Merge branch 'RED-9975' into 'main'

Red 9975: fix outline detection

See merge request fforesight/layout-parser!206

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

@@ -42,6 +42,15 @@ tasks.jacocoTestReport {
 }
 
 allprojects {
+
+    tasks.withType<Javadoc> {
+        options {
+            this as StandardJavadocDocletOptions
+            addBooleanOption("Xdoclint:none", true)
+            addStringOption("Xmaxwarns", "1")
+        }
+    }
+
     publishing {
         publications {
             create<MavenPublication>(name) {
@@ -64,6 +73,7 @@ java {
     withJavadocJar()
 }
 
+
 repositories {
     mavenLocal()
     mavenCentral()

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

@@ -43,6 +43,8 @@ public class DocumentStructure implements Serializable {
         public static final String POSITION = "position";
         public static final String ID = "id";
 
+        public static final String REPRESENTATION_HASH = "representationHash";
+
     }
 
     @Schema(description = "Object containing the extra field names, a table cell has in its properties field.")
@@ -67,11 +69,25 @@ public class DocumentStructure implements Serializable {
 
     public static Rectangle2D parseRectangle2D(String bBox) {
 
-        List<Float> floats = Arrays.stream(bBox.split(RECTANGLE_DELIMITER)).map(Float::parseFloat).toList();
+        List<Float> floats = Arrays.stream(bBox.split(RECTANGLE_DELIMITER))
+                .map(Float::parseFloat)
+                .toList();
         return new Rectangle2D.Float(floats.get(0), floats.get(1), floats.get(2), floats.get(3));
     }
 
 
+    public static double[] parseRepresentationVector(String representationHash) {
+
+        String[] stringArray = representationHash.split("[,\\s]+");
+        double[] doubleArray = new double[stringArray.length];
+        for (int i = 0; i < stringArray.length; i++) {
+            doubleArray[i] = Double.parseDouble(stringArray[i]);
+        }
+
+        return doubleArray;
+    }
+
+
     public EntryData get(List<Integer> tocId) {
 
         if (tocId.isEmpty()) {
@@ -87,19 +103,24 @@ public class DocumentStructure implements Serializable {
 
     public Stream<EntryData> streamAllEntries() {
 
-        return Stream.concat(Stream.of(root), root.children.stream()).flatMap(DocumentStructure::flatten);
+        return Stream.concat(Stream.of(root), root.children.stream())
+                .flatMap(DocumentStructure::flatten);
     }
 
 
     public String toString() {
 
-        return String.join("\n", streamAllEntries().map(EntryData::toString).toList());
+        return String.join("\n",
+                           streamAllEntries().map(EntryData::toString)
+                                   .toList());
     }
 
 
     private static Stream<EntryData> flatten(EntryData entry) {
 
-        return Stream.concat(Stream.of(entry), entry.children.stream().flatMap(DocumentStructure::flatten));
+        return Stream.concat(Stream.of(entry),
+                             entry.children.stream()
+                                     .flatMap(DocumentStructure::flatten));
     }
 
 

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

@@ -2,5 +2,6 @@ package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
 
 public enum LayoutEngine {
     ALGORITHM,
-    AI
+    AI,
+    OUTLINE
 }

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

@@ -6,6 +6,7 @@ import java.util.Locale;
 public enum NodeType implements Serializable {
     DOCUMENT,
     SECTION,
+    SUPER_SECTION,
     HEADLINE,
     PARAGRAPH,
     TABLE,

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

@@ -14,7 +14,7 @@ import lombok.NoArgsConstructor;
 public class SimplifiedSectionText {
 
     @Schema(description = "The number of this Section. This is used to map the simplified section text back to the original Section.")
-    private int sectionNumber;
+    private String sectionNumber;
     @Schema(description = "The text in this Section.")
     private String text;
 

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

@@ -19,6 +19,16 @@ public class SimplifiedText {
     @Schema(description = "Number of pages in the entire document.")
     private int numberOfPages;
     @Schema(description = "A List of simplified Sections, which contains almost exclusively text.")
+    @Builder.Default
     private List<SimplifiedSectionText> sectionTexts = new ArrayList<>();
+    @Schema(description = "A list of the main section numbers ")
+    @Builder.Default
+    private List<String> mainSectionNumbers = new ArrayList<>();
+    @Schema(description = "A list of the header section numbers ")
+    @Builder.Default
+    private List<String> headerSectionNumbers = new ArrayList<>();
+    @Schema(description = "A list of the footer section numbers ")
+    @Builder.Default
+    private List<String> footerSectionNumbers = new ArrayList<>();
 
