final float THRESHOLD_Y_VERTICAL = 2;
+    private static final float THRESHOLD_X_HORIZONTAL = 2;
+    private static final float THRESHOLD_Y_HORIZONTAL = 3;
 
 
     public CleanRulings getCleanRulings(List<TableCells> tableCells, List<Ruling> rulings) {
 
+        Rulings verticalAndHorizontalRulingLines;
+
         if (!rulings.isEmpty()) {
-            snapPoints(rulings);
+            verticalAndHorizontalRulingLines = extractVerticalAndHorizontalRulingLines(rulings);
+        } else {
+            verticalAndHorizontalRulingLines = getRulingsFromParsedCells(tableCells);
         }
 
+        verticalAndHorizontalRulingLines.verticalLines.sort(X_FIRST_RULING_COMPARATOR);
+        verticalAndHorizontalRulingLines.horizontalLines.sort(X_FIRST_RULING_COMPARATOR);
+        verticalAndHorizontalRulingLines = cleanRulings(verticalAndHorizontalRulingLines);
+
+        return CleanRulings.builder().vertical(verticalAndHorizontalRulingLines.verticalLines()).horizontal(verticalAndHorizontalRulingLines.horizontalLines()).build();
+    }
+
+
+    private Rulings cleanRulings(Rulings rulings) {
+
+        List<List<Rectangle>> groupedOverlappingVerticalRectangles = groupOverlappingRectangles(rulings.verticalLines.stream()
+                                                                                                        .map(RulingCleaningService::getOverlapRectangle)
+                                                                                                        .distinct()
+                                                                                                        .toList());
+        List<Ruling> cleanedVerticalRulings = groupedOverlappingVerticalRectangles.stream()
+                .map(rectList -> getXCenteredRuling(Rectangle.boundingBoxOf(rectList)))
+                .toList();
+
+        List<List<Rectangle>> groupedOverlappingHorizontalRectangles = groupOverlappingRectangles(rulings.horizontalLines.stream()
+                                                                                                          .map(RulingCleaningService::getOverlapRectangle)
+                                                                                                          .distinct()
+                                                                                                          .toList());
+
+        List<Ruling> cleanedHorizontalRulings = groupedOverlappingHorizontalRectangles.stream()
+                .map(rectList -> getYCenteredRuling(Rectangle.boundingBoxOf(rectList)))
+                .collect(Collectors.toList());
+
+        return new Rulings(cleanedVerticalRulings, cleanedHorizontalRulings);
+    }
+
+
+    private List<List<Rectangle>> groupOverlappingRectangles(List<Rectangle> rectangles) {
+
+        UnionFind<Rectangle> unionFind = new UnionFind<>();
+        for (int i = 0; i < rectangles.size(); i++) {
+            for (int j = i + 1; j < rectangles.size(); j++) {
+                Rectangle rectangle1 = rectangles.get(i);
+                Rectangle rectangle2 = rectangles.get(j);
+
+                // we can stop early when we are too far off because of x-y-sorting
+                if(rectangle1.getRight() < rectangle2.getLeft() && rectangle1.getBottom() < rectangle2.getTop()) {
+                    break;
+                }
+
+                if (rectangle1.intersects(rectangle2)) {
+                    unionFind.union(rectangle1, rectangle2);
+                }
+            }
+        }
+
+        Map<Rectangle, List<Rectangle>> groups = new HashMap<>();
+        for (Rectangle rectangle : rectangles) {
+            Rectangle root = unionFind.find(rectangle);
+            groups.computeIfAbsent(root, k -> new ArrayList<>()).add(rectangle);
+        }
+        return new ArrayList<>(groups.values());
+    }
+
+
+    private static Rectangle getOverlapRectangle(Ruling ruling) {
+
+        float top;
+        float left;
+        float w;
+        float h;
+
+        if (ruling.x1 < ruling.x2) {
+            left = ruling.x1;
+            w = ruling.x2 - ruling.x1;
+        } else {
+            left = ruling.x2;
+            w = ruling.x1 - ruling.x2;
+        }
+        if (ruling.y1 < ruling.y2) {
+            top = ruling.y1;
+            h = ruling.y2 - ruling.y1;
+        } else {
+            top = ruling.y2;
+            h = ruling.y1 - ruling.y2;
+        }
+
+        if (ruling.horizontal()) {
+            return new Rectangle(top - THRESHOLD_Y_HORIZONTAL, left - THRESHOLD_X_HORIZONTAL, w + 2 * THRESHOLD_X_HORIZONTAL, h + 2 * THRESHOLD_Y_HORIZONTAL);
+        } else {
+            return new Rectangle(top - THRESHOLD_Y_VERTICAL, left - THRESHOLD_X_VERTICAL, w + 2 * THRESHOLD_X_VERTICAL, h + 2 * THRESHOLD_Y_VERTICAL);
+        }
+    }
+
+
+    public static Ruling getXCenteredRuling(Rectangle rectangle) {
+
+        float x = (float) rectangle.getCenterX();
+        float y1 = rectangle.getTop();
+        float y2 = rectangle.getBottom();
+
+        Point2D point1 = new Point2D.Float(x, y1 + THRESHOLD_Y_VERTICAL);
+        Point2D point2 = new Point2D.Float(x, y2 - THRESHOLD_Y_VERTICAL);
+
+        return new Ruling(point1, point2);
+    }
+
+
+    public static Ruling getYCenteredRuling(Rectangle rectangle) {
+
+        float x1 = rectangle.getLeft();
+        float x2 = rectangle.getRight();
+        float y = (float) rectangle.getCenterY();
+
+        Point2D point1 = new Point2D.Float(x1 + THRESHOLD_X_HORIZONTAL, y);
+        Point2D point2 = new Point2D.Float(x2 - THRESHOLD_X_HORIZONTAL, y);
+
+        return new Ruling(point1, point2);
+    }
+
+
+    private Rulings extractVerticalAndHorizontalRulingLines(List<Ruling> rulings) {
+
         List<Ruling> vrs = new ArrayList<>();
         for (Ruling vr : rulings) {
             if (vr.vertical()) {
                 vrs.add(vr);
             }
         }
-        if (vrs.isEmpty()) {
-            vrs.addAll(extractVerticalRulings(tableCells));
-        }
-        List<Ruling> verticalRulingLines = collapseOrientedRulings(vrs);
 
         List<Ruling> hrs = new ArrayList<>();
         for (Ruling hr : rulings) {
@@ -51,98 +171,26 @@ public class RulingCleaningService {
                 hrs.add(hr);
             }
         }
-        if (hrs.isEmpty()) {
-            hrs.addAll(extractHorizontalRulings(tableCells));
-        }
-        List<Ruling> horizontalRulingLines = collapseOrientedRulings(hrs);
-
-        return CleanRulings.builder().vertical(verticalRulingLines).horizontal(horizontalRulingLines).build();
+        return new Rulings(vrs, hrs);
     }
 
 
-    public void snapPoints(List<? extends Line2D.Float> rulings) {
+    private Rulings getRulingsFromParsedCells(List<TableCells> tableCells) {
 
-        // collect points and keep a Line -> p1,p2 map
-        Map<Line2D.Float, Point2D[]> linesToPoints = new HashMap<>();
-        List<Point2D> points = new ArrayList<>();
-        for (Line2D.Float r : rulings) {
-            Point2D p1 = r.getP1();
-            Point2D p2 = r.getP2();
-            linesToPoints.put(r, new Point2D[]{p1, p2});
-            points.add(p1);
-            points.add(p2);
-        }
-
-        // snap by X
-        points.sort(Comparator.comparingDouble(Point2D::getX));
-
-        List<List<Point2D>> groupedPoints = new ArrayList<>();
-        groupedPoints.add(new ArrayList<>(Collections.singletonList(points.get(0))));
-
-        for (Point2D p : points.subList(1, points.size() - 1)) {
-            List<Point2D> last = groupedPoints.get(groupedPoints.size() - 1);
-            if (Math.abs(p.getX() - last.get(0).getX()) < THRESHOLD) {
-                groupedPoints.get(groupedPoints.size() - 1).add(p);
-            } else {
-                groupedPoints.add(new ArrayList<>(Collections.singletonList(p)));
-            }
-        }
-
-        for (List<Point2D> group : groupedPoints) {
-            float avgLoc = 0;
-            for (Point2D p : group) {
-                avgLoc += p.getX();
-            }
-            avgLoc /= group.size();
-            for (Point2D p : group) {
-                p.setLocation(avgLoc, p.getY());
-            }
-        }
-        // ---
-
-        // snap by Y
-        points.sort(Comparator.comparingDouble(Point2D::getY));
-
-        groupedPoints = new ArrayList<>();
-        groupedPoints.add(new ArrayList<>(Collections.singletonList(points.get(0))));
-
-        for (Point2D p : points.subList(1, points.size() - 1)) {
-            List<Point2D> last = groupedPoints.get(groupedPoints.size() - 1);
-            if (Math.abs(p.getY() - last.get(0).getY()) < THRESHOLD) {
-                groupedPoints.get(groupedPoints.size() - 1).add(p);
-            } else {
-                groupedPoints.add(new ArrayList<>(Collections.singletonList(p)));
-            }
-        }
-
-        for (List<Point2D> group : groupedPoints) {
-            float avgLoc = 0;
-            for (Point2D p : group) {
-                avgLoc += p.getY();
-            }
-            avgLoc /= group.size();
-            for (Point2D p : group) {
-                p.setLocation(p.getX(), avgLoc);
-            }
-        }
-        // ---
-
-        // finally, modify lines
-        for (Map.Entry<Line2D.Float, Point2D[]> ltp : linesToPoints.entrySet()) {
-            Point2D[] p = ltp.getValue();
-            ltp.getKey().setLine(p[0], p[1]);
-        }
+        List<Ruling> vrs = extractVerticalRulingsFromParsedCells(tableCells);
+        List<Ruling> hrs = extractHorizontalRulingsFromParsedCells(tableCells);
+        return new Rulings(vrs, hrs);
     }
 
 
-    private Collection<? extends Ruling> extractVerticalRulings(List<TableCells> cvParsedTableCells) {
+    private List<Ruling> extractVerticalRulingsFromParsedCells(List<TableCells> tableCells) {
 
         List<Ruling> vrs = new ArrayList<>();
 
-        if (cvParsedTableCells != null) {
-            for (TableCells cvParsedTableCell : cvParsedTableCells) {
-                Ruling leftLine = createRuling(cvParsedTableCell.getX0(), cvParsedTableCell.getX0(), cvParsedTableCell.getY0(), cvParsedTableCell.getY1());
-                Ruling rightLine = createRuling(cvParsedTableCell.getX1(), cvParsedTableCell.getX1(), cvParsedTableCell.getY0(), cvParsedTableCell.getY1());
+        if (tableCells != null) {
+            for (TableCells tableCell : tableCells) {
+                Ruling leftLine = createRuling(tableCell.getX0(), tableCell.getX0(), tableCell.getY0(), tableCell.getY1());
+                Ruling rightLine = createRuling(tableCell.getX1(), tableCell.getX1(), tableCell.getY0(), tableCell.getY1());
                 vrs.add(leftLine);
                 vrs.add(rightLine);
             }
@@ -151,19 +199,18 @@ public class RulingCleaningService {
     }
 
 
-    private Collection<? extends Ruling> extractHorizontalRulings(List<TableCells> cvParsedTableCells) {
+    private List<Ruling> extractHorizontalRulingsFromParsedCells(List<TableCells> tableCells) {
 
         List<Ruling> hrs = new ArrayList<>();
 
-        if (cvParsedTableCells != null) {
-            for (TableCells cvParsedTableCell : cvParsedTableCells) {
-                Ruling topLine = createRuling(cvParsedTableCell.getX0(), cvParsedTableCell.getX1(), cvParsedTableCell.getY1(), cvParsedTableCell.getY1());
-                Ruling baseLine = createRuling(cvParsedTableCell.getX0(), cvParsedTableCell.getX1(), cvParsedTableCell.getY0(), cvParsedTableCell.getY0());
+        if (tableCells != null) {
+            for (TableCells tableCell : tableCells) {
+                Ruling topLine = createRuling(tableCell.getX0(), tableCell.getX1(), tableCell.getY1(), tableCell.getY1());
+                Ruling baseLine = createRuling(tableCell.getX0(), tableCell.getX1(), tableCell.getY0(), tableCell.getY0());
                 hrs.add(topLine);
                 hrs.add(baseLine);
             }
         }
-
         return hrs;
     }
 
@@ -189,46 +236,8 @@ public class RulingCleaningService {
     }
 
 
-    private List<Ruling> collapseOrientedRulings(List<Ruling> lines) {
+    private record Rulings(List<Ruling> verticalLines, List<Ruling> horizontalLines) {
 
-        int COLINEAR_OR_PARALLEL_PIXEL_EXPAND_AMOUNT = 1;
-        return collapseOrientedRulings(lines, COLINEAR_OR_PARALLEL_PIXEL_EXPAND_AMOUNT);
-    }
-
-
-    private List<Ruling> collapseOrientedRulings(List<Ruling> lines, int expandAmount) {
-
-        ArrayList<Ruling> rv = new ArrayList<>();
-        lines.sort((a, b) -> {
-            final float diff = a.getPosition() - b.getPosition();
-            return Float.compare(diff == 0 ? a.getStart() - b.getStart() : diff, 0f);
-        });
-
-        for (Ruling next_line : lines) {
-            Ruling last = rv.isEmpty() ? null : rv.get(rv.size() - 1);
-            // if current line colinear with next, and are "close enough": expand current line
-            if (last != null && DoubleComparisons.feq(next_line.getPosition(), last.getPosition()) && last.nearlyIntersects(next_line, expandAmount)) {
-                final float lastStart = last.getStart();
-                final float lastEnd = last.getEnd();
-
-                final boolean lastFlipped = lastStart > lastEnd;
-                final boolean nextFlipped = next_line.getStart() > next_line.getEnd();
-
-                boolean differentDirections = nextFlipped != lastFlipped;
-                float nextS = differentDirections ? next_line.getEnd() : next_line.getStart();
-                float nextE = differentDirections ? next_line.getStart() : next_line.getEnd();
-
-                final float newStart = lastFlipped ? Math.max(nextS, lastStart) : Math.min(nextS, lastStart);
-                final float newEnd = lastFlipped ? Math.min(nextE, lastEnd) : Math.max(nextE, lastEnd);
-                last.setStartEnd(newStart, newEnd);
-                assert !last.oblique();
-            } else if (next_line.length() == 0) {
-                continue;
-            } else {
-                rv.add(next_line);
-            }
-        }
-        return rv;
     }
 
 }

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

@@ -7,6 +7,7 @@ import java.util.List;
 import java.util.Map;
 import java.util.stream.Collectors;
 
+import org.apache.logging.log4j.util.Strings;
 import org.springframework.stereotype.Service;
 
 import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
@@ -43,7 +44,6 @@ public class SectionsBuilderService {
         for (ClassificationPage page : document.getPages()) {
             List<TextPageBlock> header = new ArrayList<>();
             List<TextPageBlock> footer = new ArrayList<>();
-            List<TextPageBlock> unclassifiedText = new ArrayList<>();
             for (AbstractPageBlock current : page.getTextBlocks()) {
 
                 if (current.getClassification() == null) {
@@ -62,11 +62,6 @@ public class SectionsBuilderService {
                     continue;
                 }
 
-                if (current.getClassification().equals(PageBlockType.OTHER)) {
-                    unclassifiedText.add((TextPageBlock) current);
-                    continue;
-                }
-
                 if (prev != null && current.getClassification().isHeadline() && !prev.getClassification().isHeadline() || !document.isHeadlines()) {
                     ClassificationSection chunkBlock = buildTextBlock(chunkWords, lastHeadline);
                     chunkBlock.setHeadline(lastHeadline);
@@ -94,9 +89,6 @@ public class SectionsBuilderService {
             if (!footer.isEmpty()) {
                 footers.add(new ClassificationFooter(footer));
             }
-            if (!unclassifiedText.isEmpty()) {
-                unclassifiedTexts.add(new UnclassifiedText(unclassifiedText));
-            }
         }
 
         ClassificationSection chunkBlock = buildTextBlock(chunkWords, lastHeadline);
@@ -110,6 +102,20 @@ public class SectionsBuilderService {
     }
 
 
+    public void buildParagraphDebugSections(ClassificationDocument document) {
+
+        List<ClassificationSection> sections = new ArrayList<>();
+        for (var page : document.getPages()) {
+            page.getTextBlocks().forEach(block -> {
+                block.setPage(page.getPageNumber());
+                var section = buildTextBlock(List.of(block), Strings.EMPTY);
+                sections.add(section);
+            });
+        }
+        document.setSections(sections);
+    }
+
+
     public void addImagesToSections(ClassificationDocument document) {
 
         Map<Integer, List<ClassificationSection>> sectionMap = new HashMap<>();
@@ -200,6 +206,7 @@ public class SectionsBuilderService {
                             .getX() <= xMax && image.getPosition().getY() >= yMin && image.getPosition().getY() <= yMax) {
                         section.getImages().add(image);
                         image.setAppendedToSection(true);
+                        break;
                     }
                 }
                 if (!image.isAppendedToSection()) {

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

@@ -1,26 +1,51 @@
 package com.knecon.fforesight.service.layoutparser.processor.services;
 
 import java.util.List;
+import java.util.stream.Collectors;
+import java.util.stream.Stream;
 
 import org.springframework.stereotype.Service;
 
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.SimplifiedSectionText;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Section;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.SimplifiedText;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.GenericSemanticNode;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.SemanticNode;
+
+import lombok.extern.slf4j.Slf4j;
 
 @Service
+@Slf4j
 public class SimplifiedSectionTextService {
 
     public SimplifiedText toSimplifiedText(Document document) {
 
-        List<SimplifiedSectionText> simplifiedSectionTexts = document.getMainSections().stream().map(this::toSimplifiedSectionText).toList();
-        return SimplifiedText.builder().numberOfPages(document.getNumberOfPages()).sectionTexts(simplifiedSectionTexts).build();
+        List<SimplifiedSectionText> simplifiedMainSectionsList = document.getMainSections()
+                .stream()
+                .map(this::toSimplifiedSectionText)
+                .toList();
+        List<SimplifiedSectionText> simplifiedHeadersList = document.getHeaders()
+                .stream()
+                .map(this::toSimplifiedSectionText)
+                .toList();
+        List<SimplifiedSectionText> simplifiedFootersList = document.getFooters()
+                .stream()
+                .map(this::toSimplifiedSectionText)
+                .toList();
+        List<SimplifiedSectionText> simplifiedText = Stream.of(simplifiedMainSectionsList, simplifiedHeadersList, simplifiedFootersList)
+                .flatMap(List::stream)
+                .collect(Collectors.toList());
+        return SimplifiedText.builder().numberOfPages(document.getNumberOfPages()).sectionTexts(simplifiedText).build();
     }
 
 
-    private SimplifiedSectionText toSimplifiedSectionText(Section section) {
+    private SimplifiedSectionText toSimplifiedSectionText(SemanticNode section) {
 
-        return SimplifiedSectionText.builder().sectionNumber(section.getTreeId().get(0)).text(section.getTextBlock().getSearchText()).build();
+        return SimplifiedSectionText.builder()
+                .sectionNumber(section.getTreeId()
+                                       .get(0))
+                .text(section.getTextBlock().getSearchText())
+                .build();
     }
+
 }

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

@@ -1,79 +1,33 @@
 package com.knecon.fforesight.service.layoutparser.processor.services;
 
-import java.awt.geom.Point2D;
+import static com.knecon.fforesight.service.layoutparser.processor.utils.GeometricComparators.CELL_SIZE_COMPARATOR;
+import static com.knecon.fforesight.service.layoutparser.processor.utils.GeometricComparators.RECTANGLE_SIZE_COMPARATOR;
+
 import java.util.ArrayList;
-import java.util.Comparator;
-import java.util.HashMap;
 import java.util.HashSet;
-import java.util.Iterator;
 import java.util.List;
 import java.util.Map;
-import java.util.Set;
+import java.util.stream.Collectors;
 
 import org.springframework.stereotype.Service;
 
 import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
 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.table.Rectangle;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.Ruling;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.TablePageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.utils.DoubleComparisons;
+import com.knecon.fforesight.service.layoutparser.processor.utils.RectangularIntersectionFinder;
+import com.knecon.fforesight.service.layoutparser.processor.utils.SpreadsheetFinder;
 
 @Service
 public class TableExtractionService {
 
-    private static final Comparator<Point2D> X_FIRST_POINT_COMPARATOR = (arg0, arg1) -> {
-
-        int rv = 0;
-        float arg0X = DoubleComparisons.round(arg0.getX(), 2);
-        float arg0Y = DoubleComparisons.round(arg0.getY(), 2);
-        float arg1X = DoubleComparisons.round(arg1.getX(), 2);
-        float arg1Y = DoubleComparisons.round(arg1.getY(), 2);
-
-        if (arg0X > arg1X) {
-            rv = 1;
-        } else if (arg0X < arg1X) {
-            rv = -1;
-        } else if (arg0Y > arg1Y) {
-            rv = 1;
-        } else if (arg0Y < arg1Y) {
-            rv = -1;
-        }
-        return rv;
-    };
-    private static final Comparator<Point2D> POINT_COMPARATOR = (arg0, arg1) -> {
-
-        int rv = 0;
-        float arg0X = DoubleComparisons.round(arg0.getX(), 2);
-        float arg0Y = DoubleComparisons.round(arg0.getY(), 2);
-        float arg1X = DoubleComparisons.round(arg1.getX(), 2);
-        float arg1Y = DoubleComparisons.round(arg1.getY(), 2);
-
-        if (arg0Y > arg1Y) {
-            rv = 1;
-        } else if (arg0Y < arg1Y) {
-            rv = -1;
-        } else if (arg0X > arg1X) {
-            rv = 1;
-        } else if (arg0X < arg1X) {
-            rv = -1;
-        }
-        return rv;
-    };
-
-
-    public boolean contains(Cell cell, double x, double y, double w, double h) {
-
-        if (cell.isEmpty() || w <= 0 || h <= 0) {
-            return false;
-        }
-        double x0 = cell.getX();
-        double y0 = cell.getY();
-        return (x >= x0 - 2 && y >= y0 - 2 && (x + w) <= x0 + cell.getWidth() + 2 && (y + h) <= y0 + cell.getHeight() + 2);
-    }
+    private static final int MAX_TABLE_CONTAINED_CELLS_WITH_TEXT = 1;
+    private static final int TEXT_BLOCK_CONTAINMENT_TOLERANCE = 2;
+    private static final double TABLE_UNIFORMITY_THRESHOLD = 0.7;
 
 
     /**
@@ -86,272 +40,104 @@ public class TableExtractionService {
      * <p>
      * DirAdj (Text direction adjusted) values can not be used here.
      *
-     * @param cleanRulings The lines used to build the table.
-     * @param page         Page object that contains textblocks and statistics.
+     * @param emptyCells The cells used to build the table.
+     * @param page       Page object that contains textblocks and statistics.
      */
-    public void extractTables(CleanRulings cleanRulings, ClassificationPage page) {
 
-        List<Cell> cells = findCells(cleanRulings.getHorizontal(), cleanRulings.getVertical());
+    public void extractTables(List<Cell> emptyCells, ClassificationPage page) {
 
-        List<TextPageBlock> toBeRemoved = new ArrayList<>();
+        // sort cells by size (height * width) ascending so that textBlocks are always assigned to the smallest cells that contain them
+        emptyCells.sort(CELL_SIZE_COMPARATOR);
 
         for (AbstractPageBlock abstractPageBlock : page.getTextBlocks()) {
             TextPageBlock textBlock = (TextPageBlock) abstractPageBlock;
-            for (Cell cell : cells) {
-                if (cell.hasMinimumSize() && contains(cell,
-                        textBlock.getPdfMinX(),
-                        textBlock.getPdfMinY(),
-                        textBlock.getPdfMaxX() - textBlock.getPdfMinX(),
-                        textBlock.getPdfMaxY() - textBlock.getPdfMinY())) {
+            for (Cell cell : emptyCells) {
+                if (cell.hasMinimumSize() && doesCellContainTextBlock(cell, textBlock)) {
                     cell.addTextBlock(textBlock);
-                    toBeRemoved.add(textBlock);
                     break;
                 }
             }
         }
 