 }

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

@@ -1,5 +1,7 @@
 package com.knecon.fforesight.service.layoutparser.internal.api.data.taas;
 
+import java.util.List;
+
 import io.swagger.v3.oas.annotations.media.Schema;
 import lombok.AllArgsConstructor;
 import lombok.Builder;
@@ -13,6 +15,8 @@ public class StructureObject {
 
     @Schema(description = "The ID of this StructureObject.")
     Integer structureObjectNumber;
+    @Schema(description = "The Tree ID of this StructureObject.")
+    List<Integer> treeId;
     @Schema(description = "This value indicates the start of the string offsets in this Object, with respect to the reading order.")
     int page;
     @Schema(description = "This stringOffset indicates the start of the string offsets in this Object, with respect to the reading order of the entire document. It is equal to the previous' StructureObject stringOffset + its length.")

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

@@ -2,6 +2,9 @@ package com.knecon.fforesight.service.layoutparser.internal.api.queue;
 
 public class LayoutParsingQueueNames {
 
-    public static final String LAYOUT_PARSING_REQUEST_QUEUE = "layout_parsing_request_queue";
-    public static final String LAYOUT_PARSING_FINISHED_EVENT_QUEUE = "layout_parsing_response_queue";
+    public static final String LAYOUT_PARSING_REQUEST_QUEUE_PREFIX = "layout_parsing_request";
+    public static final String LAYOUT_PARSING_REQUEST_EXCHANGE = "layout_parsing_request_exchange";
+    public static final String LAYOUT_PARSING_RESPONSE_QUEUE_PREFIX = "layout_parsing_response";
+    public static final String LAYOUT_PARSING_RESPONSE_EXCHANGE = "layout_parsing_response_exchange";
+    public static final String LAYOUT_PARSING_DLQ = "layout_parsing_error";
 }

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

@@ -19,7 +19,6 @@ 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.")//
@@ -37,9 +36,12 @@ public record LayoutParsingRequest(
         @NonNull String positionBlockFileStorageId,//
         @Schema(description = "Path where the Document Pages File will be stored.")//
         @NonNull String pageFileStorageId,//
+        @Schema(description = "Path where the Document Markdown File will be stored.")//
+        Optional<String> documentMarkdownFileStorageId,//
         @Schema(description = "Path where the Simplified Text File will be stored.")//
         @NonNull String simplifiedTextStorageId,//
         @Schema(description = "Path where the Viewer Document PDF will be stored.")//
-        @NonNull String viewerDocumentStorageId) {
+        @NonNull String viewerDocumentStorageId
+) {
 
 }

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

@@ -4,6 +4,7 @@ public enum LayoutParsingType {
     REDACT_MANAGER,
     REDACT_MANAGER_OLD,
     REDACT_MANAGER_PARAGRAPH_DEBUG,
+    REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH,
     DOCUMINE,
     DOCUMINE_OLD,
     CLARIFYND,

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

@@ -16,7 +16,7 @@ dependencies {
         exclude("org.springframework.boot", "spring-boot-starter-security")
         exclude("org.springframework.boot", "spring-boot-starter-validation")
     }
-    implementation("com.knecon.fforesight:tenant-commons:0.21.0")
+    implementation("com.knecon.fforesight:tenant-commons:0.29.0")
     implementation("com.iqser.red.commons:storage-commons:2.45.0")
 
     implementation("org.apache.pdfbox:pdfbox:${pdfBoxVersion}")
@@ -26,4 +26,9 @@ dependencies {
     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")
+    implementation("org.commonmark:commonmark:0.22.0")
+    implementation("org.commonmark:commonmark-ext-gfm-tables:0.22.0")
+    implementation("com.pdftron:PDFNet:10.11.0")
+    implementation("org.apache.commons:commons-text:1.12.0")
+
 }

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

@@ -2,12 +2,14 @@ package com.knecon.fforesight.service.layoutparser.processor;
 
 import static java.lang.String.format;
 
+import java.awt.geom.Point2D;
 import java.awt.geom.Rectangle2D;
 import java.io.File;
 import java.io.IOException;
 import java.nio.file.Files;
 import java.util.ArrayList;
 import java.util.HashMap;
+import java.util.HashSet;
 import java.util.List;
 import java.util.Map;
 import java.util.concurrent.atomic.AtomicReference;
@@ -23,12 +25,18 @@ import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.No
 import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingFinishedEvent;
 import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingRequest;
 import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingType;
+import com.knecon.fforesight.service.layoutparser.processor.markdown.MarkdownMapper;
 import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationDocument;
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.ImageType;
 import com.knecon.fforesight.service.layoutparser.processor.model.image.ClassifiedImage;
+import com.knecon.fforesight.service.layoutparser.processor.model.outline.OutlineExtractorService;
+import com.knecon.fforesight.service.layoutparser.processor.model.outline.OutlineObject;
+import com.knecon.fforesight.service.layoutparser.processor.model.outline.OutlineValidationService;
+import com.knecon.fforesight.service.layoutparser.processor.model.outline.TOCEnrichmentService;
+import com.knecon.fforesight.service.layoutparser.processor.model.outline.TableOfContents;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.Cell;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRulings;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
@@ -46,6 +54,7 @@ import com.knecon.fforesight.service.layoutparser.processor.services.SectionsBui
 import com.knecon.fforesight.service.layoutparser.processor.services.SimplifiedSectionTextService;
 import com.knecon.fforesight.service.layoutparser.processor.services.TableExtractionService;
 import com.knecon.fforesight.service.layoutparser.processor.services.TextRulingsClassifier;
+import com.knecon.fforesight.service.layoutparser.processor.services.blockification.BlockificationPostprocessingService;
 import com.knecon.fforesight.service.layoutparser.processor.services.blockification.DocstrumBlockificationService;
 import com.knecon.fforesight.service.layoutparser.processor.services.blockification.DocuMineBlockificationService;
 import com.knecon.fforesight.service.layoutparser.processor.services.blockification.RedactManagerBlockificationService;
@@ -61,6 +70,7 @@ import com.knecon.fforesight.service.layoutparser.processor.services.parsing.PDF
 import com.knecon.fforesight.service.layoutparser.processor.services.visualization.LayoutGridService;
 import com.knecon.fforesight.service.layoutparser.processor.utils.MarkedContentUtils;
 import com.knecon.fforesight.service.layoutparser.processor.utils.PageInformation;
+import com.knecon.fforesight.service.layoutparser.processor.utils.TextPositionOperations;
 
 import io.micrometer.observation.Observation;
 import io.micrometer.observation.ObservationRegistry;
@@ -90,12 +100,16 @@ public class LayoutParsingPipeline {
     TableExtractionService tableExtractionService;
     DocuMineBlockificationService docuMineBlockificationService;
     RedactManagerBlockificationService redactManagerBlockificationService;
+    BlockificationPostprocessingService blockificationPostprocessingService;
     DocstrumBlockificationService docstrumBlockificationService;
     LayoutGridService layoutGridService;
     ObservationRegistry observationRegistry;
     VisualLayoutParsingAdapter visualLayoutParsingAdapter;
     ClarifyndClassificationService clarifyndClassificationService;
     GraphicExtractorService graphicExtractorService;
+    OutlineExtractorService outlineExtractorService;
+    OutlineValidationService outlineValidationService;
+    TOCEnrichmentService tocEnrichmentService;
     LayoutparserSettings settings;
 
 
@@ -107,20 +121,12 @@ public class LayoutParsingPipeline {
         File originFile = layoutParsingStorageService.getOriginFile(layoutParsingRequest.originFileStorageId());
         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()) {
-            imageServiceResponse = layoutParsingStorageService.getImagesFile(layoutParsingRequest.imagesFileStorageId().get());
-        }
-
-        TableServiceResponse tableServiceResponse = new TableServiceResponse();
-        if (layoutParsingRequest.tablesFileStorageId().isPresent()) {
-            tableServiceResponse = layoutParsingStorageService.getTablesFile(layoutParsingRequest.tablesFileStorageId().get());
-        }
+        VisualLayoutParsingResponse visualLayoutParsingResponse = layoutParsingRequest.visualLayoutParsingFileId()
+                .map(layoutParsingStorageService::getVisualLayoutParsingFile).orElse(new VisualLayoutParsingResponse());
+        ImageServiceResponse imageServiceResponse = layoutParsingRequest.imagesFileStorageId()
+                .map(layoutParsingStorageService::getImagesFile).orElse(new ImageServiceResponse());
+        TableServiceResponse tableServiceResponse = layoutParsingRequest.tablesFileStorageId()
+                .map(layoutParsingStorageService::getTablesFile).orElse(new TableServiceResponse());
 