-        cells = new ArrayList<>(new HashSet<>(cells));
+        var cells = new ArrayList<>(new HashSet<>(emptyCells));
         DoubleComparisons.sort(cells, Rectangle.ILL_DEFINED_ORDER);
 
-        List<Rectangle> spreadsheetAreas = findSpreadsheetsFromCells(cells);
+        List<Rectangle> spreadsheetAreas = SpreadsheetFinder.findSpreadsheetsFromCells(cells);
+        // sort spreadsheetAreas by size (height * width) ascending so that cells are placed in the smallest tables first
+        // this way no cell duplication occurs when tables are contained in other tables and only the most inner table contains the cells
+        spreadsheetAreas.sort(RECTANGLE_SIZE_COMPARATOR);
 
         List<TablePageBlock> tables = new ArrayList<>();
         for (Rectangle area : spreadsheetAreas) {
 
-            List<Cell> overlappingCells = new ArrayList<>();
+            List<Cell> containedCells = new ArrayList<>();
             for (Cell c : cells) {
-                if (c.hasMinimumSize() && c.intersects(area)) {
-                    overlappingCells.add(c);
+                if (c.hasMinimumSize() && area.contains(c)) {
+                    containedCells.add(c);
                 }
             }
-            tables.add(new TablePageBlock(overlappingCells, area, page.getRotation()));
+
+            var containedCellsWithText = containedCells.stream().filter(cell -> !cell.getTextBlocks().isEmpty()).toList();
+
+            // verify if table would contain fewer cells with text than the threshold allows
+            if (containedCellsWithText.size() >= MAX_TABLE_CONTAINED_CELLS_WITH_TEXT && checkIfTableCellsAreUniform(containedCells)) {
+                tables.add(new TablePageBlock(containedCells, area, page.getRotation()));
+                cells.removeAll(containedCells);
+            }
         }
 
         for (TablePageBlock table : tables) {
             int position = -1;
 
-            Iterator<AbstractPageBlock> itty = page.getTextBlocks().iterator();
-            while (itty.hasNext()) {
-                AbstractPageBlock textBlock = itty.next();
-                if (textBlock instanceof TextPageBlock ? table.containsBlock((TextPageBlock) textBlock) : table.contains(textBlock) && position == -1) {
-                    position = page.getTextBlocks().indexOf(textBlock);
+            for (AbstractPageBlock pageBlock : page.getTextBlocks()) {
+                if (pageBlock instanceof TextPageBlock ? table.containsBlock((TextPageBlock) pageBlock) : table.contains(pageBlock) && position == -1) {
+                    position = page.getTextBlocks().indexOf(pageBlock);
                 }
             }
             if (position != -1) {
                 page.getTextBlocks().add(position, table);
+
+                var toBeRemoved = table.getCells().stream().map(Cell::getTextBlocks).flatMap(List::stream).toList();
+                // remove text blocks from the page that were also added with the table (from its contained cells)
+                page.getTextBlocks().removeAll(toBeRemoved);
             }
         }
-
-        page.getTextBlocks().removeAll(toBeRemoved);
     }
 
 
-    public List<Cell> findCells(List<Ruling> horizontalRulingLines, List<Ruling> verticalRulingLines) {
+    private boolean checkIfTableCellsAreUniform(List<Cell> containedCells) {
 
-        // Fix for 211.pdf
-        for (Ruling r : horizontalRulingLines) {
-            if (r.getX2() < r.getX1()) {
-                double a = r.getX2();
-                r.x2 = (float) r.getX1();
-                r.x1 = (float) a;
-            }
+        if (containedCells.size() <= 2) {
+            return true;
         }
 
-        List<Cell> cellsFound = new ArrayList<>();
-        Map<Point2D, Ruling[]> intersectionPoints = Ruling.findIntersections(horizontalRulingLines, verticalRulingLines);
-        List<Point2D> intersectionPointsList = new ArrayList<>(intersectionPoints.keySet());
-        intersectionPointsList.sort(POINT_COMPARATOR);
+        Map<Long, List<Long>> cellsGroupedByRoundedWidth = containedCells.stream()
+                .map(Rectangle::getWidth)
+                .map(size -> Math.round(size / 10.0) * 10)
+                .collect(Collectors.groupingBy(Long::longValue));
 
-        for (int i = 0; i < intersectionPointsList.size(); i++) {
-            Point2D topLeft = intersectionPointsList.get(i);
-            Ruling[] hv = intersectionPoints.get(topLeft);
-
-            // CrossingPointsDirectlyBelow( topLeft );
-            List<Point2D> xPoints = new ArrayList<>();
-            // CrossingPointsDirectlyToTheRight( topLeft );
-            List<Point2D> yPoints = new ArrayList<>();
-
-            for (Point2D p : intersectionPointsList.subList(i, intersectionPointsList.size())) {
-                if (p.getX() == topLeft.getX() && p.getY() > topLeft.getY()) {
-                    xPoints.add(p);
-                }
-                if (p.getY() == topLeft.getY() && p.getX() > topLeft.getX()) {
-                    yPoints.add(p);
-                }
-            }
-            outer:
-            for (Point2D xPoint : xPoints) {
-                // is there a vertical edge b/w topLeft and xPoint?
-                if (!hv[1].equals(intersectionPoints.get(xPoint)[1])) {
-                    continue;
-                }
-                for (Point2D yPoint : yPoints) {
-                    // is there an horizontal edge b/w topLeft and yPoint ?
-                    if (!hv[0].equals(intersectionPoints.get(yPoint)[0])) {
-                        continue;
-                    }
-                    Point2D btmRight = new Point2D.Float((float) yPoint.getX(), (float) xPoint.getY());
-                    if (intersectionPoints.containsKey(btmRight) && intersectionPoints.get(btmRight)[0].equals(intersectionPoints.get(xPoint)[0]) && intersectionPoints.get(btmRight)[1].equals(
-                            intersectionPoints.get(yPoint)[1])) {
-                        cellsFound.add(new Cell(topLeft, btmRight));
-                        break outer;
-                    }
-                }
-            }
-        }
-
-        // TODO create cells for vertical ruling lines with aligned endpoints at the top/bottom of a grid
-        // that aren't connected with an horizontal ruler?
-        // see: https://github.com/jazzido/tabula-extractor/issues/78#issuecomment-41481207
-
-        return cellsFound;
+        return (double) cellsGroupedByRoundedWidth.size() / containedCells.size() <= TABLE_UNIFORMITY_THRESHOLD;
     }
 
 
-    private List<Rectangle> findSpreadsheetsFromCells(List<? extends Rectangle> cells) {
-        // via: http://stackoverflow.com/questions/13746284/merging-multiple-adjacent-rectangles-into-one-polygon
-        List<Rectangle> rectangles = new ArrayList<>();
-        Set<Point2D> pointSet = new HashSet<>();
-        Map<Point2D, Point2D> edgesH = new HashMap<>();
-        Map<Point2D, Point2D> edgesV = new HashMap<>();
-        int i = 0;
+    private boolean doesCellContainTextBlock(Cell cell, TextPageBlock textBlock) {
 
-        for (Rectangle cell : cells) {
-            for (Point2D pt : cell.getPoints()) {
-                if (pointSet.contains(pt)) { // shared vertex, remove it
-                    pointSet.remove(pt);
-                } else {
-                    pointSet.add(pt);
-                }
-            }
+        double x = textBlock.getPdfMinX();
+        double y = textBlock.getPdfMinY();
+        double w = textBlock.getPdfMaxX() - textBlock.getPdfMinX();
+        double h = textBlock.getPdfMaxY() - textBlock.getPdfMinY();
+        if (cell.isEmpty() || w <= 0 || h <= 0) {
+            return false;
         }
-
-        // X first sort
-        List<Point2D> pointsSortX = new ArrayList<>(pointSet);
-        pointsSortX.sort(X_FIRST_POINT_COMPARATOR);
-        // Y first sort
-        List<Point2D> pointsSortY = new ArrayList<>(pointSet);
-        pointsSortY.sort(POINT_COMPARATOR);
-
-        while (i < pointSet.size()) {
-            float currY = (float) pointsSortY.get(i).getY();
-            while (i < pointSet.size() && DoubleComparisons.feq(pointsSortY.get(i).getY(), currY)) {
-                edgesH.put(pointsSortY.get(i), pointsSortY.get(i + 1));
-                edgesH.put(pointsSortY.get(i + 1), pointsSortY.get(i));
-                i += 2;
-            }
-        }
-
-        i = 0;
-        while (i < pointSet.size()) {
-            float currX = (float) pointsSortX.get(i).getX();
-            while (i < pointSet.size() && DoubleComparisons.feq(pointsSortX.get(i).getX(), currX)) {
-                edgesV.put(pointsSortX.get(i), pointsSortX.get(i + 1));
-                edgesV.put(pointsSortX.get(i + 1), pointsSortX.get(i));
-                i += 2;
-            }
-        }
-
-        // Get all the polygons
-        List<List<PolygonVertex>> polygons = new ArrayList<>();
-        Point2D nextVertex;
-        while (!edgesH.isEmpty()) {
-            ArrayList<PolygonVertex> polygon = new ArrayList<>();
-            Point2D first = edgesH.keySet().iterator().next();
-            polygon.add(new PolygonVertex(first, Direction.HORIZONTAL));
-            edgesH.remove(first);
-
-            while (true) {
-                PolygonVertex curr = polygon.get(polygon.size() - 1);
-                PolygonVertex lastAddedVertex;
-                if (curr.direction == Direction.HORIZONTAL) {
-                    nextVertex = edgesV.get(curr.point);
-                    edgesV.remove(curr.point);
-                    lastAddedVertex = new PolygonVertex(nextVertex, Direction.VERTICAL);
-                } else {
-                    nextVertex = edgesH.get(curr.point);
-                    edgesH.remove(curr.point);
-                    lastAddedVertex = new PolygonVertex(nextVertex, Direction.HORIZONTAL);
-                }
-                polygon.add(lastAddedVertex);
-
-                if (lastAddedVertex.equals(polygon.get(0))) {
-                    // closed polygon
-                    polygon.remove(polygon.size() - 1);
-                    break;
-                }
-            }
-
-            for (PolygonVertex vertex : polygon) {
-                edgesH.remove(vertex.point);
-                edgesV.remove(vertex.point);
-            }
-            polygons.add(polygon);
-        }
-
-        // calculate grid-aligned minimum area rectangles for each found polygon
-        for (List<PolygonVertex> poly : polygons) {
-            float top = Float.MAX_VALUE;
-            float left = Float.MAX_VALUE;
-            float bottom = Float.MIN_VALUE;
-            float right = Float.MIN_VALUE;
-            for (PolygonVertex pt : poly) {
-                top = (float) Math.min(top, pt.point.getY());
-                left = (float) Math.min(left, pt.point.getX());
-                bottom = (float) Math.max(bottom, pt.point.getY());
-                right = (float) Math.max(right, pt.point.getX());
-            }
-            rectangles.add(new Rectangle(top, left, right - left, bottom - top));
-        }
-
-        return rectangles;
+        double x0 = cell.getX();
+        double y0 = cell.getY();
+        return (x >= x0 - TEXT_BLOCK_CONTAINMENT_TOLERANCE && y >= y0 - TEXT_BLOCK_CONTAINMENT_TOLERANCE && (x + w) <= x0 + cell.getWidth() + 2 * TEXT_BLOCK_CONTAINMENT_TOLERANCE && (y + h) <= y0 + cell.getHeight() + 2 * TEXT_BLOCK_CONTAINMENT_TOLERANCE);
     }
 
 
-    private enum Direction {
-        HORIZONTAL,
-        VERTICAL
-    }
-
-    static class PolygonVertex {
-
-        Point2D point;
-        Direction direction;
-
-
-        PolygonVertex(Point2D point, Direction direction) {
-
-            this.direction = direction;
-            this.point = point;
-        }
-
-
-        @Override
-        public boolean equals(Object other) {
-
-            if (this == other) {
-                return true;
-            }
-            if (!(other instanceof PolygonVertex)) {
-                return false;
-            }
-            return this.point.equals(((PolygonVertex) other).point);
-        }
-
-
-        @Override
-        public int hashCode() {
-
-            return this.point.hashCode();
-        }
-
-
-        @Override
-        public String toString() {
-
-            return String.format("%s[point=%s,direction=%s]", this.getClass().getName(), this.point.toString(), this.direction.toString());
-        }
+    public static List<Cell> findCells(List<Ruling> horizontalRulingLines, List<Ruling> verticalRulingLines) {
 
+        return RectangularIntersectionFinder.find(horizontalRulingLines, verticalRulingLines).stream().map(Cell::new).collect(Collectors.toList());
     }
 
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/blockification/DocstrumBlockificationService.java

@@ -0,0 +1,429 @@
+package com.knecon.fforesight.service.layoutparser.processor.services.blockification;
+
+import static java.util.stream.Collectors.toSet;
+
+import java.util.ArrayList;
+import java.util.Comparator;
+import java.util.List;
+import java.util.ListIterator;
+
+import org.springframework.stereotype.Service;
+
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.DocstrumSegmentationService;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Zone;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.utils.DoubleUtils;
+import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
+import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
+import com.knecon.fforesight.service.layoutparser.processor.model.FloatFrequencyCounter;
+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.table.Ruling;
+import com.knecon.fforesight.service.layoutparser.processor.model.table.TablePageBlock;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.StringFrequencyCounter;
+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.RectangleTransformations;
+import com.knecon.fforesight.service.layoutparser.processor.utils.RulingTextDirAdjustUtil;
+
+import lombok.RequiredArgsConstructor;
+
+@SuppressWarnings("all")
+@Service
+@RequiredArgsConstructor
+public class DocstrumBlockificationService {
+
+    private final DocstrumSegmentationService docstrumSegmentationService;
+
+    static final float THRESHOLD = 1f;
+
+
+    public ClassificationPage blockify(List<TextPositionSequence> textPositions, List<Cell> cells, boolean xyOrder) {
+
+        CleanRulings usedRulings = RectangleTransformations.extractRulings(cells);
+
+        var zones = docstrumSegmentationService.segmentPage(textPositions, xyOrder);
+        var pageBlocks = toAbstractPageBlocks(zones, usedRulings.getHorizontal(), usedRulings.getVertical(), xyOrder);
+
+        var classificationPage = new ClassificationPage(pageBlocks);
+
+        mergeIntersectingBlocks(classificationPage.getTextBlocks(), 0, 0);
+
+        return classificationPage;
+    }
+
+
+    private List<AbstractPageBlock> toAbstractPageBlocks(List<Zone> zones, List<Ruling> horizontalRulings, List<Ruling> verticalRulings, boolean xyOrder) {
+
+        List<AbstractPageBlock> abstractPageBlocks = new ArrayList<>();
+        zones.forEach(zone -> {
+
+            List<TextPositionSequence> textPositionSequences = new ArrayList<>();
+            zone.getLines().forEach(line -> {
+                line.getWords().forEach(word -> {
+                    textPositionSequences.add(new TextPositionSequence(word.getTextPositions(), word.getPage()));
+                });
+            });
+
+            abstractPageBlocks.addAll(splitZonesAtRulings(textPositionSequences, horizontalRulings, verticalRulings));
+        });
+
+        if (xyOrder) {
+            abstractPageBlocks.sort(Comparator.comparing(AbstractPageBlock::getMinY, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD))
+                    .thenComparing(AbstractPageBlock::getMinX, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD)));
+            abstractPageBlocks.sort(new Comparator<AbstractPageBlock>() {
+                @Override
+                public int compare(AbstractPageBlock o1, AbstractPageBlock o2) {
+
+                    return Math.abs(o1.getMinY() - o2.getMinY()) < 5 && o1.getMinX() < o2.getMinX() == true ? -1 : 0;
+                }
+            });
+        }
+
+        return abstractPageBlocks;
+    }
+
+
+    public void combineBlocks(ClassificationPage page) {
+
+        TextPageBlock previous = new TextPageBlock();
+        ListIterator<AbstractPageBlock> itty = page.getTextBlocks().listIterator();
+        while (itty.hasNext()) {
+
+            AbstractPageBlock block = itty.next();
+            if (block instanceof TablePageBlock) {
+                previous = new TextPageBlock();
+                continue;
+            }
+            TextPageBlock current = (TextPageBlock) block;
+
+            if (previous != null && !previous.getSequences().isEmpty()) {
+
+                if (current.getDir() != previous.getDir()) {
+                    previous = current;
+                    continue;
+                }
+
+                if (areTheOnlyTwoBlocksOnHeightsWithBothMoreThanTwoLines(previous, current, page)) {
+                    previous = combineBlocksAndResetIterator(previous, current, itty, true);
+                    continue;
+                }
+
+                if (previous.almostIntersects(current, 0, 0)) {
+                    previous = combineBlocksAndResetIterator(previous, current, itty, previous.isToDuplicate());
+                    continue;
+                }
+
+                if (isSameTopOrBottomWithPreviousHavingMaxTwoLinesAndCurrentThanOneAndMax4OtherBlocksOnHeight(previous, current, page)) {
+                    previous = combineBlocksAndResetIterator(previous, current, itty, previous.isToDuplicate());
+                    continue;
+                }
+
+                if (isOnlyIntersectingYAndOnelineOrPrevoiusTwoLines(previous, current, page)) {
+                    previous = combineBlocksAndResetIterator(previous, current, itty, previous.isToDuplicate());
+                    continue;
+                }
+
+            }
+            previous = current;
+        }
+
+        mergeIntersectingBlocks(page.getTextBlocks(), 0, 6.5f);
+    }
+
+
+    private boolean isOnlyIntersectingYAndOnelineOrPrevoiusTwoLines(TextPageBlock previous, TextPageBlock current, ClassificationPage page) {
+
+        return current.intersectsY(previous) //
+                && (previous.getNumberOfLines() == 1 && current.getNumberOfLines() >= 1 || previous.getNumberOfLines() == 2 && current.getNumberOfLines() == 1) //
+                && numberOfYIntersectionsOfSmallerBlocksWithOtherBlocks(current, previous, page.getTextBlocks()) <= 0;
+    }
+
+
+    private boolean isSameTopOrBottomWithPreviousHavingMaxTwoLinesAndCurrentThanOneAndMax4OtherBlocksOnHeight(TextPageBlock previous,
+                                                                                                              TextPageBlock current,
+                                                                                                              ClassificationPage page) {
+
+        return (Math.abs(previous.getMaxY() - current.getMaxY()) < THRESHOLD || Math.abs(previous.getMinY() - current.getMinY()) < THRESHOLD) //
+                && (previous.getNumberOfLines() == 1 && current.getNumberOfLines() >= 1 || previous.getNumberOfLines() >= 2 && current.getNumberOfLines() == 1) //
+                && !hasBetween(current, previous, page.getTextBlocks()) && numberOfYIntersectionsOfSmallerBlocksWithOtherBlocks(current, previous, page.getTextBlocks()) <= 4;
+    }
+
+
+    private boolean areTheOnlyTwoBlocksOnHeightsWithBothMoreThanTwoLines(TextPageBlock previous, TextPageBlock current, ClassificationPage page) {
+
+        return previous.getNumberOfLines() >= 2 && current.getNumberOfLines() >= 2 //
+                && previous.intersectsY(current) //
+                && numberOfYIntersectionsOfSmallerBlocksWithOtherBlocks(current, previous, page.getTextBlocks()) == 0;
+    }
+
+
+    private TextPageBlock combineBlocksAndResetIterator(TextPageBlock previous, TextPageBlock current, ListIterator<AbstractPageBlock> itty, boolean toDuplicate) {
+
+        previous.getSequences().addAll(current.getSequences());
+        previous = buildTextBlock(previous.getSequences(), 0);
+        previous.setToDuplicate(toDuplicate);
+        itty.remove();
+        itty.previous();
+        itty.set(previous);
+        itty.next();
+        return previous;
+    }
+
+
+    private boolean hasBetween(TextPageBlock block, TextPageBlock other, List<AbstractPageBlock> allBlocks) {
+
+        for (AbstractPageBlock current : allBlocks) {
+
+            if (current == other || current == block) {
+                continue;
+            }
+
+            if (other.intersectsY(current) && other.getMaxX() <= current.getMinX() && current.getMaxX() <= block.getMinX()) {
+                return true;
+            }
+        }
+
+        return false;
+    }
+
+
+    private int numberOfYIntersectionsOfSmallerBlocksWithOtherBlocks(TextPageBlock block, TextPageBlock other, List<AbstractPageBlock> allBlocks) {
+
+        double minY = Math.min(block.getMinY(), other.getMinY());
+        double maxY = Math.min(block.getMaxY(), other.getMaxY());
+
+        int numberOfYIntersections = 0;
+        for (AbstractPageBlock current : allBlocks) {
+
+            if (current == other || current == block) {
+                continue;
+            }
+
+            if (minY <= current.getMaxY() && maxY >= current.getMinY()) {
+                numberOfYIntersections++;
+            }
+        }
+
+        return numberOfYIntersections;
+    }
+
+
+    public void mergeIntersectingBlocks(List<AbstractPageBlock> blocks, float xThreshold, float yThreshold) {
+
+        ListIterator<AbstractPageBlock> itty = blocks.listIterator();
+        while (itty.hasNext()) {
+            AbstractPageBlock block = itty.next();
+            if(block == null){
+                continue;
+            }
+            if (block instanceof TablePageBlock) {
+                continue;
+            }
+
+            TextPageBlock current = (TextPageBlock) block;
+
+            for (int i = 0; i < blocks.size(); i++) {
+
+                if(blocks.get(i) == null){
+                    continue;
+                }
+                if (blocks.get(i) == current) {
+                    continue;
+                }
+                if (blocks.get(i) instanceof TablePageBlock) {
+                    continue;
+                }
+
+                TextPageBlock inner = (TextPageBlock) blocks.get(i);
+
+                if (current.getDir() == inner.getDir() && current.almostIntersects(inner, yThreshold, xThreshold)) {
+
+                    boolean toDuplicate = current.isToDuplicate() || inner.isToDuplicate();
+                    current.getSequences().addAll(inner.getSequences());
+                    current = buildTextBlock(current.getSequences(), 0);
+
+                    current.setToDuplicate(toDuplicate);
+                    blocks.set(i, null);
+                    itty.set(current);
+                }
+            }
+        }
+        var blocksIterator = blocks.iterator();
+        while(blocksIterator.hasNext()){
+            if(blocksIterator.next() == null){
+                blocksIterator.remove();
+            }
+        }
+    }
+
+
+    public List<AbstractPageBlock> splitZonesAtRulings(List<TextPositionSequence> textPositions, List<Ruling> horizontalRulingLines, List<Ruling> verticalRulingLines) {
+
+        int indexOnPage = 0;
+        List<TextPositionSequence> chunkWords = new ArrayList<>();
+        List<AbstractPageBlock> chunkBlockList = new ArrayList<>();
+
+        float minX = 1000, maxX = 0, minY = 1000, maxY = 0;
+        TextPositionSequence prev = null;
+
+        for (TextPositionSequence word : textPositions) {
+
+            boolean isSplitByRuling = isSplitByRuling(minX, minY, maxX, maxY, word, horizontalRulingLines, verticalRulingLines);
+            boolean splitByDir = prev != null && !prev.getDir().equals(word.getDir());
+
+            if (prev != null && (splitByDir || isSplitByRuling)) {
+
+                TextPageBlock cb1 = buildTextBlock(chunkWords, indexOnPage);
+                indexOnPage++;
+
+                chunkBlockList.add(cb1);
+                chunkWords = new ArrayList<>();
+
+                minX = 1000;
+                maxX = 0;
+                minY = 1000;
+                maxY = 0;
+                prev = null;
+            }
+
+            chunkWords.add(word);
+
+            prev = word;
+            if (word.getMinXDirAdj() < minX) {
+                minX = word.getMinXDirAdj();
+            }
+            if (word.getMaxXDirAdj() > maxX) {
+                maxX = word.getMaxXDirAdj();
+            }
+            if (word.getMinYDirAdj() < minY) {
+                minY = word.getMinYDirAdj();
+            }
+            if (word.getMaxYDirAdj() > maxY) {
+                maxY = word.getMaxYDirAdj();
+            }
+        }
+
+        TextPageBlock cb1 = buildTextBlock(chunkWords, indexOnPage);
+        if (cb1 != null) {
+            chunkBlockList.add(cb1);
+        }
+
+        return chunkBlockList;
+    }
+
+
+    private boolean equalsWithThreshold(float f1, float f2) {
+
+        return Math.abs(f1 - f2) < THRESHOLD;
+    }
+
+
+    private TextPageBlock buildTextBlock(List<TextPositionSequence> wordBlockList, int indexOnPage) {
+
+        TextPageBlock textBlock = null;
+
+        FloatFrequencyCounter lineHeightFrequencyCounter = new FloatFrequencyCounter();
+        FloatFrequencyCounter fontSizeFrequencyCounter = new FloatFrequencyCounter();
+        FloatFrequencyCounter spaceFrequencyCounter = new FloatFrequencyCounter();
+        StringFrequencyCounter fontFrequencyCounter = new StringFrequencyCounter();
+        StringFrequencyCounter styleFrequencyCounter = new StringFrequencyCounter();
+
+        for (TextPositionSequence wordBlock : wordBlockList) {
+
+            lineHeightFrequencyCounter.add(wordBlock.getTextHeight());
+            fontSizeFrequencyCounter.add(wordBlock.getFontSize());
+            spaceFrequencyCounter.add(wordBlock.getSpaceWidth());
+            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());
+        }
+
+        if (textBlock != null && textBlock.getSequences() != null && textBlock.getSequences().stream().map(t -> round(t.getMinYDirAdj(), 3)).collect(toSet()).size() == 1) {
+            textBlock.getSequences().sort(Comparator.comparing(TextPositionSequence::getMinXDirAdj));
+        }
+        return textBlock;
+    }
+
+
+    private boolean isSplitByRuling(float minX,
+                                    float minY,
+                                    float maxX,
+                                    float maxY,
+                                    TextPositionSequence word,
+                                    List<Ruling> horizontalRulingLines,
+                                    List<Ruling> verticalRulingLines) {
+
+        return isSplitByRuling(maxX,
+                minY,
+                word.getMinXDirAdj(),
+                word.getMinYDirAdj(),
+                verticalRulingLines,
+                word.getDir().getDegrees(),
+                word.getPageWidth(),
+                word.getPageHeight()) //
+                || isSplitByRuling(minX,
+                minY,
+                word.getMinXDirAdj(),
+                word.getMaxYDirAdj(),
+                horizontalRulingLines,
+                word.getDir().getDegrees(),
+                word.getPageWidth(),
+                word.getPageHeight()) //
+                || isSplitByRuling(maxX,
+                minY,
+                word.getMinXDirAdj(),
+                word.getMinYDirAdj(),
+                horizontalRulingLines,
+                word.getDir().getDegrees(),
+                word.getPageWidth(),
+                word.getPageHeight()) //
+                || isSplitByRuling(minX,
+                minY,
+                word.getMinXDirAdj(),
+                word.getMaxYDirAdj(),
+                verticalRulingLines,
+                word.getDir().getDegrees(),
+                word.getPageWidth(),
+                word.getPageHeight());
+    }
+
+
+    private boolean isSplitByRuling(float previousX2, float previousY1, float currentX1, float currentY1, List<Ruling> rulingLines, float dir, float pageWidth, float pageHeight) {
+
+        for (Ruling ruling : rulingLines) {
+            var line = RulingTextDirAdjustUtil.convertToDirAdj(ruling, dir, pageWidth, pageHeight);
+            if (line.intersectsLine(previousX2, previousY1, currentX1, currentY1)) {
+                return true;
+            }
+        }
+        return false;
+    }
+
+
+    private double round(float value, int decimalPoints) {
+
+        var d = Math.pow(10, decimalPoints);
+        return Math.round(value * d) / d;
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/blockification/DocuMineBlockificationService.java