         ClassificationDocument classificationDocument = parseLayout(settings.getLayoutParsingTypeOverride() == null //
                                                                             ? layoutParsingRequest.layoutParsingType() : settings.getLayoutParsingTypeOverride(),
@@ -142,10 +148,13 @@ public class LayoutParsingPipeline {
         log.info("Storing resulting files for {}", layoutParsingRequest.identifier());
 
         layoutParsingStorageService.storeDocumentData(layoutParsingRequest, DocumentDataMapper.toDocumentData(documentGraph));
+        if (layoutParsingRequest.documentMarkdownFileStorageId().isPresent()) {
+            layoutParsingStorageService.storeMarkdownFile(layoutParsingRequest.documentMarkdownFileStorageId().get(), new MarkdownMapper().toMarkdownContent(documentGraph));
+        }
         layoutParsingStorageService.storeSimplifiedText(layoutParsingRequest, simplifiedSectionTextService.toSimplifiedText(documentGraph));
         layoutParsingStorageService.storeViewerDocument(layoutParsingRequest, viewerDocumentFile);
 
-        if (layoutParsingRequest.layoutParsingType().equals(LayoutParsingType.CLARIFYND)) {
+        if (layoutParsingRequest.researchDocumentStorageId() != null) {
             log.info("Building research document data for {}", layoutParsingRequest.identifier());
             var researchDocumentData = TaasDocumentDataMapper.fromDocument(documentGraph);
             layoutParsingStorageService.storeResearchDocumentData(layoutParsingRequest, researchDocumentData);
@@ -199,15 +208,15 @@ public class LayoutParsingPipeline {
 
     private String buildSemanticNodeCountMessage(int numberOfPages, Map<NodeType, Long> semanticNodeCounts) {
 
-        return String.format("%d pages with %d sections, %d headlines, %d paragraphs, %d tables with %d cells, %d headers, and %d footers parsed",
-                             numberOfPages,
-                             semanticNodeCounts.get(NodeType.SECTION) == null ? 0 : semanticNodeCounts.get(NodeType.SECTION),
-                             semanticNodeCounts.get(NodeType.HEADLINE) == null ? 0 : semanticNodeCounts.get(NodeType.HEADLINE),
-                             semanticNodeCounts.get(NodeType.PARAGRAPH) == null ? 0 : semanticNodeCounts.get(NodeType.PARAGRAPH),
-                             semanticNodeCounts.get(NodeType.TABLE) == null ? 0 : semanticNodeCounts.get(NodeType.TABLE),
-                             semanticNodeCounts.get(NodeType.TABLE_CELL) == null ? 0 : semanticNodeCounts.get(NodeType.TABLE_CELL),
-                             semanticNodeCounts.get(NodeType.HEADER) == null ? 0 : semanticNodeCounts.get(NodeType.HEADER),
-                             semanticNodeCounts.get(NodeType.FOOTER) == null ? 0 : semanticNodeCounts.get(NodeType.FOOTER));
+        return format("%d pages with %d sections, %d headlines, %d paragraphs, %d tables with %d cells, %d headers, and %d footers parsed",
+                      numberOfPages,
+                      semanticNodeCounts.get(NodeType.SECTION) == null ? 0 : semanticNodeCounts.get(NodeType.SECTION),
+                      semanticNodeCounts.get(NodeType.HEADLINE) == null ? 0 : semanticNodeCounts.get(NodeType.HEADLINE),
+                      semanticNodeCounts.get(NodeType.PARAGRAPH) == null ? 0 : semanticNodeCounts.get(NodeType.PARAGRAPH),
+                      semanticNodeCounts.get(NodeType.TABLE) == null ? 0 : semanticNodeCounts.get(NodeType.TABLE),
+                      semanticNodeCounts.get(NodeType.TABLE_CELL) == null ? 0 : semanticNodeCounts.get(NodeType.TABLE_CELL),
+                      semanticNodeCounts.get(NodeType.HEADER) == null ? 0 : semanticNodeCounts.get(NodeType.HEADER),
+                      semanticNodeCounts.get(NodeType.FOOTER) == null ? 0 : semanticNodeCounts.get(NodeType.FOOTER));
     }
 
 
@@ -222,16 +231,24 @@ public class LayoutParsingPipeline {
 
         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();
 
         if (settings.isDebug() || identifier.containsKey("debug")) {
-            classificationDocument.getVisualizations().setActive(true);
+            classificationDocument.getLayoutDebugLayer().setActive(true);
         }
 