@@ -61,8 +61,8 @@ public class DocuMineBlockificationService {
             boolean newLineAfterSplit = prev != null && word.getMinYDirAdj() != prev.getMinYDirAdj() && wasSplitted && splitX1 != word.getMinXDirAdj();
             boolean isSplitByRuling = isSplitByRuling(minX, minY, maxX, maxY, word, horizontalRulingLines, verticalRulingLines);
             boolean splitByDir = prev != null && !prev.getDir().equals(word.getDir());
-            boolean splitByOtherFontAndOtherY = prev != null && prev.getMaxYDirAdj() != word.getMaxYDirAdj() && (word.getFontStyle().contains("bold") && !prev.getFontStyle()
-                    .contains("bold") || prev.getFontStyle().contains("bold") && !word.getFontStyle().contains("bold"));
+            boolean splitByOtherFontAndOtherY = prev != null && Math.abs(prev.getMaxYDirAdj() - word.getMaxYDirAdj()) > word.getTextHeight() * 0.2 && (word.getFontStyle()
+                    .contains("bold") && !prev.getFontStyle().contains("bold") || prev.getFontStyle().contains("bold") && !word.getFontStyle().contains("bold"));
 
             Matcher matcher = pattern.matcher(chunkWords.stream().collect(Collectors.joining(" ")).toString());
             boolean startsOnSameX = Math.abs(minX - word.getMinXDirAdj()) < 5 && matcher.matches();

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/blockification/RedactManagerBlockificationService.java

@@ -13,10 +13,13 @@ import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBl
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
 import com.knecon.fforesight.service.layoutparser.processor.model.FloatFrequencyCounter;
 import com.knecon.fforesight.service.layoutparser.processor.model.Orientation;
+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.table.Ruling;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.StringFrequencyCounter;
 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.RectangleTransformations;
 import com.knecon.fforesight.service.layoutparser.processor.utils.RulingTextDirAdjustUtil;
 
 @SuppressWarnings("all")
@@ -31,12 +34,12 @@ public class RedactManagerBlockificationService {
      * This method must use text direction adjusted postions (DirAdj). Where {0,0} is on the upper left. Never try to change this!
      * Rulings (TablePageBlock lines) must be adjusted to the text directions as well, when checking if a block is split by a ruling.
      *
-     * @param textPositions         The words of a page.
-     * @param horizontalRulingLines Horizontal table lines.
-     * @param verticalRulingLines   Vertical table lines.
+     * @param textPositions The words of a page.
      * @return Page object that contains the Textblock and text statistics.
      */
-    public ClassificationPage blockify(List<TextPositionSequence> textPositions, List<Ruling> horizontalRulingLines, List<Ruling> verticalRulingLines) {
+    public ClassificationPage blockify(List<TextPositionSequence> textPositions, List<Cell> cells) {
+
+        CleanRulings usedRulings = RectangleTransformations.extractRulings(cells);
 
         int indexOnPage = 0;
         List<TextPositionSequence> chunkWords = new ArrayList<>();
@@ -54,7 +57,7 @@ public class RedactManagerBlockificationService {
             boolean splitByX = prev != null && maxX + 50 < word.getMinXDirAdj() && prev.getMinYDirAdj() == word.getMinYDirAdj();
             boolean xIsBeforeFirstX = prev != null && word.getMinXDirAdj() < minX;
             boolean newLineAfterSplit = prev != null && word.getMinYDirAdj() != prev.getMinYDirAdj() && wasSplitted && splitX1 != word.getMinXDirAdj();
-            boolean isSplitByRuling = isSplitByRuling(minX, minY, maxX, maxY, word, horizontalRulingLines, verticalRulingLines);
+            boolean isSplitByRuling = isSplitByRuling(minX, minY, maxX, maxY, word, usedRulings.getHorizontal(), usedRulings.getVertical());
             boolean splitByDir = prev != null && !prev.getDir().equals(word.getDir());
 
             if (prev != null && (lineSeparation || startFromTop || splitByX || splitByDir || isSplitByRuling)) {

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/blockification/TaasBlockificationService.java

@@ -1,330 +0,0 @@
-package com.knecon.fforesight.service.layoutparser.processor.services.blockification;
-
-
-// TODO: figure out, why this fails the build
-// import static com.knecon.fforesight.service.layoutparser.processor.services.factory.SearchTextWithTextPositionFactory.HEIGHT_PADDING;
-
-import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
-import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
-import com.knecon.fforesight.service.layoutparser.processor.model.Orientation;
-import com.knecon.fforesight.service.layoutparser.processor.model.table.Ruling;
-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.RulingTextDirAdjustUtil;
-import org.springframework.stereotype.Service;
-
-import java.util.*;
-import java.util.regex.Matcher;
-import java.util.regex.Pattern;
-import java.util.stream.Stream;
-
-@Service
-@SuppressWarnings("all")
-public class TaasBlockificationService {
-
-    private static final float THRESHOLD = 1f;
-    private static final float Y_GAP_SPLIT_HEIGHT_MODIFIER = 1.25f; // multiplied with text height
-    private static final float INTERSECTS_Y_THRESHOLD = 4;// 2 * HEIGHT_PADDING // This is exactly 2 times our position height padding. This is required to find boxes that are visually intersecting.
-    private static final int X_GAP_SPLIT_CONSTANT = 50;
-    public static final int X_ALIGNMENT_THRESHOLD = 1;
-    public static final int NEGATIVE_X_GAP_THRESHOLD = -5;
-
-    private Pattern listIdentifier = Pattern.compile("^(?:(?:[1-9]|1\\d|20|[ivxlc]|[a-z])\\s*(?:[.)]))|\\uF0B7", Pattern.CASE_INSENSITIVE);
-
-
-    /**
-     * This method is building blocks by expanding the minX/maxX and minY/maxY value on each word that is not split by the conditions.
-     * This method must use text direction adjusted postions (DirAdj). Where {0,0} is on the upper left. Never try to change this!
-     * Rulings (Table lines) must be adjusted to the text directions as well, when checking if a block is split by a ruling.
-     *
-     * @param textPositions         The words of a page.
-     * @param horizontalRulingLines Horizontal table lines.
-     * @param verticalRulingLines   Vertical table lines.
-     * @return ClassificationPage object that contains the Textblock and text statistics.
-     */
-    public ClassificationPage blockify(List<TextPositionSequence> textPositions, List<Ruling> horizontalRulingLines, List<Ruling> verticalRulingLines) {
-
-        List<TextPageBlock> classificationTextBlocks = constructFineGranularTextPageBlocks(textPositions, horizontalRulingLines, verticalRulingLines);
-        classificationTextBlocks = mergeTextPageBlocksAligningX(classificationTextBlocks);
-        classificationTextBlocks = mergeIntersectingTextBlocksUntilConvergence(classificationTextBlocks);
-
-        return new ClassificationPage(new ArrayList<>(classificationTextBlocks.stream().map(classificationTextBlock -> (AbstractPageBlock) classificationTextBlock).toList()));
-    }
-
-
-    private List<TextPageBlock> mergeIntersectingTextBlocksUntilConvergence(List<TextPageBlock> classificationTextBlocks) {
-
-        int currentSize = classificationTextBlocks.size();
-        while (true) {
-            classificationTextBlocks = mergeTextPageBlocksAlmostIntersecting(classificationTextBlocks);
-            if (classificationTextBlocks.size() == currentSize) {
-                break;
-            }
-            currentSize = classificationTextBlocks.size();
-        }
-        return classificationTextBlocks;
-    }
-
-
-    private List<TextPageBlock> mergeTextPageBlocksAligningX(List<TextPageBlock> classificationTextBlocks) {
-
-        if (classificationTextBlocks.isEmpty()) {
-            return new ArrayList<>();
-        }
-        List<List<TextPageBlock>> textBlocksToMerge = new LinkedList<>();
-        List<TextPageBlock> currentTextBlocksToMerge = new LinkedList<>();
-        textBlocksToMerge.add(currentTextBlocksToMerge);
-        TextPageBlock previousTextBlock = null;
-        Float lastLineGap = null;
-        for (TextPageBlock currentTextBlock : classificationTextBlocks) {
-            if (previousTextBlock == null) {
-                currentTextBlocksToMerge.add(currentTextBlock);
-                previousTextBlock = currentTextBlock;
-                continue;
-            }
-
-
-            Matcher listIdentifierPattern = listIdentifier.matcher(currentTextBlock.getText());
-            boolean isListIdentifier = listIdentifierPattern.find();
-
-            boolean yGap = Math.abs(currentTextBlock.getPdfMaxY() - previousTextBlock.getPdfMinY()) < previousTextBlock.getMostPopularWordHeight() * Y_GAP_SPLIT_HEIGHT_MODIFIER;
-
-            boolean sameFont = previousTextBlock.getMostPopularWordFont().equals(currentTextBlock.getMostPopularWordFont()) && previousTextBlock.getMostPopularWordFontSize() == currentTextBlock.getMostPopularWordFontSize();
-//            boolean yGap = previousTextBlock != null && currentTextBlock.getMinYDirAdj() - maxY > Math.min(word.getHeight(), prev.getHeight()) * Y_GAP_SPLIT_HEIGHT_MODIFIER;
-
-            boolean alignsXRight = Math.abs(currentTextBlock.getPdfMaxX() - previousTextBlock.getPdfMaxX()) < X_ALIGNMENT_THRESHOLD;
-            boolean alignsXLeft = Math.abs(currentTextBlock.getPdfMinX() - previousTextBlock.getPdfMinX()) < X_ALIGNMENT_THRESHOLD;
-//            boolean smallYGap = Math.abs(currentTextBlock.getPdfMaxY() - previousTextBlock.getPdfMinY()) < yGap;
-            if (yGap && sameFont && !isListIdentifier) {
-                currentTextBlocksToMerge.add(currentTextBlock);
-
-            } else {
-                currentTextBlocksToMerge = new LinkedList<>();
-                currentTextBlocksToMerge.add(currentTextBlock);
-                textBlocksToMerge.add(currentTextBlocksToMerge);
-            }
-            previousTextBlock = currentTextBlock;
-        }
-        return textBlocksToMerge.stream().map(TextPageBlock::merge).toList();
-    }
-
-
-    private List<TextPageBlock> mergeTextPageBlocksAlmostIntersecting(List<TextPageBlock> textPageBlocks) {
-
-        Set<TextPageBlock> alreadyMerged = new HashSet<>();
-        List<List<TextPageBlock>> textBlocksToMerge = new LinkedList<>();
-        for (TextPageBlock textPageBlock : textPageBlocks) {
-            if (alreadyMerged.contains(textPageBlock)) {
-                continue;
-            }
-            alreadyMerged.add(textPageBlock);
-            textBlocksToMerge.add(Stream.concat(Stream.of(textPageBlock),
-                            textPageBlocks.stream().filter(textPageBlock2 -> textPageBlock.almostIntersects(textPageBlock2, INTERSECTS_Y_THRESHOLD, 0) && !alreadyMerged.contains(textPageBlock2)).peek(alreadyMerged::add))
-                    .toList());
-        }
-        return textBlocksToMerge.stream().map(TextPageBlock::merge).toList();
-    }
-
-
-    private void assignOrientations(List<TextPageBlock> classificationTextBlocks) {
-
-        Iterator<TextPageBlock> itty = classificationTextBlocks.iterator();
-
-        TextPageBlock previousLeft = null;
-        TextPageBlock previousRight = null;
-        while (itty.hasNext()) {
-            TextPageBlock block = (TextPageBlock) itty.next();
-
-            if (previousLeft != null && block.getOrientation().equals(Orientation.LEFT)) {
-                if (previousLeft.getMinY() > block.getMinY() && block.getMaxY() + block.getMostPopularWordHeight() > previousLeft.getMinY()) {
-                    previousLeft.add(block);
-                    itty.remove();
-                    continue;
-                }
-            }
-
-            if (previousRight != null && block.getOrientation().equals(Orientation.RIGHT)) {
-                if (previousRight.getMinY() > block.getMinY() && block.getMaxY() + block.getMostPopularWordHeight() > previousRight.getMinY()) {
-                    previousRight.add(block);
-                    itty.remove();
-                    continue;
-                }
-            }
-
-            if (block.getOrientation().equals(Orientation.LEFT)) {
-                previousLeft = block;
-            } else if (block.getOrientation().equals(Orientation.RIGHT)) {
-                previousRight = block;
-            }
-        }
-
-        itty = classificationTextBlocks.iterator();
-        TextPageBlock previous = null;
-        while (itty.hasNext()) {
-            TextPageBlock block = (TextPageBlock) itty.next();
-
-            if (previous != null && previous.getOrientation().equals(Orientation.LEFT) && block.getOrientation().equals(Orientation.LEFT) && equalsWithThreshold(
-                    block.getMaxY(),
-                    previous.getMaxY()) || previous != null && previous.getOrientation().equals(Orientation.LEFT) && block.getOrientation()
-                    .equals(Orientation.RIGHT) && equalsWithThreshold(block.getMaxY(), previous.getMaxY())) {
-                previous.add(block);
-                itty.remove();
-                continue;
-            }
-
-            previous = block;
-        }
-    }
-
-
-    private List<TextPageBlock> constructFineGranularTextPageBlocks(List<TextPositionSequence> textPositions,
-                                                                    List<Ruling> horizontalRulingLines,
-                                                                    List<Ruling> verticalRulingLines) {
-
-        int indexOnPage = 0;
-        List<TextPositionSequence> wordClusterToCombine = new ArrayList<>();
-        List<TextPageBlock> classificationTextBlocks = new ArrayList<>();
-
-        float minX = 1000, maxX = 0, minY = 1000, maxY = 0;
-        TextPositionSequence prev = null;
-        // TODO: make static final constant
-
-
-        boolean wasSplitted = false;
-        Float splitX1 = null;
-        for (TextPositionSequence word : textPositions) {
-
-            Matcher listIdentifierPattern = listIdentifier.matcher(word.toString());
-
-            boolean yGap = prev != null && word.getMinYDirAdj() - maxY > Math.min(word.getHeight(), prev.getHeight()) * Y_GAP_SPLIT_HEIGHT_MODIFIER;
-            boolean sameLine = prev != null && equalsWithThreshold(prev.getMinYDirAdj(), word.getMinYDirAdj());
-            boolean positiveXGapInline = prev != null && maxX + X_GAP_SPLIT_CONSTANT < word.getMinXDirAdj() && sameLine;
-            boolean negativeXGap = prev != null && word.getMinXDirAdj() - minX < NEGATIVE_X_GAP_THRESHOLD;
-            boolean startFromTop = prev != null && word.getMinYDirAdj() < prev.getMinYDirAdj() - prev.getTextHeight();
-            boolean newLineAfterSplit = prev != null && word.getMinYDirAdj() != prev.getMinYDirAdj() && wasSplitted && splitX1 != word.getMinXDirAdj();
-            boolean splitByRuling = isSplitByRuling(minX, minY, maxX, maxY, word, horizontalRulingLines, verticalRulingLines);
-            boolean splitByDir = prev != null && !prev.getDir().equals(word.getDir());
-            boolean fontChange = prev != null && (!word.getFont().equals(prev.getFont()) || !word.getFontStyle()
-                    .equals(prev.getFontStyle()) || word.getFontSize() != prev.getFontSize());
-            boolean newline = prev != null && Math.abs(word.getMinYDirAdj() - prev.getMinYDirAdj()) > word.getHeight();
-            boolean isListIdentifier = listIdentifierPattern.matches();
-
-            if (prev != null && (prev.isParagraphStart() || negativeXGap || positiveXGapInline || yGap || startFromTop || splitByRuling || (newline && (fontChange || isListIdentifier)))) {
-//            if (prev != null && (lineSeparation || startFromTop || splitByX || splitByDir || isSplitByRuling)) {
-
-                Orientation prevOrientation = null;
-                if (!classificationTextBlocks.isEmpty()) {
-                    prevOrientation = classificationTextBlocks.get(classificationTextBlocks.size() - X_ALIGNMENT_THRESHOLD).getOrientation();
-                }
-
-                TextPageBlock classificationTextBlock = TextPageBlock.fromTextPositionSequences(wordClusterToCombine);
-
-                classificationTextBlocks.add(classificationTextBlock);
-                wordClusterToCombine = new ArrayList<>();
-
-                if (positiveXGapInline && !splitByRuling) {
-                    wasSplitted = true;
-                    classificationTextBlock.setOrientation(Orientation.LEFT);
-                    splitX1 = word.getMinXDirAdj();
-                } else if (newLineAfterSplit && !splitByRuling) {
-                    wasSplitted = false;
-                    classificationTextBlock.setOrientation(Orientation.RIGHT);
-                    splitX1 = null;
-                } else if (prevOrientation != null && prevOrientation.equals(Orientation.RIGHT) && (yGap || !startFromTop || !positiveXGapInline || !newLineAfterSplit || !splitByRuling)) {
-                    classificationTextBlock.setOrientation(Orientation.LEFT);
-                }
-
-                minX = 1000;
-                maxX = 0;
-                minY = 1000;
-                maxY = 0;
-                prev = null;
-            }
-
-            wordClusterToCombine.add(word);
-
-            prev = word;
-            if (word.getMinXDirAdj() < minX) {
-                minX = word.getMinXDirAdj();
-            }
-            if (word.getMaxXDirAdj() > maxX) {
-                maxX = word.getMaxXDirAdj();
-            }
-            if (word.getMinYDirAdj() < minY) {
-                minY = word.getMinYDirAdj();
-            }
-            if (word.getMaxYDirAdj() > maxY) {
-                maxY = word.getMaxYDirAdj();
-            }
-        }
-
-        TextPageBlock classificationTextBlock = TextPageBlock.fromTextPositionSequences(wordClusterToCombine);
-        if (classificationTextBlock != null) {
-            classificationTextBlocks.add(classificationTextBlock);
-        }
-        return classificationTextBlocks;
-    }
-
-
-    private boolean equalsWithThreshold(float f1, float f2) {
-
-        return Math.abs(f1 - f2) < THRESHOLD;
-    }
-
-
-    private boolean isSplitByRuling(float minX,
-                                    float minY,
-                                    float maxX,
-                                    float maxY,
-                                    TextPositionSequence word,
-                                    List<Ruling> horizontalRulingLines,
-                                    List<Ruling> verticalRulingLines) {
-
-        return isSplitByRuling(maxX,
-                minY,
-                word.getMinXDirAdj(),
-                word.getMinYDirAdj(),
-                verticalRulingLines,
-                word.getDir().getDegrees(),
-                word.getPageWidth(),
-                word.getPageHeight()) //
-                || isSplitByRuling(minX,
-                minY,
-                word.getMinXDirAdj(),
-                word.getMaxYDirAdj(),
-                horizontalRulingLines,
-                word.getDir().getDegrees(),
-                word.getPageWidth(),
-                word.getPageHeight()) //
-                || isSplitByRuling(maxX,
-                minY,
-                word.getMinXDirAdj(),
-                word.getMinYDirAdj(),
-                horizontalRulingLines,
-                word.getDir().getDegrees(),
-                word.getPageWidth(),
-                word.getPageHeight()) //
-                || isSplitByRuling(minX,
-                minY,
-                word.getMinXDirAdj(),
-                word.getMaxYDirAdj(),
-                verticalRulingLines,
-                word.getDir().getDegrees(),
-                word.getPageWidth(),
-                word.getPageHeight()); //
-    }
-
-
-    private boolean isSplitByRuling(float previousX2, float previousY1, float currentX1, float currentY1, List<Ruling> rulingLines, float dir, float pageWidth, float pageHeight) {
-
-        for (Ruling ruling : rulingLines) {
-            var line = RulingTextDirAdjustUtil.convertToDirAdj(ruling, dir, pageWidth, pageHeight);
-            if (line.intersectsLine(previousX2, previousY1, currentX1, currentY1)) {
-                return true;
-            }
-        }
-        return false;
-    }
-
-}
-

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/classification/ClarifyndClassificationService.java

@@ -3,7 +3,6 @@ package com.knecon.fforesight.service.layoutparser.processor.services.classifica
 import java.util.List;
 import java.util.regex.Pattern;
 
-import com.knecon.fforesight.service.layoutparser.processor.utils.MarkedContentUtils;
 import org.springframework.stereotype.Service;
 
 import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
@@ -11,7 +10,7 @@ import com.knecon.fforesight.service.layoutparser.processor.model.Classification
 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.text.TextPageBlock;
-import com.knecon.fforesight.service.layoutparser.processor.services.BodyTextFrameService;
+import com.knecon.fforesight.service.layoutparser.processor.utils.MarkedContentUtils;
 import com.knecon.fforesight.service.layoutparser.processor.utils.PositionUtils;
 
 import lombok.RequiredArgsConstructor;
@@ -20,26 +19,21 @@ import lombok.extern.slf4j.Slf4j;
 @Slf4j
 @Service
 @RequiredArgsConstructor
-public class TaasClassificationService {
-
-    private final BodyTextFrameService bodyTextFrameService;
-
+public class ClarifyndClassificationService {
 
     public void classifyDocument(ClassificationDocument document) {
 
-
         List<Float> headlineFontSizes = document.getFontSizeCounter().getHighterThanMostPopular();
 
         log.debug("Document FontSize counters are: {}", document.getFontSizeCounter().getCountPerValue());
 
         for (ClassificationPage page : document.getPages()) {
-
             classifyPage(page, document, headlineFontSizes);
         }
     }
 
 
-    public void classifyPage(ClassificationPage page, ClassificationDocument document, List<Float> headlineFontSizes) {
+    private void classifyPage(ClassificationPage page, ClassificationDocument document, List<Float> headlineFontSizes) {
 
         for (AbstractPageBlock textBlock : page.getTextBlocks()) {
             if (textBlock instanceof TextPageBlock) {
@@ -49,20 +43,25 @@ public class TaasClassificationService {
     }
 
 
-    public void classifyBlock(TextPageBlock textBlock, ClassificationPage page, ClassificationDocument document, List<Float> headlineFontSizes) {
+    private void classifyBlock(TextPageBlock textBlock, ClassificationPage page, ClassificationDocument document, List<Float> headlineFontSizes) {
 
         var bodyTextFrame = page.getBodyTextFrame();
 
         if (document.getFontSizeCounter().getMostPopular() == null) {
-            textBlock.setClassification(PageBlockType.OTHER);
+            textBlock.setClassification(PageBlockType.PARAGRAPH);
             return;
         }
-        if (MarkedContentUtils.intersects(textBlock, page.getMarkedContentBboxPerType(), MarkedContentUtils.HEADER)
-                || PositionUtils.isOverBodyTextFrame(bodyTextFrame, textBlock, page.getRotation())) {
-            textBlock.setClassification(PageBlockType.HEADER);
-        } else if (MarkedContentUtils.intersects(textBlock, page.getMarkedContentBboxPerType(), MarkedContentUtils.FOOTER)
-                || PositionUtils.isUnderBodyTextFrame(bodyTextFrame, textBlock, page.getRotation())) {
-            textBlock.setClassification(PageBlockType.FOOTER);
+        if (MarkedContentUtils.intersects(textBlock, page.getMarkedContentBboxPerType(), MarkedContentUtils.HEADER) || PositionUtils.isOverBodyTextFrame(bodyTextFrame,
+                textBlock,
+                page.getRotation()) && (document.getFontSizeCounter().getMostPopular() == null || textBlock.getHighestFontSize() <= document.getFontSizeCounter()
+                .getMostPopular())) {
+            textBlock.setClassification(PageBlockType.PARAGRAPH);
+
+        } else if (MarkedContentUtils.intersects(textBlock, page.getMarkedContentBboxPerType(), MarkedContentUtils.FOOTER) || PositionUtils.isUnderBodyTextFrame(bodyTextFrame,
+                textBlock,
+                page.getRotation()) && (document.getFontSizeCounter().getMostPopular() == null || textBlock.getHighestFontSize() <= document.getFontSizeCounter()
+                .getMostPopular())) {
+            textBlock.setClassification(PageBlockType.PARAGRAPH);
         } else if (page.getPageNumber() == 1 && (PositionUtils.getHeightDifferenceBetweenChunkWordAndDocumentWord(textBlock,
                 document.getTextHeightCounter().getMostPopular()) > 2.5 && textBlock.getHighestFontSize() > document.getFontSizeCounter().getMostPopular() || page.getTextBlocks()
                 .size() == 1)) {
@@ -107,8 +106,9 @@ public class TaasClassificationService {
         } else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock)) {
             textBlock.setClassification(PageBlockType.PARAGRAPH_UNKNOWN);
         } else {
-            textBlock.setClassification(PageBlockType.OTHER);
+            textBlock.setClassification(PageBlockType.PARAGRAPH);
         }
     }
 
 }
+

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/classification/RedactManagerClassificationService.java

@@ -3,14 +3,15 @@ package com.knecon.fforesight.service.layoutparser.processor.services.classifica
 import java.util.List;
 import java.util.regex.Pattern;
 
-import com.knecon.fforesight.service.layoutparser.processor.utils.MarkedContentUtils;
 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.ClassificationPage;
 import com.knecon.fforesight.service.layoutparser.processor.model.PageBlockType;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.ImageType;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
+import com.knecon.fforesight.service.layoutparser.processor.utils.MarkedContentUtils;
 import com.knecon.fforesight.service.layoutparser.processor.utils.PositionUtils;
 
 import lombok.RequiredArgsConstructor;
@@ -21,7 +22,6 @@ import lombok.extern.slf4j.Slf4j;
 @RequiredArgsConstructor
 public class RedactManagerClassificationService {
 
-
     public void classifyDocument(ClassificationDocument document) {
 
         List<Float> headlineFontSizes = document.getFontSizeCounter().getHighterThanMostPopular();
@@ -52,14 +52,25 @@ public class RedactManagerClassificationService {
             textBlock.setClassification(PageBlockType.OTHER);
             return;
         }
-        if (MarkedContentUtils.intersects(textBlock, page.getMarkedContentBboxPerType(), MarkedContentUtils.HEADER)
-                || PositionUtils.isOverBodyTextFrame(bodyTextFrame, textBlock, page.getRotation()) && (document.getFontSizeCounter()
-                .getMostPopular() == null || textBlock.getHighestFontSize() <= document.getFontSizeCounter().getMostPopular())) {
+
+        if (page.getImages()
+                .stream()
+                .filter(image -> image.getImageType().equals(ImageType.GRAPHIC))
+                .anyMatch(graphic -> graphic.getPosition().intersects(textBlock.getPdfMinX(), textBlock.getPdfMinY(), textBlock.getWidth(), textBlock.getHeight()))) {
+            textBlock.setClassification(PageBlockType.PARAGRAPH);
+            return;
+        }
+
+        if (MarkedContentUtils.intersects(textBlock, page.getMarkedContentBboxPerType(), MarkedContentUtils.HEADER) || PositionUtils.isOverBodyTextFrame(bodyTextFrame,
+                textBlock,
+                page.getRotation()) && (document.getFontSizeCounter().getMostPopular() == null || textBlock.getHighestFontSize() <= document.getFontSizeCounter()
+                .getMostPopular())) {
             textBlock.setClassification(PageBlockType.HEADER);
 
-        } else if (MarkedContentUtils.intersects(textBlock, page.getMarkedContentBboxPerType(), MarkedContentUtils.FOOTER)
-                || PositionUtils.isUnderBodyTextFrame(bodyTextFrame, textBlock, page.getRotation()) && (document.getFontSizeCounter()
-                .getMostPopular() == null || textBlock.getHighestFontSize() <= document.getFontSizeCounter().getMostPopular())) {
+        } else if (MarkedContentUtils.intersects(textBlock, page.getMarkedContentBboxPerType(), MarkedContentUtils.FOOTER) || PositionUtils.isUnderBodyTextFrame(bodyTextFrame,
+                textBlock,
+                page.getRotation()) && (document.getFontSizeCounter().getMostPopular() == null || textBlock.getHighestFontSize() <= document.getFontSizeCounter()
+                .getMostPopular())) {
             textBlock.setClassification(PageBlockType.FOOTER);
         } else if (page.getPageNumber() == 1 && (PositionUtils.getHeightDifferenceBetweenChunkWordAndDocumentWord(textBlock,
                 document.getTextHeightCounter().getMostPopular()) > 2.5 && textBlock.getHighestFontSize() > document.getFontSizeCounter().getMostPopular() || page.getTextBlocks()

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/factory/DocumentGraphFactory.java

@@ -7,21 +7,24 @@ import static java.util.stream.Collectors.toList;
 import java.awt.geom.Rectangle2D;
 import java.util.ArrayList;
 import java.util.HashMap;
+import java.util.HashSet;
 import java.util.LinkedList;
 import java.util.List;
 import java.util.Map;
 import java.util.NoSuchElementException;
 import java.util.Set;
+import java.util.stream.Collectors;
 
+import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
+import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingType;
 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.image.ClassifiedImage;
-import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.DuplicatedParagraph;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Footer;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.GenericSemanticNode;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Header;
@@ -31,6 +34,8 @@ import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Pa
 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.textblock.AtomicTextBlock;
+import com.knecon.fforesight.service.layoutparser.processor.model.image.ClassifiedImage;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.utils.IdBuilder;
 import com.knecon.fforesight.service.layoutparser.processor.utils.TextPositionOperations;
 
@@ -44,14 +49,19 @@ import lombok.experimental.UtilityClass;
 @UtilityClass
 public class DocumentGraphFactory {
 
-    public Document buildDocumentGraph(ClassificationDocument document) {
+    public Document buildDocumentGraph(LayoutParsingType layoutParsingType, ClassificationDocument document) {
 
         Document documentGraph = new Document();
         Context context = new Context(documentGraph);
 
-        document.getPages().forEach(context::buildAndAddPageWithCounter);
-        document.getSections().stream().flatMap(section -> section.getImages().stream()).forEach(image -> context.getImages().add(image));
-        addSections(document, context);
+        document.getPages()
+                .forEach(context::buildAndAddPageWithCounter);
+        document.getSections()
+                .stream()
+                .flatMap(section -> section.getImages()
+                        .stream())
+                .forEach(image -> context.getImages().add(image));
+        addSections(layoutParsingType, document, context, documentGraph);
         addHeaderAndFooterToEachPage(document, context);
 
         documentGraph.setNumberOfPages(context.pages.size());
@@ -62,9 +72,10 @@ public class DocumentGraphFactory {
     }
 
 
-    private void addSections(ClassificationDocument document, Context context) {
+    private void addSections(LayoutParsingType layoutParsingType, ClassificationDocument classificationDocument, Context context, Document document) {
 
-        document.getSections().forEach(section -> SectionNodeFactory.addSection(null, section.getNonEmptyPageBlocks(), section.getImages(), context));
+        classificationDocument.getSections()
+                .forEach(section -> SectionNodeFactory.addSection(layoutParsingType, null, section.getNonEmptyPageBlocks(), section.getImages(), context, document));
     }
 
 
@@ -74,9 +85,14 @@ public class DocumentGraphFactory {
 
         GenericSemanticNode node;
         if (originalTextBlock.isHeadline()) {
-            node = Headline.builder().documentTree(context.getDocumentTree()).build();
+            node = Headline.builder().documentTree(context.getDocumentTree())
+                    .build();
+        } else if (originalTextBlock.isToDuplicate()) {
+            node = DuplicatedParagraph.builder().documentTree(context.getDocumentTree())
+                    .build();
         } else {
-            node = Paragraph.builder().documentTree(context.getDocumentTree()).build();
+            node = Paragraph.builder().documentTree(context.getDocumentTree())
+                    .build();
         }
 
         page.getMainBody().add(node);
@@ -84,14 +100,33 @@ public class DocumentGraphFactory {
         List<TextPageBlock> textBlocks = new ArrayList<>();
         textBlocks.add(originalTextBlock);
         textBlocks.addAll(textBlocksToMerge);
+
         AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock(TextPositionOperations.mergeAndSortTextPositionSequenceByYThenX(textBlocks), node, context, page);
+
+        if (node instanceof DuplicatedParagraph duplicatedParagraph) {
+            AtomicTextBlock unsortedTextBlock = context.textBlockFactory.buildAtomicTextBlock(textBlocks.stream()
+                                                                                                      .flatMap(tb -> tb.getSequences()
+                                                                                                              .stream())
+                                                                                                      .collect(Collectors.toList()), node, context, page);
+            duplicatedParagraph.setUnsortedLeafTextBlock(unsortedTextBlock);
+        }
+
         List<Integer> treeId = context.documentTree.createNewChildEntryAndReturnId(parentNode, node);
         node.setLeafTextBlock(textBlock);
         node.setTreeId(treeId);
     }
 
 
-    public void addImage(Section section, ClassifiedImage image, Context context) {
+    public void addImage(GenericSemanticNode parent, ClassifiedImage image, Context context) {
+
+        Image imageNode = createImage(image, context);
+        List<Integer> treeId = context.getDocumentTree().createNewChildEntryAndReturnId(parent, imageNode);
+        imageNode.setTreeId(treeId);
+        imageNode.setLeafTextBlock(context.textBlockFactory.emptyTextBlock(parent, context, context.getPage(image.getPage())));
+    }
+
+
+    private Image createImage(ClassifiedImage image, Context context) {
 
         Rectangle2D position = image.getPosition();
         Page page = context.getPage(image.getPage());
@@ -104,9 +139,7 @@ public class DocumentGraphFactory {
                 .documentTree(context.getDocumentTree())
                 .build();
         page.getMainBody().add(imageNode);
-
-        List<Integer> tocId = context.getDocumentTree().createNewChildEntryAndReturnId(section, imageNode);
-        imageNode.setTreeId(tocId);
+        return imageNode;
     }
 
 
@@ -145,11 +178,12 @@ public class DocumentGraphFactory {
     private void addFooter(List<TextPageBlock> textBlocks, Context context) {
 
         Page page = context.getPage(textBlocks.get(0).getPage());
-        Footer footer = Footer.builder().documentTree(context.getDocumentTree()).build();
+        Footer footer = Footer.builder().documentTree(context.getDocumentTree())
+                .build();
         AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock(TextPositionOperations.mergeAndSortTextPositionSequenceByYThenX(textBlocks),
-                footer,
-                context,
-                page);
+                                                                                  footer,
+                                                                                  context,
+                                                                                  page);
         List<Integer> tocId = context.getDocumentTree().createNewMainEntryAndReturnId(footer);
         footer.setTreeId(tocId);
         footer.setLeafTextBlock(textBlock);
@@ -160,7 +194,8 @@ public class DocumentGraphFactory {
     public void addHeader(List<TextPageBlock> textBlocks, Context context) {
 
         Page page = context.getPage(textBlocks.get(0).getPage());
-        Header header = Header.builder().documentTree(context.getDocumentTree()).build();
+        Header header = Header.builder().documentTree(context.getDocumentTree())
+                .build();
         AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock(TextPositionOperations.mergeAndSortTextPositionSequenceByYThenX(textBlocks), header, 0, page);
         List<Integer> tocId = context.getDocumentTree().createNewMainEntryAndReturnId(header);
         header.setTreeId(tocId);
@@ -172,7 +207,8 @@ public class DocumentGraphFactory {
     private void addEmptyFooter(int pageIndex, Context context) {
 
         Page page = context.getPage(pageIndex);
-        Footer footer = Footer.builder().documentTree(context.getDocumentTree()).build();
+        Footer footer = Footer.builder().documentTree(context.getDocumentTree())
+                .build();
         AtomicTextBlock textBlock = context.textBlockFactory.emptyTextBlock(footer, context, page);
         List<Integer> tocId = context.getDocumentTree().createNewMainEntryAndReturnId(footer);
         footer.setTreeId(tocId);
@@ -184,7 +220,8 @@ public class DocumentGraphFactory {
     private void addEmptyHeader(int pageIndex, Context context) {
 
         Page page = context.getPage(pageIndex);
-        Header header = Header.builder().documentTree(context.getDocumentTree()).build();
+        Header header = Header.builder().documentTree(context.getDocumentTree())
+                .build();
         AtomicTextBlock textBlock = context.textBlockFactory.emptyTextBlock(header, 0, page);
         List<Integer> tocId = context.getDocumentTree().createNewMainEntryAndReturnId(header);
         header.setTreeId(tocId);

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/factory/SearchTextWithTextPositionFactory.java

@@ -8,10 +8,10 @@ import java.util.List;
 import java.util.Locale;
 import java.util.Objects;
 
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.Boundary;
 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.model.graph.Boundary;
 
 import lombok.experimental.UtilityClass;
 
@@ -110,6 +110,7 @@ public class SearchTextWithTextPositionFactory {
         return context.stringIdx - context.lastHyphenIdx < MAX_HYPHEN_LINEBREAK_DISTANCE;
     }
 
+
     private static List<Boundary> mergeToBoundaries(List<Integer> integers) {
 
         if (integers.isEmpty()) {
@@ -125,8 +126,9 @@ public class SearchTextWithTextPositionFactory {
             }
             end = current + 1;
         }
-        if (boundaries.isEmpty())
+        if (boundaries.isEmpty()) {
             boundaries.add(new Boundary(start, end));
+        }
         return boundaries;
     }
 
@@ -138,6 +140,7 @@ public class SearchTextWithTextPositionFactory {
         }
     }
 
+
     private boolean isLineBreak(RedTextPosition currentTextPosition, RedTextPosition previousTextPosition) {
 
         return Objects.equals(currentTextPosition.getUnicode(), "\n") || isDeltaYLargerThanTextHeight(currentTextPosition, previousTextPosition);
@@ -177,7 +180,7 @@ public class SearchTextWithTextPositionFactory {
     }
 
 
-    private Rectangle2D mapRedTextPositionToInitialUserSpace(RedTextPosition textPosition, TextPositionSequence sequence) {
+    public Rectangle2D mapRedTextPositionToInitialUserSpace(RedTextPosition textPosition, TextPositionSequence sequence) {
 
         float textHeight = sequence.getTextHeight() + HEIGHT_PADDING;
         Rectangle2D rectangle2D = new Rectangle2D.Double(textPosition.getXDirAdj(),

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/factory/SectionNodeFactory.java

@@ -4,19 +4,22 @@ import static java.lang.String.format;
 import static java.util.Collections.emptyList;
 import static java.util.stream.Collectors.groupingBy;
 
+import java.util.ArrayList;
 import java.util.HashSet;
 import java.util.LinkedList;
 import java.util.List;
 import java.util.Map;
 import java.util.Set;
 
+import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingType;
 import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
-import com.knecon.fforesight.service.layoutparser.processor.model.image.ClassifiedImage;
-import com.knecon.fforesight.service.layoutparser.processor.model.table.TablePageBlock;
-import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.GenericSemanticNode;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Page;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Section;
+import com.knecon.fforesight.service.layoutparser.processor.model.image.ClassifiedImage;
+import com.knecon.fforesight.service.layoutparser.processor.model.table.TablePageBlock;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.utils.TableMergingUtility;
 
 import lombok.experimental.UtilityClass;
@@ -24,27 +27,52 @@ import lombok.experimental.UtilityClass;
 @UtilityClass
 public class SectionNodeFactory {
 
-    public void addSection(GenericSemanticNode parentNode, List<AbstractPageBlock> pageBlocks, List<ClassifiedImage> images, DocumentGraphFactory.Context context) {
+    public void addSection(LayoutParsingType layoutParsingType,
+                           GenericSemanticNode parentNode,
+                           List<AbstractPageBlock> pageBlocks,
+                           List<ClassifiedImage> images,
+                           DocumentGraphFactory.Context context,
+                           Document document) {
+
+        // This is for the case where we have images on a page without any text/footer/header.
+        // The pageBlocks list is empty, but we still need to add those images to the document.
+        if (!images.isEmpty() && pageBlocks.isEmpty()) {
+            images.stream()
+                    .distinct()
+                    .forEach(image -> DocumentGraphFactory.addImage(document, image, context));
+            return;
+        }
 
         if (pageBlocks.isEmpty()) {
             return;
         }
-        Map<Integer, List<AbstractPageBlock>> blocksPerPage = pageBlocks.stream().collect(groupingBy(AbstractPageBlock::getPage));
-        Section section = Section.builder().documentTree(context.getDocumentTree()).build();
+
+        Map<Integer, List<AbstractPageBlock>> blocksPerPage = pageBlocks.stream()
+                .collect(groupingBy(AbstractPageBlock::getPage));
+        Section section = Section.builder().documentTree(context.getDocumentTree())
+                .build();
 
         context.getSections().add(section);
-        blocksPerPage.keySet().forEach(pageNumber -> addSectionNodeToPageNode(context, section, pageNumber));
+        blocksPerPage.keySet()
+                .forEach(pageNumber -> addSectionNodeToPageNode(context, section, pageNumber));
 
         section.setTreeId(getTreeId(parentNode, context, section));
 
-        addFirstHeadlineDirectlyToSection(pageBlocks, context, section);
+        addFirstHeadlineDirectlyToSection(layoutParsingType, pageBlocks, context, section, document);
         if (containsTablesAndTextBlocks(pageBlocks)) {
-            splitPageBlocksIntoSubSections(pageBlocks).forEach(subSectionPageBlocks -> addSection(section, subSectionPageBlocks, emptyList(), context));
+            splitPageBlocksIntoSubSections(pageBlocks).forEach(subSectionPageBlocks -> addSection(layoutParsingType,
+                                                                                                  section,
+                                                                                                  subSectionPageBlocks,
+                                                                                                  emptyList(),
+                                                                                                  context,
+                                                                                                  document));
         } else {
-            addTablesAndParagraphsAndHeadlinesToSection(pageBlocks, context, section);
+            addTablesAndParagraphsAndHeadlinesToSection(layoutParsingType, pageBlocks, context, section, document);
         }
 
-        images.stream().distinct().forEach(image -> DocumentGraphFactory.addImage(section, image, context));
+        images.stream()
+                .distinct()
+                .forEach(image -> DocumentGraphFactory.addImage(section, image, context));
     }
 
 
@@ -58,16 +86,24 @@ public class SectionNodeFactory {
     }
 
 
-    private void addFirstHeadlineDirectlyToSection(List<AbstractPageBlock> pageBlocks, DocumentGraphFactory.Context context, Section section) {
+    private void addFirstHeadlineDirectlyToSection(LayoutParsingType layoutParsingType,
+                                                   List<AbstractPageBlock> pageBlocks,
+                                                   DocumentGraphFactory.Context context,
+                                                   Section section,
+                                                   Document document) {
 
         if (pageBlocks.get(0).isHeadline()) {
-            addTablesAndParagraphsAndHeadlinesToSection(List.of(pageBlocks.get(0)), context, section);
+            addTablesAndParagraphsAndHeadlinesToSection(layoutParsingType, List.of(pageBlocks.get(0)), context, section, document);
             pageBlocks.remove(0);
         }
     }
 
 
-    private void addTablesAndParagraphsAndHeadlinesToSection(List<AbstractPageBlock> pageBlocks, DocumentGraphFactory.Context context, Section section) {
+    private void addTablesAndParagraphsAndHeadlinesToSection(LayoutParsingType layoutParsingType,
+                                                             List<AbstractPageBlock> pageBlocks,
+                                                             DocumentGraphFactory.Context context,
+                                                             Section section,
+                                                             Document document) {
 
         Set<AbstractPageBlock> alreadyMerged = new HashSet<>();
         List<AbstractPageBlock> remainingBlocks = new LinkedList<>(pageBlocks);
@@ -80,13 +116,23 @@ public class SectionNodeFactory {
             remainingBlocks.removeAll(alreadyMerged);
 
             if (abstractPageBlock instanceof TextPageBlock) {
-                List<TextPageBlock> textBlocks = findTextBlocksWithSameClassificationAndAlignsY((TextPageBlock) abstractPageBlock, remainingBlocks);
-                alreadyMerged.addAll(textBlocks);
-                DocumentGraphFactory.addParagraphOrHeadline(section, (TextPageBlock) abstractPageBlock, context, textBlocks);
+
+                switch (layoutParsingType) {
+                    case REDACT_MANAGER, CLARIFYND, CLARIFYND_PARAGRAPH_DEBUG -> {
+                        alreadyMerged.add(abstractPageBlock);
+                        remainingBlocks.remove(abstractPageBlock);
+                        DocumentGraphFactory.addParagraphOrHeadline(section, (TextPageBlock) abstractPageBlock, context, new ArrayList<>());
+                    }
+                    default -> {
+                        List<TextPageBlock> textBlocks = findTextBlocksWithSameClassificationAndAlignsY((TextPageBlock) abstractPageBlock, remainingBlocks);
+                        alreadyMerged.addAll(textBlocks);
+                        DocumentGraphFactory.addParagraphOrHeadline(section, (TextPageBlock) abstractPageBlock, context, textBlocks);
+                    }
+                }
             } else if (abstractPageBlock instanceof TablePageBlock tablePageBlock) {
                 List<TablePageBlock> tablesToMerge = TableMergingUtility.findConsecutiveTablesWithSameColCountAndSameHeaders(tablePageBlock, remainingBlocks);
                 alreadyMerged.addAll(tablesToMerge);
-                TableNodeFactory.addTable(section, tablesToMerge, context);
+                TableNodeFactory.addTable(layoutParsingType, section, tablesToMerge, context, document);
             } else {
                 throw new RuntimeException(format("Unhandled AbstractPageBlockType %s!", abstractPageBlock.getClass()));
             }
@@ -96,7 +142,9 @@ public class SectionNodeFactory {
 
     private boolean containsTablesAndTextBlocks(List<AbstractPageBlock> pageBlocks) {
 