         List<ClassificationPage> classificationPages = new ArrayList<>();
+        OutlineObject lastProcessedOutlineObject = null;
+
+        // parsing the structure elements could be useful as well
+        if (layoutParsingType != LayoutParsingType.REDACT_MANAGER_OLD) {
+            classificationDocument.setOutlineObjectTree(outlineExtractorService.getOutlineObjectTree(originDocument));
+        }
 
         long pageCount = originDocument.getNumberOfPages();
 
@@ -255,12 +272,14 @@ public class LayoutParsingPipeline {
             stripper.setStartPage(pageNumber);
             stripper.setEndPage(pageNumber);
             stripper.setPdpage(pdPage);
-            if (layoutParsingType.equals(LayoutParsingType.DOCUMINE_OLD)) {
-                stripper.setSortByPosition(true);
-            }
             stripper.getText(originDocument);
             List<TextPositionSequence> words = stripper.getTextPositionSequences();
-            classificationDocument.getVisualizations().addTextVisualizations(words, pageNumber);
+            if (layoutParsingType.equals(LayoutParsingType.DOCUMINE_OLD)) {
+                var lines = TextPositionOperations.groupByLine(new HashSet<>(words));
+                classificationDocument.getLayoutDebugLayer().addLineVisualizationsFromNestedTextPosition(lines, pageNumber);
+                words = TextPositionOperations.sortLines(lines);
+            }
+            classificationDocument.getLayoutDebugLayer().addTextVisualizations(words, pageNumber);
 
             PDRectangle pdr = pdPage.getMediaBox();
 
@@ -268,12 +287,12 @@ public class LayoutParsingPipeline {
             boolean isLandscape = pdr.getWidth() > pdr.getHeight() && (rotation == 0 || rotation == 180) || pdr.getHeight() > pdr.getWidth() && (rotation == 90 || rotation == 270);
 
             PDRectangle cropbox = pdPage.getCropBox();
-            classificationDocument.getVisualizations().addRulingVisualization(stripper.getRulings(), pageNumber);
+            classificationDocument.getLayoutDebugLayer().addRulingVisualization(stripper.getRulings(), pageNumber);
             CleanRulings cleanRulings = rulingCleaningService.deduplicateAndStraightenRulings(pdfTableCells.get(pageNumber), stripper.getRulings());
 
             PageInformation pageInformation = PageInformation.fromPDPage(pageNumber, pdPage);
             List<Cell> emptyTableCells = TableExtractionService.findCells(cleanRulings.getHorizontals(), cleanRulings.getVerticals(), pageInformation);
-            classificationDocument.getVisualizations().addCellVisualizations(emptyTableCells, pageNumber);
+            classificationDocument.getLayoutDebugLayer().addCellVisualizations(emptyTableCells, pageNumber);
 
             TextRulingsClassifier.classifyUnderlinedAndStrikethroughText(words, cleanRulings);
 
@@ -281,17 +300,21 @@ public class LayoutParsingPipeline {
 
             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()))
+                                    .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_OLD ->
-                        redactManagerBlockificationService.blockify(stripper.getTextPositionSequences(), cleanRulings, classificationDocument.getVisualizations());
+                        redactManagerBlockificationService.blockify(stripper.getTextPositionSequences(), cleanRulings, classificationDocument.getLayoutDebugLayer());
                 case DOCUMINE_OLD -> docuMineBlockificationService.blockify(words, cleanRulings);
-                case DOCUMINE, REDACT_MANAGER, REDACT_MANAGER_PARAGRAPH_DEBUG ->
-                        docstrumBlockificationService.blockify(words, cleanRulings, true, classificationDocument.getVisualizations(), layoutParsingType);
+                case DOCUMINE, REDACT_MANAGER, REDACT_MANAGER_PARAGRAPH_DEBUG, REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH ->
+                        docstrumBlockificationService.blockify(words, cleanRulings, true, classificationDocument.getLayoutDebugLayer(), layoutParsingType);
                 case CLARIFYND, CLARIFYND_PARAGRAPH_DEBUG ->
-                        docstrumBlockificationService.blockify(words, cleanRulings, false, classificationDocument.getVisualizations(), layoutParsingType);
+                        docstrumBlockificationService.blockify(words, cleanRulings, false, classificationDocument.getLayoutDebugLayer(), layoutParsingType);
             };
 
             classificationPage.setCleanRulings(cleanRulings);
@@ -301,7 +324,22 @@ public class LayoutParsingPipeline {
             classificationPage.setPageWidth(cropbox.getWidth());
             classificationPage.setPageHeight(cropbox.getHeight());
 