-        return pageBlocks.stream().anyMatch(pageBlock -> pageBlock instanceof TablePageBlock) && pageBlocks.stream().anyMatch(pageBlock -> pageBlock instanceof TextPageBlock);
+        return pageBlocks.stream()
+                       .anyMatch(pageBlock -> pageBlock instanceof TablePageBlock) && pageBlocks.stream()
+                       .anyMatch(pageBlock -> pageBlock instanceof TextPageBlock);
     }
 
 
@@ -112,7 +160,9 @@ public class SectionNodeFactory {
 
         List<List<AbstractPageBlock>> splitList = splitIntoCoherentList(pageBlocks);
         movePrecedingHeadlineToTableList(splitList);
-        return splitList.stream().filter(list -> !list.isEmpty()).toList();
+        return splitList.stream()
+                .filter(list -> !list.isEmpty())
+                .toList();
     }
 
 
@@ -133,7 +183,8 @@ public class SectionNodeFactory {
 
     private boolean listIsTablesOnly(List<AbstractPageBlock> abstractPageBlocks) {
 
-        return abstractPageBlocks.stream().allMatch(abstractPageBlock -> abstractPageBlock instanceof TablePageBlock);
+        return abstractPageBlocks.stream()
+                .allMatch(abstractPageBlock -> abstractPageBlock instanceof TablePageBlock);
     }
 
 
@@ -171,6 +222,7 @@ public class SectionNodeFactory {
                 .filter(abstractTextContainer -> abstractTextContainer.intersectsY(atc))
                 .map(abstractTextContainer -> (TextPageBlock) abstractTextContainer)
                 .filter(abstractTextContainer -> abstractTextContainer.getDir() == atc.getDir())
+                .filter(abstractTextContainer -> !abstractTextContainer.isToDuplicate())
                 .toList();
     }
 

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/factory/TableNodeFactory.java

@@ -7,16 +7,18 @@ import java.util.List;
 import java.util.Set;
 import java.util.stream.Collectors;
 
+import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingType;
 import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
-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.TextPositionSequence;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.GenericSemanticNode;
 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.model.graph.nodes.Table;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.TableCell;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
+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.TextPositionSequence;
 import com.knecon.fforesight.service.layoutparser.processor.utils.TextPositionOperations;
 
 import lombok.experimental.UtilityClass;
@@ -27,23 +29,30 @@ public class TableNodeFactory {
     public final double TABLE_CELL_MERGE_CONTENTS_SIZE_THRESHOLD = 0.05;
 
 
-    public void addTable(GenericSemanticNode parentNode, List<TablePageBlock> tablesToMerge, DocumentGraphFactory.Context context) {
+    public void addTable(LayoutParsingType layoutParsingType,
+                         GenericSemanticNode parentNode,
+                         List<TablePageBlock> tablesToMerge,
+                         DocumentGraphFactory.Context context,
+                         Document document) {
 
         setPageNumberInCells(tablesToMerge);
-        Set<Page> pages = tablesToMerge.stream().map(AbstractPageBlock::getPage).map(context::getPage).collect(Collectors.toSet());
-        List<List<Cell>> mergedRows = tablesToMerge.stream().map(TablePageBlock::getRows).flatMap(Collection::stream).toList();
+        Set<Page> pages = tablesToMerge.stream()
+                .map(AbstractPageBlock::getPage)
+                .map(context::getPage)
+                .collect(Collectors.toSet());
+        List<List<Cell>> mergedRows = tablesToMerge.stream()
+                .map(TablePageBlock::getRows)
+                .flatMap(Collection::stream)
+                .toList();
 
-        Table table = Table.builder()
-                .documentTree(context.getDocumentTree())
-                .numberOfCols(mergedRows.isEmpty() ? 0 : mergedRows.get(0).size())
-                .numberOfRows(mergedRows.size())
+        Table table = Table.builder().documentTree(context.getDocumentTree()).numberOfCols(mergedRows.isEmpty() ? 0 : mergedRows.get(0).size()).numberOfRows(mergedRows.size())
                 .build();
 
         pages.forEach(page -> addTableToPage(page, parentNode, table));
 
         List<Integer> treeId = context.getDocumentTree().createNewChildEntryAndReturnId(parentNode, table);
         table.setTreeId(treeId);
-        addTableCells(mergedRows, table, context);
+        addTableCells(layoutParsingType, mergedRows, table, context, document);
 
         ifTableHasNoHeadersSetFirstRowAsHeaders(table);
     }
@@ -63,7 +72,8 @@ public class TableNodeFactory {
 
     private void setPageNumberInTextBlocksWithPageNumberSetTo0(TablePageBlock table, Cell cell) {
 
-        cell.getTextBlocks().stream()//
+        cell.getTextBlocks()
+                .stream()//
                 .filter(tb -> tb.getPage() == 0)//
                 .forEach(tb -> tb.setPage(table.getPage()));
     }
@@ -82,28 +92,44 @@ public class TableNodeFactory {
 
     private void ifTableHasNoHeadersSetFirstRowAsHeaders(Table table) {
 
-        if (table.streamHeaders().findAny().isEmpty()) {
-            table.streamRow(0).forEach(tableCellNode -> tableCellNode.setHeader(true));
+        if (table.streamHeaders()
+                .findAny().isEmpty()) {
+            table.streamRow(0)
+                    .forEach(tableCellNode -> tableCellNode.setHeader(true));
         }
     }
 
 
-    private void addTableCells(List<List<Cell>> rows, Table table, DocumentGraphFactory.Context context) {
+    private void addTableCells(LayoutParsingType layoutParsingType, List<List<Cell>> rows, Table table, DocumentGraphFactory.Context context, Document document) {
 
         for (int rowIndex = 0; rowIndex < rows.size(); rowIndex++) {
             for (int colIndex = 0; colIndex < rows.get(rowIndex).size(); colIndex++) {
-                addTableCell(rows.get(rowIndex).get(colIndex), rowIndex, colIndex, table, context);
+                addTableCell(layoutParsingType,
+                             rows.get(rowIndex)
+                                     .get(colIndex),
+                             rowIndex,
+                             colIndex,
+                             table,
+                             context,
+                             document);
             }
         }
     }
 
 
     @SuppressWarnings("PMD.UnusedPrivateMethod") // PMD actually flags this wrong
-    private void addTableCell(Cell cell, int rowIndex, int colIndex, Table tableNode, DocumentGraphFactory.Context context) {
+    private void addTableCell(LayoutParsingType layoutParsingType,
+                              Cell cell,
+                              int rowIndex,
+                              int colIndex,
+                              Table tableNode,
+                              DocumentGraphFactory.Context context,
+                              Document document) {
 
         Page page = context.getPage(cell.getPageNumber());
 
-        TableCell tableCell = TableCell.builder().documentTree(context.getDocumentTree()).row(rowIndex).col(colIndex).header(cell.isHeaderCell()).bBox(cell.getBounds2D()).build();
+        TableCell tableCell = TableCell.builder().documentTree(context.getDocumentTree()).row(rowIndex).col(colIndex).header(cell.isHeaderCell()).bBox(cell.getBounds2D())
+                .build();
         page.getMainBody().add(tableCell);
 
         List<Integer> treeId = context.getDocumentTree().createNewTableChildEntryAndReturnId(tableNode, tableCell);
@@ -113,16 +139,27 @@ public class TableNodeFactory {
         if (cell.getTextBlocks().isEmpty()) {
             tableCell.setLeafTextBlock(context.getTextBlockFactory().emptyTextBlock(tableNode, context, page));
         } else if (cell.getTextBlocks().size() == 1) {
-            textBlock = context.getTextBlockFactory().buildAtomicTextBlock(cell.getTextBlocks().get(0).getSequences(), tableCell, context, page);
+            textBlock = context.getTextBlockFactory()
+                    .buildAtomicTextBlock(cell.getTextBlocks()
+                                                  .get(0).getSequences(), tableCell, context, page);
             tableCell.setLeafTextBlock(textBlock);
         } else if (firstTextBlockIsHeadline(cell)) {
-            SectionNodeFactory.addSection(tableCell, cell.getTextBlocks().stream().map(tb -> (AbstractPageBlock) tb).toList(), emptyList(), context);
+            SectionNodeFactory.addSection(layoutParsingType,
+                                          tableCell,
+                                          cell.getTextBlocks()
+                                                  .stream()
+                                                  .map(tb -> (AbstractPageBlock) tb)
+                                                  .toList(),
+                                          emptyList(),
+                                          context,
+                                          document);
         } else if (cellAreaIsSmallerThanPageAreaTimesThreshold(cell, page)) {
             List<TextPositionSequence> sequences = TextPositionOperations.mergeAndSortTextPositionSequenceByYThenX(cell.getTextBlocks());
             textBlock = context.getTextBlockFactory().buildAtomicTextBlock(sequences, tableCell, context, page);
             tableCell.setLeafTextBlock(textBlock);
         } else {
-            cell.getTextBlocks().forEach(tb -> DocumentGraphFactory.addParagraphOrHeadline(tableCell, tb, context, emptyList()));
+            cell.getTextBlocks()
+                    .forEach(tb -> DocumentGraphFactory.addParagraphOrHeadline(tableCell, tb, context, emptyList()));
         }
     }
 
@@ -135,7 +172,8 @@ public class TableNodeFactory {
 
     private boolean firstTextBlockIsHeadline(Cell cell) {
 
-        return cell.getTextBlocks().get(0).isHeadline();
+        return cell.getTextBlocks()
+                .get(0).isHeadline();
     }
 
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/graphics/Box.java

@@ -0,0 +1,162 @@
+package com.knecon.fforesight.service.layoutparser.processor.services.graphics;
+
+import java.awt.geom.AffineTransform;
+import java.awt.geom.Point2D;
+import java.awt.geom.Rectangle2D;
+import java.util.List;
+import java.util.Optional;
+
+public class Box {
+
+    public double x1;
+    public double y1;
+    public double x2;
+    public double y2;
+
+
+    public Box(double x1, double y1, double x2, double y2) {
+
+        this.x1 = x1;
+        this.y1 = y1;
+        this.x2 = x2;
+        this.y2 = y2;
+    }
+
+
+    public Box(Rectangle2D rectangle2D) {
+
+        this.x1 = rectangle2D.getMinX();
+        this.y1 = rectangle2D.getMinY();
+        this.x2 = rectangle2D.getMaxX();
+        this.y2 = rectangle2D.getMaxY();
+    }
+
+
+    public double width() {
+
+        return x2 - x1;
+    }
+
+
+    public double height() {
+
+        return y2 - y1;
+    }
+
+
+    public double xCenter() {
+
+        return (x2 + x1) / 2;
+    }
+
+
+    public double yCenter() {
+
+        return (y2 + y1) / 2;
+    }
+
+
+    public double area() {
+
+        return width() * height();
+    }
+
+
+    public Box scale(double scale) {
+
+        return new Box(x1 * scale, y1 * scale, x2 * scale, y2 * scale);
+    }
+
+
+    public boolean horizontallyAligned(Box other, double tol) {
+
+        return !(other.x1 - tol > x2 || other.x2 + tol < x1);
+    }
+
+
+    public double yDistanceTo(Box other) {
+
+        return Math.min(Math.abs(other.y1 - y2), Math.abs(y2 - other.y1));
+    }
+
+
+    public boolean intersects(Box other, double tol) {
+
+        return !((x2 < other.x1 - tol) || (x1 > other.x2 + tol) || (y2 < other.y1 - tol) || (y1 > other.y2 + tol));
+    }
+
+
+    public boolean intersectsAndOver(Box other, double tol) {
+
+        return (!((x2 < other.x1 - tol) || (x1 > other.x2 + tol) || (y2 < other.y1 - tol) || (y1 > other.y2 + tol))) && other.y1 > y1;
+    }
+
+
+    public boolean intersectsCenter(Box other, double tol) {
+
+        return !((x2 < other.xCenter() - tol) || (x1 > other.xCenter() + tol) || (y2 < other.yCenter() - tol) || (y1 > other.yCenter() + tol));
+    }
+
+
+    public Optional<Box> intersectRegion(Box other, double tol) {
+
+        if (!intersects(other, tol)) {
+            return Optional.empty();
+        } else {
+            var overlapX1 = Math.max(x1, other.x1);
+            var overlapY1 = Math.max(y1, other.y1);
+            var overlapX2 = Math.min(x2, other.x2);
+            var overlapY2 = Math.min(y2, other.y2);
+            return Optional.of(new Box(overlapX1, overlapY1, overlapX2, overlapY2));
+        }
+    }
+
+
+    public double intersectArea(Box other, double tol) {
+
+        return intersectRegion(other, tol).map(Box::area).orElse(0d);
+    }
+
+
+    public boolean intersectsAny(List<Box> others, double tol) {
+
+        return others.stream().anyMatch(other -> intersects(other, tol));
+    }
+
+
+    public boolean intersectsAnyAndOver(List<Box> others, double tol) {
+
+        return others.stream().anyMatch(other -> intersectsAndOver(other, tol));
+    }
+
+
+    public boolean intersectsCenter(List<Box> others, double tol) {
+
+        return others.stream().anyMatch(other -> intersectsCenter(other, tol));
+    }
+
+
+    public boolean contains(Box other, double tol) {
+
+        return (x1 <= other.x1 + tol) && (y1 <= other.y1 + tol) && (x2 >= other.x2 - tol) && (y2 >= other.y2 - tol);
+    }
+
+
+    public Box container(Box other) {
+
+        var minX = Math.min(x1, other.x1);
+        var minY = Math.min(y1, other.y1);
+        var maxX = Math.max(x2, other.x2);
+        var maxY = Math.max(y2, other.y2);
+        return new Box(minX, minY, maxX, maxY);
+    }
+
+
+    public Box transform(AffineTransform affineTransform) {
+
+        Point2D point = affineTransform.transform(new Point2D.Double(x1, y1), null);
+        Point2D point2 = affineTransform.transform(new Point2D.Double(x2, y2), null);
+        return new Box(Math.min(point.getX(), point2.getX()), Math.min(point.getY(), point2.getY()), Math.max(point.getX(), point2.getX()), Math.max(point.getY(), point2.getY()));
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/graphics/DistinctQueue.java

@@ -0,0 +1,51 @@
+package com.knecon.fforesight.service.layoutparser.processor.services.graphics;
+
+import java.util.HashSet;
+import java.util.LinkedList;
+import java.util.Queue;
+import java.util.Set;
+
+public class DistinctQueue<T> {
+
+    private Queue<T> queue;
+    private Set<T> set;
+
+
+    public DistinctQueue() {
+
+        queue = new LinkedList<>();
+        set = new HashSet<>();
+    }
+
+
+    public void enqueue(T element) {
+
+        if (!set.contains(element)) {
+            queue.add(element);
+            set.add(element);
+        }
+    }
+
+
+    public T dequeue() {
+
+        T element = queue.poll();
+        if (element != null) {
+            set.remove(element);
+        }
+        return element;
+    }
+
+
+    public boolean isEmpty() {
+
+        return queue.isEmpty();
+    }
+
+
+    public int size() {
+
+        return queue.size();
+    }
+    // Other methods as needed
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/graphics/FindGraphicsRaster.java

@@ -0,0 +1,172 @@
+package com.knecon.fforesight.service.layoutparser.processor.services.graphics;
+
+import java.awt.geom.AffineTransform;
+import java.awt.geom.Rectangle2D;
+import java.awt.image.BufferedImage;
+import java.awt.image.DataBufferByte;
+import java.util.ArrayList;
+import java.util.List;
+import java.util.stream.Collectors;
+
+import org.apache.pdfbox.pdmodel.PDDocument;
+import org.apache.pdfbox.rendering.ImageType;
+import org.apache.pdfbox.rendering.PDFRenderer;
+import org.springframework.stereotype.Service;
+
+import lombok.SneakyThrows;
+
+@Service
+public class FindGraphicsRaster {
+
+    // Pixels that are lighter then this threshold are ignored
+    private final static int THRESHOLD = 240;
+
+    // DPI to render the image at, in practice sub-72 seems to risk pixels being lost
+    private final static int DPI = 72;
+
+
+    @SneakyThrows
+    public List<Box> findCCBoundingBoxes(PDDocument doc, List<Rectangle2D> remove, PageInformation pageInformation) {
+
+        var renderer = new PDFRenderer(doc);
+        var img = renderer.renderImageWithDPI(pageInformation.number() - 1, DPI, ImageType.GRAY);
+        var imageCtm = getImageCTM(pageInformation, img.getWidth());
+        return findCCBoundingBoxes(img, remove, THRESHOLD, DPI / 72, imageCtm);
+    }
+
+
+    @SneakyThrows
+    private List<Box> findCCBoundingBoxes(BufferedImage image, List<Rectangle2D> remove, int grayScaleTresh, int rescale, AffineTransform imageCTM) {
+
+        var inverseCTM = imageCTM.createInverse();
+
+        var h = image.getHeight();
+        var w = image.getWidth();
+        var pixels = new int[w * h];
+        image.getRaster().getPixels(0, 0, w, h, pixels);
+        remove.stream().map(rect -> inverseCTM.createTransformedShape(rect).getBounds2D()).forEach(box -> {
+            for (int y = (int) Math.floor(box.getMinY() / rescale); y <= (int) Math.min(Math.ceil(box.getMaxY() / rescale), h); y++) {
+                for (int x = (int) Math.floor(box.getMinX() / rescale); x <= (int) Math.min(Math.ceil(box.getMaxX() / rescale), w); x++) {
+                    pixels[w * y + x] = grayScaleTresh;
+                }
+            }
+        });
+
+//        var image2 = createImageFromMatrix(pixels, w, h);
+
+        return findCCBoundingBoxes(pixels, w, h, grayScaleTresh, rescale, imageCTM);
+    }
+
+
+    public static BufferedImage createImageFromMatrix(int[] matrix, int width, int height) {
+
+        BufferedImage image = new BufferedImage(width, height, BufferedImage.TYPE_BYTE_GRAY);
+
+        byte[] pixelData = ((DataBufferByte) image.getRaster().getDataBuffer()).getData();
+
+        for (int y = 0; y < height; y++) {
+            for (int x = 0; x < width; x++) {
+                int index = y * width + x;
+                int pixel = matrix[index]; // Assuming each element in the matrix represents a pixel color
+                pixelData[index] = (byte) pixel;
+            }
+        }
+
+        return image;
+    }
+
+
+    private List<Box> findCCBoundingBoxes(int[] pixels, int w, int h, int pixThreshold, int rescale, AffineTransform imageCTM) {
+
+        DistinctQueue<Integer> pixelsToExplore = new DistinctQueue<>();
+        var boundingBoxes = new ArrayList<Box>();
+
+        for (int y = 0; y < h; y++) {
+            for (int x = 0; x < w; x++) {
+                var pixelIndex = x + y * w;
+                if (pixels[pixelIndex] < pixThreshold) {
+                    var minX = x;
+                    var maxX = x;
+                    var minY = y;
+                    var maxY = y;
+                    pixelsToExplore.enqueue(pixelIndex);
+                    while (!pixelsToExplore.isEmpty()) {
+                        var currentPixel = pixelsToExplore.dequeue();
+                        if (currentPixel > w) {
+                            var lowerPixel = currentPixel - w;
+                            if (pixels[lowerPixel] < pixThreshold) {
+                                pixelsToExplore.enqueue(currentPixel - w);
+                                minY = Math.min(minY, lowerPixel / w);
+                            }
+                        }
+                        if (currentPixel < pixels.length - w) {
+                            var upperPixel = currentPixel + w;
+                            if (pixels[upperPixel] < pixThreshold) {
+                                pixelsToExplore.enqueue(upperPixel);
+                                maxY = Math.max(maxY, upperPixel / w);
+                            }
+                        }
+                        if (currentPixel % w != 0) {
+                            var leftPixel = currentPixel - 1;
+                            if (pixels[leftPixel] < pixThreshold) {
+                                pixelsToExplore.enqueue(leftPixel);
+                                minX = Math.min(minX, leftPixel % w);
+                            }
+                        }
+                        if ((currentPixel + 1) % w != 0) {
+                            var rightPixel = currentPixel + 1;
+                            if (pixels[rightPixel] < pixThreshold) {
+                                pixelsToExplore.enqueue(rightPixel + 1);
+                                maxX = Math.max(maxX, rightPixel % w);
+                            }
+                        }
+                        // Set the current pixel to white so we don't visit it again.
+                        pixels[currentPixel] = pixThreshold;
+                    }
+                    boundingBoxes.add(new Box(minX * rescale, minY * rescale, maxX * rescale, maxY * rescale));
+                }
+            }
+        }
+        return boundingBoxes.stream().filter(box -> box.area() > 0).map(box -> box.transform(imageCTM)).collect(Collectors.toList());
+    }
+
+
+    public AffineTransform getImageCTM(PageInformation pageInformation, int imageWidth) {
+
+        double scalingFactor = calculateScalingFactor(pageInformation, imageWidth);
+        AffineTransform imageToCropBoxScaling = new AffineTransform(scalingFactor, 0, 0, scalingFactor, -pageInformation.minX(), -pageInformation.minY());
+
+        AffineTransform mirrorMatrix = new AffineTransform(1, 0, 0, -1, 0, pageInformation.height());
+
+        AffineTransform rotationMatrix = switch (pageInformation.rotationDegrees()) {
+            case 90 -> new AffineTransform(0, 1, -1, 0, pageInformation.height(), 0);
+            case 180 -> new AffineTransform(-1, 0, 0, -1, pageInformation.width(), pageInformation.height());
+            case 270 -> new AffineTransform(0, -1, 1, 0, pageInformation.width() - pageInformation.height(), pageInformation.height()); // results from 90 + 180 rotations
+            default -> new AffineTransform();
+        };
+
+        // matrix multiplication is performed from right to left, so the order is reversed.
+        // scaling -> mirror -> rotation
+        AffineTransform resultMatrix = new AffineTransform();
+
+        resultMatrix.concatenate(rotationMatrix);
+        resultMatrix.concatenate(mirrorMatrix);
+        resultMatrix.concatenate(imageToCropBoxScaling);
+        return resultMatrix;
+    }
+
+
+    private double calculateScalingFactor(PageInformation pageInformation, int imageWidth) {
+
+        // PDFBox always returns page height and width based on rotation
+        double pageWidth;
+        if (pageInformation.rotationDegrees() == 90 || pageInformation.rotationDegrees() == 270) {
+            pageWidth = pageInformation.height();
+        } else {
+            pageWidth = pageInformation.width();
+        }
+
+        return pageWidth / imageWidth;
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/graphics/GraphicBBDetector.java

@@ -0,0 +1,247 @@
+package com.knecon.fforesight.service.layoutparser.processor.services.graphics;
+
+import java.awt.Rectangle;
+import java.awt.color.CMMException;
+import java.awt.geom.GeneralPath;
+import java.awt.geom.Point2D;
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.List;
+import java.util.stream.Collectors;
+
+import org.apache.pdfbox.contentstream.PDFGraphicsStreamEngine;
+import org.apache.pdfbox.contentstream.PDFStreamEngine;
+import org.apache.pdfbox.contentstream.operator.Operator;
+import org.apache.pdfbox.contentstream.operator.OperatorProcessor;
+import org.apache.pdfbox.cos.COSBase;
+import org.apache.pdfbox.cos.COSName;
+import org.apache.pdfbox.pdmodel.PDPage;
+import org.apache.pdfbox.pdmodel.graphics.color.PDColor;
+import org.apache.pdfbox.pdmodel.graphics.image.PDImage;
+
+import lombok.SneakyThrows;
+import lombok.extern.slf4j.Slf4j;
+
+@Slf4j
+public class GraphicBBDetector extends PDFGraphicsStreamEngine {
+
+    private int clipWindingRule = -1;
+    private final GeneralPath linePath = new GeneralPath();
+    private final List<Rectangle> bounds = new ArrayList<>();
+    private final PDColor emptyPattern = new PDColor(new float[3], null);
+
+
+    public GraphicBBDetector(PDPage page, boolean ignoreWhite) {
+
+        super(page);
+
+        if (!ignoreWhite) {
+            addOperator(new NullOp("d", this));
+            addOperator(new NullOp("k", this));
+            addOperator(new NullOp("K", this));
+            addOperator(new NullOp("g", this));
+            addOperator(new NullOp("G", this));
+            addOperator(new NullOp("CS", this));
+            addOperator(new NullOp("cs", this));
+            addOperator(new NullOp("RG", this));
+            addOperator(new NullOp("rg", this));
+            addOperator(new NullOp("sc", this));
+            addOperator(new NullOp("SC", this));
+            addOperator(new NullOp("scn", this));
+            addOperator(new NullOp("SCN", this));
+        }
+
+        // Ignore text and font ops:
+        addOperator(new NullOp("Tf", this));
+        addOperator(new NullOp("Tj", this));
+        addOperator(new NullOp("TJ", this));
+        addOperator(new NullOp("T*", this));
+        addOperator(new NullOp("'", this));
+        addOperator(new NullOp("\"", this));
+    }
+
+
+    public List<Box> findGraphicBB() throws IOException {
+
+        processPage(getPage());
+        return bounds.stream().map(r -> new Box(r.x, r.y, r.x + r.width, r.y + r.height)).filter(box -> box.area() > 0).collect(Collectors.toList());
+    }
+
+
+    @Override
+    public void appendRectangle(Point2D p0, Point2D p1, Point2D p2, Point2D p3) {
+
+        linePath.moveTo((float) p0.getX(), (float) p0.getY());
+        linePath.lineTo((float) p1.getX(), (float) p1.getY());
+        linePath.lineTo((float) p2.getX(), (float) p2.getY());
+        linePath.lineTo((float) p3.getX(), (float) p3.getY());
+        linePath.closePath();
+    }
+
+
+    @Override
+    public void drawImage(PDImage pdImage) {
+        // Do nothing
+    }
+
+
+    @Override
+    public void clip(int windingRule) {
+
+        clipWindingRule = windingRule;
+    }
+
+
+    @Override
+    public void moveTo(float x, float y) {
+
+        linePath.moveTo(x, y);
+    }
+
+
+    @Override
+    public void lineTo(float x, float y) {
+
+        linePath.lineTo(x, y);
+    }
+
+
+    @Override
+    public void curveTo(float x1, float y1, float x2, float y2, float x3, float y3) {
+
+        linePath.curveTo(x1, y1, x2, y2, x3, y3);
+    }
+
+
+    @Override
+    public Point2D getCurrentPoint() {
+
+        return linePath.getCurrentPoint();
+    }
+
+
+    @Override
+    public void closePath() {
+
+        linePath.closePath();
+    }
+
+
+    @Override
+    public void endPath() {
+
+        if (clipWindingRule != -1) {
+            linePath.setWindingRule(clipWindingRule);
+            getGraphicsState().intersectClippingPath(linePath);
+            clipWindingRule = -1;
+        }
+        linePath.reset();
+    }
+
+
+    @Override
+    public void strokePath() throws IOException {
+
+        addLinePath(true, false);
+        linePath.reset();
+    }
+
+
+    @Override
+    public void fillPath(int windingRule) throws IOException {
+
+        linePath.setWindingRule(windingRule);
+        addLinePath(false, true);
+        linePath.reset();
+    }
+
+
+    @Override
+    public void fillAndStrokePath(int windingRule) throws IOException {
+
+        linePath.setWindingRule(windingRule);
+        addLinePath(true, true);
+        linePath.reset();
+    }
+
+
+    @Override
+    public void shadingFill(COSName shadingName) {
+
+        var newBound = getGraphicsState().getCurrentClippingPath().getBounds();
+        if (newBound.getWidth() > 0 && newBound.getHeight() > 0) {
+            bounds.add(newBound);
+        }
+    }
+
+
+    private void addLinePath(boolean stroke, boolean fill) throws IOException {
+
+        var newBound = getGraphicsState().getCurrentClippingPath().getBounds().intersection(linePath.getBounds());
+        if (newBound.getWidth() > 0 && newBound.getHeight() > 0) {
+
+            if (stroke && !getGraphicsState().getStrokingColor().isPattern() && isBlack(getGraphicsState().getStrokingColor()) || //
+                    !stroke && !getGraphicsState().getNonStrokingColor().isPattern() && isBlack(getGraphicsState().getNonStrokingColor())) {
+                bounds.add(newBound);
+            }
+
+//            var skipWhiteGraphic = ignoreWhite && (!stroke || isWhite(getGraphicsState().getStrokingColor())) && (!fill || isWhite(getGraphicsState().getNonStrokingColor()));
+//            if (!skipWhiteGraphic) {
+//                bounds.add(newBound);
+//            }
+        }
+    }
+
+
+    @SneakyThrows
+    private boolean isBlack(PDColor color) {
+
+        try {
+            return color.toRGB() == 0;
+        } catch (CMMException e) {
+            // see https://github.com/haraldk/TwelveMonkeys/issues/124 or https://issues.apache.org/jira/browse/PDFBOX-3531
+            // This is a quick and dirt hack
+            // Happens for file 216.pdf
+            log.debug(e.getMessage());
+            var result = true;
+            for (var component : color.getComponents()) {
+                result = result && component == 0;
+            }
+            return result;
+        }
+    }
+
+
+    private boolean isWhite(PDColor color) throws IOException {
+
+        return !color.isPattern() && color.toRGB() == 16777215 || color.equals(emptyPattern);
+    }
+
+
+    private final class NullOp extends OperatorProcessor {
+
+        private final String name;
+
+
+        private NullOp(String name, PDFStreamEngine context) {
+
+            super(context);
+            this.name = name;
+        }
+
+
+        @Override
+        public void process(Operator operator, List<COSBase> operands) {
+            // Do nothing.
+        }
+
+
+        @Override
+        public String getName() {
+
+            return name;
+        }
+
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/graphics/GraphicExtractorService.java

@@ -0,0 +1,107 @@
+package com.knecon.fforesight.service.layoutparser.processor.services.graphics;
+
+import java.awt.geom.Rectangle2D;
+import java.util.ArrayList;
+import java.util.List;
+import java.util.stream.Collectors;
+
+import org.apache.pdfbox.pdmodel.PDDocument;
+import org.apache.pdfbox.pdmodel.PDPage;
+import org.springframework.stereotype.Service;
+
+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.TextPositionSequence;
+import com.knecon.fforesight.service.layoutparser.processor.services.factory.SearchTextWithTextPositionFactory;
+import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
+
+import lombok.RequiredArgsConstructor;
+import lombok.SneakyThrows;
+
+@Service
+@RequiredArgsConstructor
+public class GraphicExtractorService {
+
+    private final GraphicsClusteringService graphicsClusteringService;
+    private final FindGraphicsRaster findGraphicsRaster;
+
+
+    @SneakyThrows
+    public List<Box> extractPathElementGraphics(PDDocument pdDocument,
+                                                PDPage pdPage,
+                                                int pageNumber,
+                                                CleanRulings cleanRulings,
+                                                List<TextPositionSequence> textPositionSequences,
+                                                List<Cell> emptyTableCells,
+                                                boolean graphicsRaster) {
+
+        var characterBBoxes = getCharacterBBoxes(textPositionSequences);
+        var tableLineBBoxes = getLineBBoxesFromTableCells(emptyTableCells);
+        var underLineBBoxes = getUnderlineBBoxes(cleanRulings, characterBBoxes);
+        var strikeThroughBBoxes = getStrikeThroughBBoxes(cleanRulings, characterBBoxes);
+
+        GraphicBBDetector graphicBBDetector = new GraphicBBDetector(pdPage, true);
+        var graphicBBoxes = graphicBBDetector.findGraphicBB();
+
+        if (graphicsRaster) {
+            // This should only be used if ocr was performed, it is currently in an early stage and needs to be improved.
+            graphicBBoxes.addAll(findGraphicsRaster.findCCBoundingBoxes(pdDocument,
+                    characterBBoxes.stream().map(box -> new Rectangle2D.Double(box.x1 - 2, box.y1 - 2, box.width() + 4, box.height() + 4)).collect(Collectors.toList()),
+                    PageInformation.fromPDPage(pageNumber, pdPage)));
+        }
+
+        var filteredGraphicBBoxes = graphicBBoxes.stream()
+                .filter(box -> !box.intersectsAny(tableLineBBoxes, 4))
+                .filter(box -> !box.intersectsAny(underLineBBoxes, 4))
+                .filter(box -> !box.intersectsAny(strikeThroughBBoxes, 4))
+                .collect(Collectors.toList());
+
+        var clusters = graphicsClusteringService.getClusters(filteredGraphicBBoxes, 14);
+
+        return clusters.stream().filter(box -> box.area() > 500 && box.height() > 50 && box.width() > 50).toList();
+    }
+
+
+    private List<Box> getCharacterBBoxes(List<TextPositionSequence> textPositionSequences) {
+
+        return textPositionSequences.stream()
+                .map(pos -> pos.getTextPositions()
+                        .stream()
+                        .map(tp -> SearchTextWithTextPositionFactory.mapRedTextPositionToInitialUserSpace(tp, pos))
+                        .collect(RectangleTransformations.collectBBox()))
+                .map(Box::new)
+                .collect(Collectors.toList());
+    }
+
+
+    private List<Box> getLineBBoxesFromTableCells(List<Cell> emptyTableCells) {
+
+        List<Box> expandedTableLines = new ArrayList<>();
+
+        emptyTableCells.forEach(cell -> {
+            expandedTableLines.add(new Box(new Rectangle2D.Double(cell.x, cell.y - 1, cell.width, 2)));
+            expandedTableLines.add(new Box(new Rectangle2D.Double(cell.x, cell.y + cell.height - 1, cell.width, 2)));
+            expandedTableLines.add(new Box(new Rectangle2D.Double(cell.x - 1, cell.y, 2, cell.height)));
+            expandedTableLines.add(new Box(new Rectangle2D.Double(cell.x + cell.width - 1, cell.y, 2, cell.height)));
+        });
+
+        return expandedTableLines;
+    }
+
+
+    private List<Box> getUnderlineBBoxes(CleanRulings cleanRulings, List<Box> characterBBoxes) {
+
+        return cleanRulings.getHorizontal()
+                .stream()
+                .map(h -> new Box(h.x1, h.y1, h.x2, h.y2))
+                .filter(box -> box.intersectsAnyAndOver(characterBBoxes, 6))
+                .collect(Collectors.toList());
+    }
+
+
+    private List<Box> getStrikeThroughBBoxes(CleanRulings cleanRulings, List<Box> characterBBoxes) {
+
+        return cleanRulings.getHorizontal().stream().map(h -> new Box(h.x1, h.y1, h.x2, h.y2)).filter(box -> box.intersectsCenter(characterBBoxes, 2)).collect(Collectors.toList());
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/graphics/GraphicsClusteringService.java

@@ -0,0 +1,83 @@
+package com.knecon.fforesight.service.layoutparser.processor.services.graphics;
+
+import java.util.ArrayList;
+import java.util.List;
+
+import org.springframework.stereotype.Service;
+
+@Service
+public class GraphicsClusteringService {
+
+    public List<Box> getClusters(List<Box> boxes, double tol) {
+
+        if (boxes.isEmpty()) {
+            return boxes;
+        } else {
+            // We iteratively pick a Box that intersects at least one other box and replace the
+            // intersecting box with a Box containing them
+            var currentBoxes = boxes;
+            var foundIntersectingBoxes = true;
+            while (foundIntersectingBoxes) {
+                foundIntersectingBoxes = false;
+
+                // The box we are going to check to see if there are any intersecting boxes, followed by
+                // any boxes that we have already check
+                var checked = List.of(currentBoxes.get(0));
+                var unchecked = currentBoxes.subList(1, currentBoxes.size());
+
+                while (!foundIntersectingBoxes && !unchecked.isEmpty()) {
+
+                    List<Box> intersects = new ArrayList<>();
+                    List<Box> nonIntersects = new ArrayList<>();
+
+                    for (Box uncheck : unchecked) {
+                        if (uncheck.intersects(checked.get(0), tol)) {
+                            intersects.add(uncheck);
+                        } else {
+                            nonIntersects.add(uncheck);
+                        }
+                    }
+
+                    if (!intersects.isEmpty()) {
+                        List<Box> combinedIntersecting = new ArrayList<>();
+                        combinedIntersecting.add(checked.get(0));
+                        combinedIntersecting.addAll(intersects);
+                        var newBox = merge(combinedIntersecting);
+
+                        List<Box> newCurrentBoxes = new ArrayList<>();
+                        newCurrentBoxes.add(newBox);
+                        newCurrentBoxes.addAll(checked.subList(1, checked.size()));
+                        newCurrentBoxes.addAll(nonIntersects);
+                        currentBoxes = newCurrentBoxes;
+                        foundIntersectingBoxes = true; // Exit this loop and re-enter the outer loop
+                    } else {
+                        List<Box> newChecked = new ArrayList<>();
+                        newChecked.add(unchecked.get(0));
+                        newChecked.addAll(checked);
+                        checked = newChecked;
+                        unchecked = unchecked.subList(1, unchecked.size());
+                    }
+                }
+            }
+            return currentBoxes;
+        }
+    }
+
+
+    public Box merge(List<Box> boxes) {
+
+        double minX = Double.POSITIVE_INFINITY;
+        double minY = Double.POSITIVE_INFINITY;
+        double maxX = Double.NEGATIVE_INFINITY;
+        double maxY = Double.NEGATIVE_INFINITY;
+
+        for (Box box : boxes) {
+            minX = Math.min(minX, box.x1);
+            minY = Math.min(minY, box.y1);
+            maxX = Math.max(maxX, box.x2);
+            maxY = Math.max(maxY, box.y2);
+        }
+        return new Box(minX, minY, maxX, maxY);
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/graphics/PageInformation.java

@@ -0,0 +1,42 @@
+package com.knecon.fforesight.service.layoutparser.processor.services.graphics;
+
+import java.awt.geom.Rectangle2D;
+
+import org.apache.pdfbox.pdmodel.PDPage;
+import org.apache.pdfbox.pdmodel.common.PDRectangle;
+
+public record PageInformation(Rectangle2D mediabox, int number, int rotationDegrees) {
+
+    public static PageInformation fromPDPage(int pageNum, PDPage page) {
+
+        PDRectangle mediaBox = page.getMediaBox();
+        return new PageInformation(new Rectangle2D.Double(mediaBox.getLowerLeftX(), mediaBox.getLowerLeftY(), mediaBox.getWidth(), mediaBox.getHeight()),
+                pageNum,
+                page.getRotation());
+    }
+
+
+    public double height() {
+
+        return mediabox.getHeight();
+    }
+
+
+    public double width() {
+
+        return mediabox.getWidth();
+    }
+
+
+    public double minX() {
+
+        return mediabox.getX();
+    }
+
+
+    public double minY() {
+
+        return mediabox.getY();
+    }
+
+}

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/mapper/DocumentDataMapper.java

@@ -2,6 +2,7 @@ package com.knecon.fforesight.service.layoutparser.processor.services.mapper;
 
 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;
@@ -12,8 +13,10 @@ import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.Do
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentPositionData;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentStructure;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentTextData;
+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.nodes.Document;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.DuplicatedParagraph;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Image;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Page;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Table;
@@ -73,17 +76,24 @@ public class DocumentDataMapper {
             case TABLE -> PropertiesMapper.buildTableProperties((Table) entry.getNode());
             case TABLE_CELL -> PropertiesMapper.buildTableCellProperties((TableCell) entry.getNode());
             case IMAGE -> PropertiesMapper.buildImageProperties((Image) entry.getNode());
+            case PARAGRAPH ->
+                    entry.getNode() instanceof DuplicatedParagraph duplicatedParagraph ? PropertiesMapper.buildDuplicateParagraphProperties(duplicatedParagraph) : new HashMap<>();
             default -> new HashMap<>();
         };
 
-        return DocumentStructure.EntryData.builder()
+        DocumentStructure.EntryData.EntryDataBuilder documentBuilder = DocumentStructure.EntryData.builder()
                 .treeId(toPrimitiveIntArray(entry.getTreeId()))
                 .children(entry.getChildren().stream().map(DocumentDataMapper::toEntryData).toList())
                 .type(entry.getType())
                 .atomicBlockIds(atomicTextBlocks)
                 .pageNumbers(entry.getNode().getPages().stream().map(Page::getNumber).map(Integer::longValue).toArray(Long[]::new))
-                .properties(properties)
-                .build();
+                .properties(properties);
+        if (entry.getNode() != null) {
+            documentBuilder.engines(entry.getNode().getEngines());
+        } else {
+            documentBuilder.engines(new HashSet<>(Set.of(LayoutEngine.ALGORITHM)));
+        }
+        return documentBuilder.build();
     }
 
 

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/mapper/DocumentGraphMapper.java

@@ -7,13 +7,14 @@ import java.util.List;
 import java.util.Map;
 import java.util.NoSuchElementException;
 
-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.internal.api.data.redaction.DocumentData;
-import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentStructure;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentPage;
+import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentPositionData;
+import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentStructure;
+import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentTextData;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.DuplicatedParagraph;
 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;
@@ -61,7 +62,7 @@ public class DocumentGraphMapper {
 
             SemanticNode node = switch (entryData.getType()) {
                 case SECTION -> buildSection(context);
-                case PARAGRAPH -> buildParagraph(context);
+                case PARAGRAPH -> buildParagraph(context, entryData.getProperties());
                 case HEADLINE -> buildHeadline(context);
                 case HEADER -> buildHeader(context);
                 case FOOTER -> buildFooter(context);
@@ -140,7 +141,17 @@ public class DocumentGraphMapper {
     }
 
 
-    private Paragraph buildParagraph(Context context) {
+    private Paragraph buildParagraph(Context context, Map<String, String> properties) {
+
+        if (PropertiesMapper.isDuplicateParagraph(properties)) {
+
+            DuplicatedParagraph duplicatedParagraph = DuplicatedParagraph.builder().documentTree(context.documentTree).build();
+
+            Long[] unsortedTextblockIds = PropertiesMapper.getUnsortedTextblockIds(properties);
+            duplicatedParagraph.setUnsortedLeafTextBlock(toTextBlock(unsortedTextblockIds, context, duplicatedParagraph));
+            return duplicatedParagraph;
+
+        }
 
         return Paragraph.builder().documentTree(context.documentTree).build();
     }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/mapper/PropertiesMapper.java

@@ -1,17 +1,19 @@
 package com.knecon.fforesight.service.layoutparser.processor.services.mapper;
 
 import java.awt.geom.Rectangle2D;
-import java.util.Collections;
+import java.util.Arrays;
 import java.util.HashMap;
 import java.util.Locale;
 import java.util.Map;
 
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentStructure;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.DuplicatedParagraph;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Image;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.ImageType;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Page;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Table;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.TableCell;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.AtomicTextBlock;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
 
 public class PropertiesMapper {
 
@@ -76,6 +78,32 @@ public class PropertiesMapper {
     }
 
 
+    public static Map<String, String> buildDuplicateParagraphProperties(DuplicatedParagraph duplicatedParagraph) {
+
+        Map<String, String> properties = new HashMap<>();
+        properties.put(DocumentStructure.DuplicateParagraphProperties.UNSORTED_TEXTBLOCK_ID, Arrays.toString(toAtomicTextBlockIds(duplicatedParagraph.getUnsortedLeafTextBlock())));
+        return properties;
+    }
+
+
+    public static boolean isDuplicateParagraph(Map<String, String> properties) {
+
+        return properties.containsKey(DocumentStructure.DuplicateParagraphProperties.UNSORTED_TEXTBLOCK_ID);
+    }
+
+
+    public static Long[] getUnsortedTextblockIds(Map<String, String> properties) {
+
+        return toLongArray(properties.get(DocumentStructure.DuplicateParagraphProperties.UNSORTED_TEXTBLOCK_ID));
+    }
+
+
+    public static Long[] toLongArray(String ids) {
+
+        return Arrays.stream(ids.substring(1, ids.length() - 1).trim().split(",")).map(Long::valueOf).toArray(Long[]::new);
+    }
+
+
     private static ImageType parseImageType(String imageType) {
 
         return switch (imageType) {
@@ -101,4 +129,10 @@ public class PropertiesMapper {
                 rectangle2D.getHeight());
     }
 
+
+    private static Long[] toAtomicTextBlockIds(TextBlock textBlock) {
+
+        return textBlock.getAtomicTextBlocks().stream().map(AtomicTextBlock::getId).toArray(Long[]::new);
+    }
+
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/parsing/PDFLinesTextStripper.java

@@ -237,8 +237,13 @@ public class PDFLinesTextStripper extends PDFTextStripper {
         int startIndex = 0;
         RedTextPosition previous = null;
 
+        float direction = -1;
         for (int i = 0; i <= textPositions.size() - 1; i++) {
 
+            if (direction == -1) {
+                direction = textPositions.get(i).getDir();
+            }
+
             if (!textPositionSequences.isEmpty()) {
                 previous = textPositionSequences.get(textPositionSequences.size() - 1)
                         .getTextPositions()
@@ -250,6 +255,13 @@ public class PDFLinesTextStripper extends PDFTextStripper {
                 continue;
             }
 
+            if (textPositions.get(i).getDir() != direction && startIndex != i) {
+                List<TextPosition> sublist = textPositions.subList(startIndex, i);
+                textPositionSequences.add(new TextPositionSequence(sublist, pageNumber, i == textPositions.size() - 1 && isParagraphStart));
+                startIndex = i;
+                direction = textPositions.get(i).getDir();
+            }
+
             // Strange but sometimes this is happening, for example: Metolachlor2.pdf
             if (checkIfCurrentPositionIsToTheRightOfPreviousPosition(i, textPositions)) {
                 List<TextPosition> sublist = textPositions.subList(startIndex, i);
@@ -329,6 +341,7 @@ public class PDFLinesTextStripper extends PDFTextStripper {
                 .getXDirAdj() - (previous.getXDirAdj() + previous.getWidthDirAdj()) < maximumGapSize;
     }
 
+
     @Override
     public String getText(PDDocument doc) throws IOException {
 

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/visualization/LayoutGridService.java

@@ -1,12 +1,11 @@
 package com.knecon.fforesight.service.layoutparser.processor.services.visualization;
 
-import static com.knecon.fforesight.service.layoutparser.processor.services.visualization.ViewerDocumentService.LINE_WIDTH;
-
 import java.awt.Color;
 import java.awt.geom.Line2D;
 import java.awt.geom.Point2D;
 import java.awt.geom.Rectangle2D;
 import java.awt.geom.RectangularShape;
+import java.io.File;
 import java.util.ArrayList;
 import java.util.Collections;
 import java.util.Comparator;
@@ -18,55 +17,108 @@ import java.util.stream.Stream;
 
 import org.springframework.stereotype.Service;
 
+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.nodes.Document;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.DuplicatedParagraph;
 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.model.graph.nodes.Table;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.TableCell;
-import com.knecon.fforesight.service.layoutparser.processor.model.visualization.ColoredLine;
-import com.knecon.fforesight.service.layoutparser.processor.model.visualization.ColoredRectangle;
-import com.knecon.fforesight.service.layoutparser.processor.model.visualization.LayoutGrid;
-import com.knecon.fforesight.service.layoutparser.processor.model.visualization.PlacedText;
 import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
+import com.knecon.fforesight.service.viewerdoc.ContentStreams;
+import com.knecon.fforesight.service.viewerdoc.model.ColoredLine;
+import com.knecon.fforesight.service.viewerdoc.model.ColoredRectangle;
+import com.knecon.fforesight.service.viewerdoc.model.LayoutGrid;
+import com.knecon.fforesight.service.viewerdoc.model.PlacedText;
+import com.knecon.fforesight.service.viewerdoc.model.Standard14EmbeddableFont;
+import com.knecon.fforesight.service.viewerdoc.model.Visualizations;
+import com.knecon.fforesight.service.viewerdoc.service.ViewerDocumentService;
+
+import io.micrometer.observation.annotation.Observed;
+import lombok.AccessLevel;
+import lombok.RequiredArgsConstructor;
+import lombok.SneakyThrows;
+import lombok.experimental.FieldDefaults;
 
 @Service
+@RequiredArgsConstructor
+@FieldDefaults(level = AccessLevel.PRIVATE, makeFinal = true)
 public class LayoutGridService {
 
-    private static final Color INNER_LINES_COLOR = new Color(255, 175, 175);
-    private static final Color PARAGRAPH_COLOR = new Color(70, 130, 180);
-    public static final Color TABLE_COLOR = new Color(102, 205, 170);
-    public static final Color SECTION_COLOR = new Color(50, 50, 50);
-    public static final Color HEADLINE_COLOR = new Color(162, 56, 56);
-    public static final Color HEADER_COLOR = new Color(171, 131, 6);
-    public static final Color IMAGE_COLOR = new Color(253, 63, 146);
+    ViewerDocumentService viewerDocumentService;
+
+    static float FONT_SIZE = 10f;
+    static float LINE_WIDTH = 1f;
+    static Standard14EmbeddableFont FONT = Standard14EmbeddableFont.helvetica();
+
+    static Color INNER_LINES_COLOR = new Color(255, 175, 175);
+    static Color PARAGRAPH_COLOR = new Color(70, 130, 180);
+
+    static Color DUPLICATE_PARAGRAPH_COLOR = new Color(70, 180, 101);
+    static Color TABLE_COLOR = new Color(102, 205, 170);
+    static Color SECTION_COLOR = new Color(50, 50, 50);
+    static Color HEADLINE_COLOR = new Color(162, 56, 56);
+    static Color HEADER_COLOR = new Color(171, 131, 6);
+    static Color IMAGE_COLOR = new Color(253, 63, 146);
 
 
-    public LayoutGrid createLayoutGrid(Document document) {
+    @SneakyThrows
+    @Observed(name = "ViewerDocumentService", contextualName = "create-viewer-document")
+    public void addLayoutGrid(File originFile, Document document, File destinationFile, boolean layerVisibilityDefaultValue) {
+
+        Visualizations layoutGrid = this.addLayoutGrid(document, layerVisibilityDefaultValue, false);
+        Visualizations visualLayoutGrid = this.addLayoutGrid(document, layerVisibilityDefaultValue, true);
+
+        viewerDocumentService.addVisualizationsOnPage(originFile, destinationFile, List.of(layoutGrid, visualLayoutGrid));
+    }
+
+
+    @SneakyThrows
+    @Observed(name = "ViewerDocumentService", contextualName = "create-viewer-document")
+    public Visualizations addLayoutGrid(Document document, boolean layerVisibilityDefaultValue, boolean visualParsingGrid) {
+
+        LayoutGrid layoutGrid = createLayoutGrid(document, visualParsingGrid);
+
+        return Visualizations.builder()
+                .layer(visualParsingGrid ? ContentStreams.KNECON_VISUAL_PARSING : ContentStreams.KNECON_LAYOUT)
+                .visualizationsOnPages(layoutGrid.getVisualizationsPerPages())
+                .layerVisibilityDefaultValue(layerVisibilityDefaultValue)
+                .build();
+    }
+
+
+    private LayoutGrid createLayoutGrid(Document document, boolean visualParsingGrid) {
 
         LayoutGrid layoutGrid = new LayoutGrid(document.getNumberOfPages());
-        document.streamAllSubNodes().forEach(semanticNode -> {
-            Color color = switch (semanticNode.getType()) {
-                case PARAGRAPH -> PARAGRAPH_COLOR;
-                case TABLE -> TABLE_COLOR;
-                case SECTION -> SECTION_COLOR;
-                case HEADLINE -> HEADLINE_COLOR;
-                case HEADER, FOOTER -> HEADER_COLOR;
-                case IMAGE -> IMAGE_COLOR;
-                default -> null;
-            };
-            if (isNotSectionOrTableCellOrDocument(semanticNode)) {
-                addAsRectangle(semanticNode, layoutGrid, color);
-            }
-            if (semanticNode.getType().equals(NodeType.SECTION)) {
-                addSection(semanticNode, layoutGrid, color);
-            }
-            if (semanticNode.getType().equals(NodeType.TABLE)) {
-                Table table = (Table) semanticNode;
-                addInnerTableLines(table, layoutGrid);
-            }
-        });
+        document.streamAllSubNodes()
+                .filter(node -> (node.getEngines().contains(LayoutEngine.AI) && visualParsingGrid) || (node.getEngines().contains(LayoutEngine.ALGORITHM) && !visualParsingGrid))
+                .forEach(semanticNode -> {
+                    Color color = switch (semanticNode.getType()) {
+                        case PARAGRAPH -> PARAGRAPH_COLOR;
+                        case TABLE -> TABLE_COLOR;
+                        case SECTION -> SECTION_COLOR;
+                        case HEADLINE -> HEADLINE_COLOR;
+                        case HEADER, FOOTER -> HEADER_COLOR;
+                        case IMAGE -> IMAGE_COLOR;
+                        default -> null;
+                    };
 
+                    if (semanticNode instanceof DuplicatedParagraph) {
+                        color = DUPLICATE_PARAGRAPH_COLOR;
+                    }
+
+                    if (isNotSectionOrTableCellOrDocument(semanticNode)) {
+                        addAsRectangle(semanticNode, layoutGrid, color);
+                    }
+                    if (semanticNode.getType().equals(NodeType.SECTION)) {
+                        addSection(semanticNode, layoutGrid, color);
+                    }
+                    if (semanticNode.getType().equals(NodeType.TABLE)) {
+                        Table table = (Table) semanticNode;
+                        addInnerTableLines(table, layoutGrid);
+                    }
+                });
         return layoutGrid;
     }
 
@@ -77,11 +129,13 @@ public class LayoutGridService {
             return;
         }
         for (Page page : table.getPages()) {
+
             Optional<Integer> optionalFirstRowOnPage = table.streamCol(0).filter(tableCell -> tableCell.isOnPage(page.getNumber())).map(TableCell::getRow).findFirst();
             if (optionalFirstRowOnPage.isEmpty()) {
                 continue;
             }
             int firstRowOnPage = optionalFirstRowOnPage.get();
+
             Stream<Double> xStream = switch (page.getRotation()) {
                 case 90 -> streamBBoxOfCellsOnPage(table.streamCol(0), page).map(RectangularShape::getMinX);
                 case 180 -> streamBBoxOfCellsOnPage(table.streamRow(firstRowOnPage), page).map(RectangularShape::getMaxX);
@@ -90,6 +144,7 @@ public class LayoutGridService {
             };
             List<Double> xs = xStream.collect(Collectors.toList());
             xs.remove(0);
+
             Stream<Double> yStream = switch (page.getRotation()) {
                 case 90 -> streamBBoxOfCellsOnPage(table.streamRow(firstRowOnPage), page).map(RectangularShape::getMinY);
                 case 180 -> streamBBoxOfCellsOnPage(table.streamCol(0), page).map(RectangularShape::getMinY);
@@ -99,15 +154,15 @@ public class LayoutGridService {
             List<Double> ys = yStream.collect(Collectors.toList());
             ys.remove(0);
 
-            Rectangle2D tableBBox = table.getBBox().get(table.getFirstPage());
+            Rectangle2D tableBBox = table.getBBox().get(page);
             List<ColoredLine> coloredLines = layoutGrid.getVisualizationsPerPages().get(page.getNumber() - 1).getColoredLines();
             xs.forEach(x -> {
                 Line2D line = new Line2D.Double(new Point2D.Double(x, tableBBox.getMaxY()), new Point2D.Double(x, tableBBox.getMinY()));
-                coloredLines.add(new ColoredLine(line, INNER_LINES_COLOR));
+                coloredLines.add(new ColoredLine(line, INNER_LINES_COLOR, LINE_WIDTH));
             });
             ys.forEach(y -> {
                 Line2D line = new Line2D.Double(new Point2D.Double(tableBBox.getMinX(), y), new Point2D.Double(tableBBox.getMaxX(), y));
-                coloredLines.add(new ColoredLine(line, INNER_LINES_COLOR));
+                coloredLines.add(new ColoredLine(line, INNER_LINES_COLOR, LINE_WIDTH));
             });
         }
     }
@@ -135,9 +190,9 @@ public class LayoutGridService {
             List<Line2D> lines = createLinesFromRectangle(r, firstPage.getRotation());
             // add string to top line
             var firstLine = lines.remove(0);
-            coloredLines.add(new ColoredLine(firstLine, color));
+            coloredLines.add(new ColoredLine(firstLine, color, LINE_WIDTH));
             for (Line2D line : lines) {
-                coloredLines.add(new ColoredLine(line, color));
+                coloredLines.add(new ColoredLine(line, color, LINE_WIDTH));
             }
             return;
         }
@@ -152,16 +207,37 @@ public class LayoutGridService {
     }
 
 
+    @SneakyThrows
     private void addPlacedText(Page page, Rectangle2D textBBox, String s, LayoutGrid layoutGrid) {
 
-        Point2D.Float upperLeftCorner = switch (page.getRotation()) {
-            case 90 -> new Point2D.Float((float) (textBBox.getMinX()), (float) textBBox.getMinY());
-            case 180 -> new Point2D.Float((float) (textBBox.getMaxX()), (float) textBBox.getMinY());
-            case 270 -> new Point2D.Float((float) (textBBox.getMaxX()), (float) textBBox.getMaxY());
-            default -> new Point2D.Float((float) (textBBox.getMinX()), (float) textBBox.getMaxY());
-        };
+        // translates text, such that its right edge is a bit to the left of the drawn box
+        float translationAmount = ((FONT.getStringWidth(s) / 1000) * FONT_SIZE + (2 * LINE_WIDTH) + 4);
+
+        Point2D upperLeftCorner;
+        Point2D translationVector;
+        switch (page.getRotation()) {
+            case 90 -> {
+                upperLeftCorner = new Point2D.Double(textBBox.getMinX(), textBBox.getMinY());
+                translationVector = new Point2D.Double(FONT_SIZE, -translationAmount);
+            }
+            case 180 -> {
+                upperLeftCorner = new Point2D.Double(textBBox.getMaxX(), textBBox.getMinY());
+                translationVector = new Point2D.Double(translationAmount, FONT_SIZE);
+            }
+            case 270 -> {
+                upperLeftCorner = new Point2D.Double(textBBox.getMaxX(), textBBox.getMaxY());
+                translationVector = new Point2D.Double(-FONT_SIZE, translationAmount);
+            }
+            default -> {
+                upperLeftCorner = new Point2D.Double(textBBox.getMinX(), textBBox.getMaxY());
+                translationVector = new Point2D.Double(-translationAmount, -FONT_SIZE);
+            }
+        }
+
+        upperLeftCorner = add(upperLeftCorner, translationVector);
+
         var placedTexts = layoutGrid.getVisualizationsPerPages().get(page.getNumber() - 1).getPlacedTexts();
-        placedTexts.add(new PlacedText(s, upperLeftCorner));
+        placedTexts.add(PlacedText.textFacingUp(s, upperLeftCorner, FONT_SIZE, Color.BLACK, FONT));
     }
 
 
@@ -176,9 +252,9 @@ public class LayoutGridService {
         midPageLines.remove(1);
         // add string to left line
         var leftLine = midPageLines.remove(1);
-        coloredLines.add(new ColoredLine(leftLine, color));
+        coloredLines.add(new ColoredLine(leftLine, color, LINE_WIDTH));
         for (Line2D line : midPageLines) {
-            coloredLines.add(new ColoredLine(line, color));
+            coloredLines.add(new ColoredLine(line, color, LINE_WIDTH));
         }
     }
 
@@ -192,9 +268,9 @@ public class LayoutGridService {
         lastPageLines.remove(0);
         // add string to left line
         var leftLine = lastPageLines.remove(2);
-        coloredLines.add(new ColoredLine(leftLine, color));
+        coloredLines.add(new ColoredLine(leftLine, color, LINE_WIDTH));
         for (Line2D line : lastPageLines) {
-            coloredLines.add(new ColoredLine(line, color));
+            coloredLines.add(new ColoredLine(line, color, LINE_WIDTH));
         }
     }
 
@@ -208,9 +284,9 @@ public class LayoutGridService {
         firstPageLines.remove(2);
         // add string to top line
         var firstLine = firstPageLines.remove(0);
-        coloredLines.add(new ColoredLine(firstLine, color));
+        coloredLines.add(new ColoredLine(firstLine, color, LINE_WIDTH));
         for (Line2D line : firstPageLines) {
-            coloredLines.add(new ColoredLine(line, color));
+            coloredLines.add(new ColoredLine(line, color, LINE_WIDTH));
         }
     }
 
@@ -276,7 +352,16 @@ public class LayoutGridService {
     private void addAsRectangle(SemanticNode semanticNode, LayoutGrid layoutGrid, Color color) {
 
         semanticNode.getBBox()
-                .forEach((page, textBBox) -> layoutGrid.getVisualizationsPerPages().get(page.getNumber() - 1).getColoredRectangles().add(new ColoredRectangle(textBBox, color)));
+                .forEach((page, textBBox) -> layoutGrid.getVisualizationsPerPages()
+                        .get(page.getNumber() - 1)
+                        .getColoredRectangles()
+                        .add(new ColoredRectangle(textBBox, color, LINE_WIDTH)));
+    }
+
+
+    private Point2D add(Point2D a, Point2D b) {
+
+        return new Point2D.Double(a.getX() + b.getX(), a.getY() + b.getY());
     }
 
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/visualization/ViewerDocumentService.java

@@ -1,217 +0,0 @@
-package com.knecon.fforesight.service.layoutparser.processor.services.visualization;
-
-import java.awt.geom.AffineTransform;
-import java.awt.geom.Rectangle2D;
-import java.io.File;
-import java.io.IOException;
-import java.nio.file.Files;
-import java.nio.file.Path;
-import java.nio.file.StandardCopyOption;
-
-import org.apache.pdfbox.Loader;
-import org.apache.pdfbox.cos.COSName;
-import org.apache.pdfbox.pdmodel.PDDocument;
-import org.apache.pdfbox.pdmodel.PDDocumentCatalog;
-import org.apache.pdfbox.pdmodel.PDPage;
-import org.apache.pdfbox.pdmodel.PDPageContentStream;
-import org.apache.pdfbox.pdmodel.PDResources;
-import org.apache.pdfbox.pdmodel.font.PDFont;
-import org.apache.pdfbox.pdmodel.font.PDType1Font;
-import org.apache.pdfbox.pdmodel.font.Standard14Fonts;
-import org.apache.pdfbox.pdmodel.graphics.optionalcontent.PDOptionalContentGroup;
-import org.apache.pdfbox.pdmodel.graphics.optionalcontent.PDOptionalContentProperties;
-import org.apache.pdfbox.pdmodel.graphics.state.PDExtendedGraphicsState;
-import org.apache.pdfbox.util.Matrix;
-import org.springframework.stereotype.Service;
-
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
-import com.knecon.fforesight.service.layoutparser.processor.model.visualization.ColoredLine;
-import com.knecon.fforesight.service.layoutparser.processor.model.visualization.ColoredRectangle;
-import com.knecon.fforesight.service.layoutparser.processor.model.visualization.FilledRectangle;
-import com.knecon.fforesight.service.layoutparser.processor.model.visualization.LayoutGrid;
-import com.knecon.fforesight.service.layoutparser.processor.model.visualization.PlacedText;
-import com.knecon.fforesight.service.layoutparser.processor.model.visualization.VisualizationsOnPage;
-
-import io.micrometer.observation.Observation;
-import io.micrometer.observation.ObservationRegistry;
-import io.micrometer.observation.annotation.Observed;
-import lombok.RequiredArgsConstructor;
-import lombok.SneakyThrows;
-import lombok.extern.slf4j.Slf4j;
-
-@Slf4j
-@Service
-@RequiredArgsConstructor
-public class ViewerDocumentService {
-
-    private static final String LAYER_NAME = "Layout grid";
-    private static final int FONT_SIZE = 10;
-    public static final float LINE_WIDTH = 1f;
-
-    private final LayoutGridService layoutGridService;
-    private final ObservationRegistry observationRegistry;
-
-
-    @SneakyThrows
-    @Observed(name = "ViewerDocumentService", contextualName = "create-viewer-document")
-    public void createViewerDocument(File originFile, Document document, File destinationFile, boolean layerVisibilityDefaultValue) {
-
-        Path tmpFile = Files.createTempFile("tmpViewerDocument", ".pdf");
-        PDDocument pdDocument = openPDDocument(originFile);
-        LayoutGrid layoutGrid = layoutGridService.createLayoutGrid(document);
-
-        PDOptionalContentGroup layer = addLayerToDocument(pdDocument, layerVisibilityDefaultValue);
-        PDFont font = new PDType1Font(Standard14Fonts.FontName.HELVETICA);
-
-        for (int pageNumber = 0; pageNumber < pdDocument.getNumberOfPages(); pageNumber++) {
-
-            PDPage pdPage = pdDocument.getPage(pageNumber);
-//
-            AffineTransform textDeRotationMatrix = getTextDeRotationTransform(pdPage);
-            addLayerToPageResources(pdPage);
-
-            // We need to save the graphics state before, such that our appended content cannot be affected by previous content streams with side effects,
-            // e.g. not escaped matrix transformations.
-            escapePreviousContents(pdDocument, pdPage);
-
-            VisualizationsOnPage visualizationsOnPage = layoutGrid.getVisualizationsPerPages().get(pageNumber);
-            assert pageNumber == visualizationsOnPage.getPageNumber();
-            // We need to append to the content stream, otherwise the content could be overlapped by following content.
-            try (var contentStream = new PDPageContentStream(pdDocument, pdPage, PDPageContentStream.AppendMode.APPEND, true)) {
-
-                contentStream.beginMarkedContent(COSName.OC, layer);
-                contentStream.saveGraphicsState();
-
-                contentStream.setLineWidth(LINE_WIDTH);
-                for (ColoredLine coloredLine : visualizationsOnPage.getColoredLines()) {
-                    contentStream.setStrokingColor(coloredLine.color());
-                    contentStream.moveTo((float) coloredLine.line().getX1(), (float) coloredLine.line().getY1());
-                    contentStream.lineTo((float) coloredLine.line().getX2(), (float) coloredLine.line().getY2());
-                    contentStream.stroke();
-                }
-                for (ColoredRectangle coloredRectangle : visualizationsOnPage.getColoredRectangles()) {
-                    contentStream.setStrokingColor(coloredRectangle.color());
-                    Rectangle2D r = coloredRectangle.rectangle2D();
-                    contentStream.addRect((float) r.getX(), (float) r.getY(), (float) r.getWidth(), (float) r.getHeight());
-                    contentStream.stroke();
-                }
-                for (FilledRectangle filledRectangle : visualizationsOnPage.getFilledRectangles()) {
-                    contentStream.setNonStrokingColor(filledRectangle.color());
-                    PDExtendedGraphicsState graphicsState = new PDExtendedGraphicsState();
-                    graphicsState.setNonStrokingAlphaConstant(filledRectangle.alpha());
-                    contentStream.setGraphicsStateParameters(graphicsState);
-                    Rectangle2D r = filledRectangle.rectangle2D();
-                    contentStream.addRect((float) r.getX(), (float) r.getY(), (float) r.getWidth(), (float) r.getHeight());
-                    contentStream.fill();
-                }
-                for (PlacedText placedText : visualizationsOnPage.getPlacedTexts()) {
-                    contentStream.setFont(font, FONT_SIZE);
-                    contentStream.beginText();
-                    Matrix textMatrix = new Matrix((float) textDeRotationMatrix.getScaleX(),
-                            (float) textDeRotationMatrix.getShearX(),
-                            (float) textDeRotationMatrix.getShearY(),
-                            (float) textDeRotationMatrix.getScaleY(),
-                            (float) placedText.lineStart().getX(),
-                            (float) placedText.lineStart().getY());
-                    textMatrix.translate(-((font.getStringWidth(placedText.text()) / 1000) * FONT_SIZE + (2 * LINE_WIDTH) + 4), -FONT_SIZE);
-                    contentStream.setTextMatrix(textMatrix);
-                    contentStream.showText(placedText.text());
-                    contentStream.endText();
-                }
-                contentStream.restoreGraphicsState();
-                contentStream.endMarkedContent();
-            }
-
-            if (pageNumber % 500 == 0 && pageNumber != 0) { // re-open document every once in a while to save on RAM
-                log.info("Incremental save after {} pages", pageNumber);
-                observedIncrementalSave(pdDocument, destinationFile);
-                pdDocument.close();
-                Files.copy(destinationFile.toPath(), tmpFile, StandardCopyOption.REPLACE_EXISTING);
-                pdDocument = openPDDocument(tmpFile.toFile());
-                layer = addLayerToDocument(pdDocument, layerVisibilityDefaultValue);
-            }
-
-        }
-        observedIncrementalSave(pdDocument, destinationFile);
-
-        tmpFile.toFile().delete();
-        pdDocument.close();
-    }
-
-
-    private static PDDocument openPDDocument(File tmpFile) throws IOException {
-
-        PDDocument pdDocument;
-        pdDocument = Loader.loadPDF(tmpFile);
-        pdDocument.setAllSecurityToBeRemoved(true);
-        return pdDocument;
-    }
-
-
-    @SneakyThrows
-    private void observedIncrementalSave(PDDocument pdDocument, File outputFile) {
-
-        Observation.createNotStarted("ViewerDocumentService", observationRegistry).contextualName("incremental-save").observe(() -> {
-            try {
-                pdDocument.save(outputFile);
-            } catch (IOException e) {
-                throw new RuntimeException(e);
-            }
-        });
-
-    }
-
-
-    private static void addLayerToPageResources(PDPage pdPage) {
-
-        PDResources resources = pdPage.getResources();
-        if (resources == null) {
-            resources = new PDResources();
-            pdPage.setResources(resources);
-        }
-    }
-
-
-    private static void escapePreviousContents(PDDocument pdDocument, PDPage pdPage) throws IOException {
-
-        try (var contentStream = new PDPageContentStream(pdDocument, pdPage, PDPageContentStream.AppendMode.PREPEND, false)) {
-            contentStream.saveGraphicsState();
-        }
-        try (var contentStream = new PDPageContentStream(pdDocument, pdPage, PDPageContentStream.AppendMode.APPEND, false)) {
-            contentStream.restoreGraphicsState();
-        }
-    }
-
-
-    private static PDOptionalContentGroup addLayerToDocument(PDDocument pdDocument, boolean layerVisibilityDefaultValue) {
-
-        PDDocumentCatalog catalog = pdDocument.getDocumentCatalog();
-        PDOptionalContentProperties ocprops = catalog.getOCProperties();
-        if (ocprops == null) {
-            ocprops = new PDOptionalContentProperties();
-            catalog.setOCProperties(ocprops);
-        }
-        PDOptionalContentGroup layer = null;
-        if (ocprops.hasGroup(LAYER_NAME)) {
-            layer = ocprops.getGroup(LAYER_NAME);
-        } else {
-            layer = new PDOptionalContentGroup(LAYER_NAME);
-            ocprops.addGroup(layer);
-        }
-        ocprops.setGroupEnabled(layer, layerVisibilityDefaultValue);
-//        dictionariesToUpdate.add(catalog.getCOSObject());
-        return layer;
-    }
-
-
-    private static AffineTransform getTextDeRotationTransform(PDPage page) {
-
-        return AffineTransform.getQuadrantRotateInstance(switch (page.getRotation()) {
-            case 90 -> 3;
-            case 180 -> 2;
-            case 270 -> 1;
-            default -> 0;
-        });
-    }
-
-}

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

@@ -0,0 +1,28 @@
+package com.knecon.fforesight.service.layoutparser.processor.utils;
+
+import java.awt.Color;
+
+import lombok.AccessLevel;
+import lombok.AllArgsConstructor;
+import lombok.Builder;
+import lombok.Getter;
+import lombok.experimental.FieldDefaults;
+
+@Builder
+@AllArgsConstructor
+@Getter
+@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
+public class DrawingOptions {
+
+    boolean stroke;
+    @Builder.Default
+    Color strokeColor = Color.BLACK;
+    @Builder.Default
+    float strokeWidth = 1f;
+
+    boolean fill;
+    @Builder.Default
+    Color fillColor = Color.BLACK;
+
+}
+

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

@@ -0,0 +1,88 @@
+package com.knecon.fforesight.service.layoutparser.processor.utils;
+
+import java.awt.geom.Point2D;
+import java.util.Comparator;
+
+import com.knecon.fforesight.service.layoutparser.processor.model.table.Cell;
+import com.knecon.fforesight.service.layoutparser.processor.model.table.Rectangle;
+import com.knecon.fforesight.service.layoutparser.processor.model.table.Ruling;
+
+public class GeometricComparators {
+
+    private static final int COMPARATOR_ROUNDING = 2;
+
+    public static final Comparator<Point2D> X_FIRST_POINT_COMPARATOR = (point1, point2) -> {
+
+        int rv = 0;
+        float point1X = DoubleComparisons.round(point1.getX(), COMPARATOR_ROUNDING);
+        float point1Y = DoubleComparisons.round(point1.getY(), COMPARATOR_ROUNDING);
+        float point2X = DoubleComparisons.round(point2.getX(), COMPARATOR_ROUNDING);
+        float point2Y = DoubleComparisons.round(point2.getY(), COMPARATOR_ROUNDING);
+
+        if (point1X > point2X) {
+            rv = 1;
+        } else if (point1X < point2X) {
+            rv = -1;
+        } else if (point1Y > point2Y) {
+            rv = 1;
+        } else if (point1Y < point2Y) {
+            rv = -1;
+        }
+        return rv;
+    };
+
+    public static final Comparator<Point2D> Y_FIRST_POINT_COMPARATOR = (point1, point2) -> {
+
+        int rv = 0;
+        float point1X = DoubleComparisons.round(point1.getX(), COMPARATOR_ROUNDING);
+        float point1Y = DoubleComparisons.round(point1.getY(), COMPARATOR_ROUNDING);
+        float point2X = DoubleComparisons.round(point2.getX(), COMPARATOR_ROUNDING);
+        float point2Y = DoubleComparisons.round(point2.getY(), COMPARATOR_ROUNDING);
+
+        if (point1Y > point2Y) {
+            rv = 1;
+        } else if (point1Y < point2Y) {
+            rv = -1;
+        } else if (point1X > point2X) {
+            rv = 1;
+        } else if (point1X < point2X) {
+            rv = -1;
+        }
+        return rv;
+    };
+
+    public static final Comparator<Cell> CELL_SIZE_COMPARATOR = (cell1, cell2) -> {
+
+        Double cell1Size = cell1.getHeight() * cell1.getWidth();
+        Double cell2Size = cell2.getHeight() * cell2.getWidth();
+        return cell1Size.compareTo(cell2Size);
+    };
+
+    public static final Comparator<Rectangle> RECTANGLE_SIZE_COMPARATOR = (rect1, rect2) -> {
+
+        Double rect1Size = rect1.getHeight() * rect1.getWidth();
+        Double rect2Size = rect2.getHeight() * rect2.getWidth();
+        return rect1Size.compareTo(rect2Size);
+    };
+
+    public static final Comparator<Ruling> X_FIRST_RULING_COMPARATOR = (ruling1, ruling2) -> {
+
+        int rv = 0;
+        float point1X = DoubleComparisons.round(Math.min(ruling1.getLeft(), ruling1.getRight()), COMPARATOR_ROUNDING);
+        float point1Y = DoubleComparisons.round(Math.min(ruling1.getTop(), ruling1.getBottom()), COMPARATOR_ROUNDING);
+        float point2X = DoubleComparisons.round(Math.min(ruling2.getLeft(), ruling2.getRight()), COMPARATOR_ROUNDING);
+        float point2Y = DoubleComparisons.round(Math.min(ruling2.getTop(), ruling2.getBottom()), COMPARATOR_ROUNDING);
+
+        if (point1X > point2X) {
+            rv = 1;
+        } else if (point1X < point2X) {
+            rv = -1;
+        } else if (point1Y > point2Y) {
+            rv = 1;
+        } else if (point1Y < point2Y) {
+            rv = -1;
+        }
+        return rv;
+    };
+
+}

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

@@ -21,11 +21,7 @@ import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Pa
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.AtomicTextBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
 
-import lombok.AccessLevel;
-import lombok.Builder;
-import lombok.Getter;
 import lombok.SneakyThrows;
-import lombok.experimental.FieldDefaults;
 import lombok.experimental.UtilityClass;
 import lombok.extern.slf4j.Slf4j;
 
@@ -41,20 +37,20 @@ public class PdfVisualisationUtility {
 
     public void drawNode(PDDocument document, DocumentTree.Entry entry) {
 
-        Options options = buildStandardOptionsForNodes(entry);
+        DrawingOptions options = buildStandardOptionsForNodes(entry);
 
         drawBBoxAndLabelAndNumberOnPage(document, entry, options);
 
     }
 
 
-    public void drawTextBlock(PDDocument document, TextBlock textBlock, Options options) {
+    public void drawTextBlock(PDDocument document, TextBlock textBlock, DrawingOptions options) {
 
         textBlock.getAtomicTextBlocks().forEach(atb -> drawAtomicTextBlock(document, atb, options));
     }
 
 
-    public void drawAtomicTextBlock(PDDocument document, AtomicTextBlock atomicTextBlock, Options options) {
+    public void drawAtomicTextBlock(PDDocument document, AtomicTextBlock atomicTextBlock, DrawingOptions options) {
 
         drawRectangle2DList(document, atomicTextBlock.getPage().getNumber(), atomicTextBlock.getPositions().stream().toList(), options);
 
@@ -62,7 +58,7 @@ public class PdfVisualisationUtility {
 
 
     @SneakyThrows
-    public void drawText(String string, PDDocument document, Point2D location, Integer pageNumber, Options options) {
+    public void drawText(String string, PDDocument document, Point2D location, Integer pageNumber, DrawingOptions options) {
 
         var pdPage = document.getPage(pageNumber - 1);
         var contentStream = new PDPageContentStream(document, pdPage, PDPageContentStream.AppendMode.APPEND, true);
@@ -80,14 +76,14 @@ public class PdfVisualisationUtility {
 
 
     @SneakyThrows
-    public void drawRectangle2DList(PDDocument document, int pageNumber, List<Rectangle2D> rectCollection, Options options) {
+    public void drawRectangle2DList(PDDocument document, int pageNumber, List<Rectangle2D> rectCollection, DrawingOptions options) {
 
         var pdPage = document.getPage(pageNumber - 1);
         drawRectangle2DList(document, rectCollection, options, pdPage);
     }
 
 
-    private void drawRectangle2DList(PDDocument document, List<Rectangle2D> rectCollection, Options options, PDPage pdPage) throws IOException {
+    private void drawRectangle2DList(PDDocument document, List<Rectangle2D> rectCollection, DrawingOptions options, PDPage pdPage) throws IOException {
 
         var contentStream = new PDPageContentStream(document, pdPage, PDPageContentStream.AppendMode.APPEND, true);
 
@@ -110,9 +106,9 @@ public class PdfVisualisationUtility {
     }
 
 
-    private Options buildStandardOptionsForNodes(DocumentTree.Entry entry) {
+    private DrawingOptions buildStandardOptionsForNodes(DocumentTree.Entry entry) {
 
-        return Options.builder().stroke(true).strokeColor(switch (entry.getType()) {
+        return DrawingOptions.builder().stroke(true).strokeColor(switch (entry.getType()) {
             case DOCUMENT -> Color.LIGHT_GRAY;
             case HEADER, FOOTER -> Color.GREEN;
             case PARAGRAPH -> Color.BLUE;
@@ -125,7 +121,7 @@ public class PdfVisualisationUtility {
     }
 
 
-    private void drawBBoxAndLabelAndNumberOnPage(PDDocument document, DocumentTree.Entry entry, Options options) {
+    private void drawBBoxAndLabelAndNumberOnPage(PDDocument document, DocumentTree.Entry entry, DrawingOptions options) {
 
         Map<Page, Rectangle2D> rectanglesPerPage = entry.getNode().getBBox();
         rectanglesPerPage.forEach((page, rectangle2D) -> {
@@ -152,7 +148,7 @@ public class PdfVisualisationUtility {
 
 
     @SneakyThrows
-    public static void drawLine2DList(PDDocument pdDocument, int pageNumber, List<? extends Line2D> line2DS, Options options) {
+    public static void drawLine2DList(PDDocument pdDocument, int pageNumber, List<? extends Line2D> line2DS, DrawingOptions options) {
 
         var pdPage = pdDocument.getPage(pageNumber - 1);
         var contentStream = new PDPageContentStream(pdDocument, pdPage, PDPageContentStream.AppendMode.APPEND, true);
@@ -176,21 +172,4 @@ public class PdfVisualisationUtility {
         contentStream.close();
     }
 
-
-    @Builder
-    @Getter
-    @FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
-    public static class Options {
-
-        boolean fill;
-        boolean stroke;
-        @Builder.Default
-        Color strokeColor = Color.BLACK;
-        @Builder.Default
-        float strokeWidth = 1f;
-        @Builder.Default
-        Color fillColor = Color.BLACK;
-
-    }
-
 }

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

@@ -2,8 +2,10 @@ package com.knecon.fforesight.service.layoutparser.processor.utils;
 
 import static java.lang.String.format;
 
+import java.awt.geom.Point2D;
 import java.awt.geom.Rectangle2D;
 import java.awt.geom.RectangularShape;
+import java.util.ArrayList;
 import java.util.Collections;
 import java.util.LinkedList;
 import java.util.List;
@@ -19,6 +21,8 @@ import org.apache.pdfbox.pdmodel.common.PDRectangle;
 import com.iqser.red.service.persistence.service.v1.api.shared.model.redactionlog.Point;
 import com.iqser.red.service.persistence.service.v1.api.shared.model.redactionlog.Rectangle;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.AtomicTextBlock;
+import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRulings;
+import com.knecon.fforesight.service.layoutparser.processor.model.table.Ruling;
 
 import lombok.AllArgsConstructor;
 import lombok.NoArgsConstructor;
@@ -37,15 +41,27 @@ public class RectangleTransformations {
     }
 
 
+    public static double calculateIntersectedArea(Rectangle2D r1, Rectangle2D r2) {
+
+        double xOverlap = Math.max(0, Math.min(r1.getMaxX(), r2.getMaxX()) - Math.max(r1.getMinX(), r2.getMinX()));
+        double yOverlap = Math.max(0, Math.min(r1.getMaxY(), r2.getMaxY()) - Math.max(r1.getY(), r2.getY()));
+
+        return xOverlap * yOverlap;
+    }
+
+
     public static Rectangle2D bBoxUnionAtomicTextBlock(List<AtomicTextBlock> atomicTextBlocks) {
 
         return atomicTextBlocks.stream().flatMap(atomicTextBlock -> atomicTextBlock.getPositions().stream()).collect(new Rectangle2DBBoxCollector());
     }
+
+
     public static Collector<Rectangle2D, Rectangle2DBBoxCollector.BBox, Rectangle2D> collectBBox() {
 
         return new Rectangle2DBBoxCollector();
     }
 
+
     public static PDRectangle toPDRectangleBBox(List<Rectangle> rectangles) {
 
         Rectangle2D rectangle2D = RectangleTransformations.rectangleBBox(rectangles);
@@ -70,6 +86,7 @@ public class RectangleTransformations {
         return format("%f,%f,%f,%f", rectangle2D.getX(), rectangle2D.getY(), rectangle2D.getWidth(), rectangle2D.getHeight());
     }
 
+
     public static Rectangle2D rectangleBBox(List<Rectangle> rectangles) {
 
         return rectangles.stream().map(RectangleTransformations::toRectangle2D).collect(new Rectangle2DBBoxCollector());
@@ -84,6 +101,7 @@ public class RectangleTransformations {
                 -redactionLogRectangle.getHeight());
     }
 
+
     public static Rectangle2D toRectangle2D(PDRectangle rectangle) {
 
         return new Rectangle2D.Double(rectangle.getLowerLeftX(), rectangle.getLowerLeftY(), rectangle.getWidth(), rectangle.getHeight());
@@ -133,7 +151,27 @@ public class RectangleTransformations {
                 previousRectangle = currentRectangle;
             }
         }
-        return rectangleListsWithGaps.stream().map(RectangleTransformations::rectangle2DBBox).toList();
+        return rectangleListsWithGaps.stream()
+                .map(RectangleTransformations::rectangle2DBBox)
+                .toList();
+    }
+
+
+    public static CleanRulings extractRulings(List<? extends Rectangle2D.Float> rectangles) {
+
+        // Underlined or strikethrough are also in rulings but we dont want to split blocks with them so we use cells.
+        List<Ruling> horizontalRulings = new ArrayList<>();
+        List<Ruling> verticalRulings = new ArrayList<>();
+
+        rectangles.forEach(rectangle -> {
+            horizontalRulings.add(new Ruling(new Point2D.Float(rectangle.x, rectangle.y), new Point2D.Float(rectangle.x + rectangle.width, rectangle.y)));
+            horizontalRulings.add(new Ruling(new Point2D.Float(rectangle.x, rectangle.y + rectangle.height),
+                                             new Point2D.Float(rectangle.x + rectangle.width, rectangle.y + rectangle.height)));
+            verticalRulings.add(new Ruling(new Point2D.Float(rectangle.x, rectangle.y), new Point2D.Float(rectangle.x, rectangle.y + rectangle.height)));
+            verticalRulings.add(new Ruling(new Point2D.Float(rectangle.x + rectangle.width, rectangle.y),
+                                           new Point2D.Float(rectangle.x + rectangle.width, rectangle.y + rectangle.height)));
+        });
+        return CleanRulings.builder().vertical(verticalRulings).horizontal(horizontalRulings).build();
     }
 
 

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

@@ -0,0 +1,77 @@
+package com.knecon.fforesight.service.layoutparser.processor.utils;
+
+import static com.knecon.fforesight.service.layoutparser.processor.utils.GeometricComparators.Y_FIRST_POINT_COMPARATOR;
+
+import java.awt.geom.Point2D;
+import java.awt.geom.Rectangle2D;
+import java.util.ArrayList;
+import java.util.List;
+import java.util.Map;
+
+import com.knecon.fforesight.service.layoutparser.processor.model.table.Ruling;
+
+public class RectangularIntersectionFinder {
+
+    public static List<Rectangle2D> find(List<Ruling> horizontalRulingLines, List<Ruling> verticalRulingLines) {
+
+        // Fix for 211.pdf
+        for (Ruling r : horizontalRulingLines) {
+            if (r.getX2() < r.getX1()) {
+                double a = r.getX2();
+                r.x2 = (float) r.getX1();
+                r.x1 = (float) a;
+            }
+        }
+
+        List<Rectangle2D> foundRectangles = new ArrayList<>();
+        Map<Point2D, Ruling[]> intersectionPoints = Ruling.findIntersections(horizontalRulingLines, verticalRulingLines);
+        List<Point2D> intersectionPointsList = new ArrayList<>(intersectionPoints.keySet());
+        intersectionPointsList.sort(Y_FIRST_POINT_COMPARATOR);
+
+        for (int i = 0; i < intersectionPointsList.size(); i++) {
+            Point2D topLeft = intersectionPointsList.get(i);
+            Ruling[] hv = intersectionPoints.get(topLeft);
+
+            // CrossingPointsDirectlyBelow( topLeft );
+            List<Point2D> xPoints = new ArrayList<>();
+            // CrossingPointsDirectlyToTheRight( topLeft );
+            List<Point2D> yPoints = new ArrayList<>();
+
+            for (Point2D p : intersectionPointsList.subList(i, intersectionPointsList.size())) {
+                if (p.getX() == topLeft.getX() && p.getY() > topLeft.getY()) {
+                    xPoints.add(p);
+                }
+                if (p.getY() == topLeft.getY() && p.getX() > topLeft.getX()) {
+                    yPoints.add(p);
+                }
+            }
+            outer:
+            for (Point2D xPoint : xPoints) {
+                // is there a vertical edge b/w topLeft and xPoint?
+                if (!hv[1].equals(intersectionPoints.get(xPoint)[1])) {
+                    continue;
+                }
+                for (Point2D yPoint : yPoints) {
+                    // is there a horizontal edge b/w topLeft and yPoint ?
+                    if (!hv[0].equals(intersectionPoints.get(yPoint)[0])) {
+                        continue;
+                    }
+                    Point2D btmRight = new Point2D.Float((float) yPoint.getX(), (float) xPoint.getY());
+                    if (intersectionPoints.containsKey(btmRight)
+                        && intersectionPoints.get(btmRight)[0].equals(intersectionPoints.get(xPoint)[0])
+                        && intersectionPoints.get(btmRight)[1].equals(intersectionPoints.get(yPoint)[1])) {
+                        foundRectangles.add(new Rectangle2D.Double(topLeft.getX(), topLeft.getY(), btmRight.getX() - topLeft.getX(), btmRight.getY() - topLeft.getY()));
+                        break outer;
+                    }
+                }
+            }
+        }
+
+        // TODO create cells for vertical ruling lines with aligned endpoints at the top/bottom of a grid
+        // that aren't connected with an horizontal ruler?
+        // see: https://github.com/jazzido/tabula-extractor/issues/78#issuecomment-41481207
+
+        return foundRectangles;
+    }
+
+}

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

@@ -0,0 +1,172 @@
+package com.knecon.fforesight.service.layoutparser.processor.utils;
+
+import static com.knecon.fforesight.service.layoutparser.processor.utils.GeometricComparators.X_FIRST_POINT_COMPARATOR;
+import static com.knecon.fforesight.service.layoutparser.processor.utils.GeometricComparators.Y_FIRST_POINT_COMPARATOR;
+
+import java.awt.geom.Point2D;
+import java.util.ArrayList;
+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.processor.model.table.Rectangle;
+
+public class SpreadsheetFinder {
+
+    private static final int MAX_OUTER_POINT_TOLERANCE = 10;
+    private static final float AREA_TOLERANCE = 0.001f;
+
+
+    public static List<Rectangle> findSpreadsheetsFromCells(List<? extends Rectangle> cells) {
+        // via: http://stackoverflow.com/questions/13746284/merging-multiple-adjacent-rectangles-into-one-polygon
+        List<Rectangle> rectangles = new ArrayList<>();
+        Set<Point2D> pointSet = new HashSet<>();
+        Map<Point2D, Point2D> edgesH = new HashMap<>();
+        Map<Point2D, Point2D> edgesV = new HashMap<>();
+
+        for (Rectangle cell : cells) {
+            for (Point2D pt : cell.getPoints()) {
+                if (pointSet.contains(pt)) { // shared vertex, remove it
+                    pointSet.remove(pt);
+                } else {
+                    pointSet.add(pt);
+                }
+            }
+        }
+
+        // X first sort
+        List<Point2D> pointsSortX = new ArrayList<>(pointSet);
+        pointsSortX.sort(X_FIRST_POINT_COMPARATOR);
+        // Y first sort
+        List<Point2D> pointsSortY = new ArrayList<>(pointSet);
+        pointsSortY.sort(Y_FIRST_POINT_COMPARATOR);
+
+        int i = 0;
+        while (i < pointSet.size()) {
+            float currY = (float) pointsSortY.get(i).getY();
+            while (i < pointSet.size() && DoubleComparisons.feq(pointsSortY.get(i).getY(), currY)) {
+                edgesH.put(pointsSortY.get(i), pointsSortY.get(i + 1));
+                edgesH.put(pointsSortY.get(i + 1), pointsSortY.get(i));
+                i += 2;
+            }
+        }
+
+        i = 0;
+        while (i < pointSet.size()) {
+            float currX = (float) pointsSortX.get(i).getX();
+            while (i < pointSet.size() && DoubleComparisons.feq(pointsSortX.get(i).getX(), currX)) {
+                edgesV.put(pointsSortX.get(i), pointsSortX.get(i + 1));
+                edgesV.put(pointsSortX.get(i + 1), pointsSortX.get(i));
+                i += 2;
+            }
+        }
+
+        // Get all the polygons
+        List<List<PolygonVertex>> polygons = new ArrayList<>();
+        Point2D nextVertex;
+        while (!edgesH.isEmpty()) {
+            ArrayList<PolygonVertex> polygon = new ArrayList<>();
+            Point2D first = edgesH.keySet()
+                    .iterator().next();
+            polygon.add(new PolygonVertex(first, Direction.HORIZONTAL));
+            edgesH.remove(first);
+
+            while (true) {
+                PolygonVertex curr = polygon.get(polygon.size() - 1);
+                PolygonVertex lastAddedVertex;
+                if (curr.direction == Direction.HORIZONTAL) {
+                    nextVertex = edgesV.get(curr.point);
+                    edgesV.remove(curr.point);
+                    lastAddedVertex = new PolygonVertex(nextVertex, Direction.VERTICAL);
+                } else {
+                    nextVertex = edgesH.get(curr.point);
+                    edgesH.remove(curr.point);
+                    lastAddedVertex = new PolygonVertex(nextVertex, Direction.HORIZONTAL);
+                }
+                polygon.add(lastAddedVertex);
+
+                if (lastAddedVertex.equals(polygon.get(0))) {
+                    // closed polygon
+                    polygon.remove(polygon.size() - 1);
+                    break;
+                }
+            }
+
+            for (PolygonVertex vertex : polygon) {
+                edgesH.remove(vertex.point);
+                edgesV.remove(vertex.point);
+            }
+            polygons.add(polygon);
+        }
+
+        // calculate grid-aligned minimum area rectangles for each found polygon
+        for (List<PolygonVertex> poly : polygons) {
+            float top = Float.MAX_VALUE;
+            float left = Float.MAX_VALUE;
+            float bottom = Float.MIN_VALUE;
+            float right = Float.MIN_VALUE;
+            for (PolygonVertex pt : poly) {
+                top = (float) Math.min(top, pt.point.getY());
+                left = (float) Math.min(left, pt.point.getX());
+                bottom = (float) Math.max(bottom, pt.point.getY());
+                right = (float) Math.max(right, pt.point.getX());
+            }
+
+            // do not add polygons with too many outer points as they are unlikely to be tables
+            if (poly.size() <= MAX_OUTER_POINT_TOLERANCE) {
+                rectangles.add(new Rectangle(top - AREA_TOLERANCE, left - AREA_TOLERANCE, right - left + 2 * AREA_TOLERANCE, bottom - top + 2 * AREA_TOLERANCE));
+            }
+        }
+        return rectangles;
+    }
+
+
+    private enum Direction {
+        HORIZONTAL,
+        VERTICAL
+    }
+
+    static class PolygonVertex {
+
+        Point2D point;
+        Direction direction;
+
+
+        PolygonVertex(Point2D point, Direction direction) {
+
+            this.direction = direction;
+            this.point = point;
+        }
+
+
+        @Override
+        public boolean equals(Object other) {
+
+            if (this == other) {
+                return true;
+            }
+            if (!(other instanceof PolygonVertex)) {
+                return false;
+            }
+            return this.point.equals(((PolygonVertex) other).point);
+        }
+
+
+        @Override
+        public int hashCode() {
+
+            return this.point.hashCode();
+        }
+
+
+        @Override
+        public String toString() {
+
+            return String.format("%s[point=%s,direction=%s]", this.getClass().getName(), this.point.toString(), this.direction.toString());
+        }
+
+    }
+
+}

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

@@ -30,8 +30,6 @@ public class TableMergingUtility {
             if (consecutiveTable.getColCount() == originalTablePageBlock.getColCount() && !hasTableHeader(consecutiveTable) && outerBoundaryAlignsX(originalTablePageBlock,
                     consecutiveTable)) {
                 consecutiveTablesWithSameColCountAndHeaders.add(consecutiveTable);
-            } else {
-                break;
             }
         }
         return Stream.concat(Stream.of(originalTablePageBlock), consecutiveTablesWithSameColCountAndHeaders.stream()).toList();

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

@@ -3,7 +3,6 @@ package com.knecon.fforesight.service.layoutparser.processor.utils;
 import java.util.List;
 import java.util.stream.Collectors;
 
-
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;