-            classificationDocument.getVisualizations().addMarkedContentVisualizations(stripper.getMarkedContents(), pageNumber);
+            if (layoutParsingType != LayoutParsingType.REDACT_MANAGER_OLD) {
+                List<OutlineObject> outlineObjects = classificationDocument.getOutlineObjectTree().getOutlineObjectsPerPage().getOrDefault(pageNumber, new ArrayList<>());
+
+                OutlineObject notFoundOutlineObject = null;
+                if (lastProcessedOutlineObject != null && !lastProcessedOutlineObject.isFound()) {
+                    lastProcessedOutlineObject.resetPoint();
+                    notFoundOutlineObject = lastProcessedOutlineObject;
+                }
+                if (!outlineObjects.isEmpty()) {
+                    classificationPage.setOutlineObjects(outlineObjects);
+                    lastProcessedOutlineObject = blockificationPostprocessingService.sanitizeOutlineBlocks(classificationPage, notFoundOutlineObject);
+                }
+                classificationDocument.getLayoutDebugLayer().addOutlineObjects(outlineObjects, pageInformation);
+            }
+
+            classificationDocument.getLayoutDebugLayer().addMarkedContentVisualizations(stripper.getMarkedContents(), pageNumber);
             // MarkedContent needs to be converted at this point, otherwise it leads to GC Problems in Pdfbox.
             classificationPage.setMarkedContentBboxPerType(convertMarkedContents(stripper.getMarkedContents()));
 
@@ -332,24 +370,37 @@ public class LayoutParsingPipeline {
         log.info("Calculating BodyTextFrame for {}", identifier);
         bodyTextFrameService.setBodyTextFrames(classificationDocument, layoutParsingType);
         for (ClassificationPage page : classificationDocument.getPages()) {
-            classificationDocument.getVisualizations().addCleanRulingVisualization(page.getCleanRulings(), page.getPageNumber());
+            classificationDocument.getLayoutDebugLayer().addCleanRulingVisualization(page.getCleanRulings(), page.getPageNumber());
         }
         log.info("Classify TextBlocks for {}", identifier);
         switch (layoutParsingType) {
-            case REDACT_MANAGER, REDACT_MANAGER_PARAGRAPH_DEBUG, REDACT_MANAGER_OLD, CLARIFYND_PARAGRAPH_DEBUG ->
+            case REDACT_MANAGER, REDACT_MANAGER_PARAGRAPH_DEBUG, REDACT_MANAGER_OLD, CLARIFYND_PARAGRAPH_DEBUG, REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH ->
                     redactManagerClassificationService.classifyDocument(classificationDocument);
             case DOCUMINE_OLD, DOCUMINE -> docuMineClassificationService.classifyDocument(classificationDocument);
             case CLARIFYND -> clarifyndClassificationService.classifyDocument(classificationDocument);
         }
 
+        if (layoutParsingType.equals(LayoutParsingType.DOCUMINE_OLD)) {
+            for (ClassificationPage page : classificationDocument.getPages()) {
+                docuMineBlockificationService.mergeblocks(page, page.getCleanRulings().withoutTextRulings(), 0, 10);
+            }
+        }
+
+        List<TextPageBlock> headlines = classificationDocument.getPages()
+                .stream()
+                .flatMap(classificationPage -> classificationPage.getTextBlocks()
+                        .stream()
+                        .filter(tb -> tb instanceof TextPageBlock && tb.getClassification() != null && tb.getClassification().isHeadline())
+                        .map(tb -> (TextPageBlock) tb))
+                .toList();
+        TableOfContents tableOfContents = outlineValidationService.createToC(headlines);
+        classificationDocument.setTableOfContents(tableOfContents);
+
         log.info("Building Sections for {}", identifier);
 
         switch (layoutParsingType) {
             case CLARIFYND_PARAGRAPH_DEBUG, REDACT_MANAGER_PARAGRAPH_DEBUG -> sectionsBuilderService.buildParagraphDebugSections(classificationDocument);
-            default -> {
-                sectionsBuilderService.buildSections(classificationDocument);
-                sectionsBuilderService.addImagesToSections(classificationDocument);
-            }
+            default -> tocEnrichmentService.assignSectionBlocksAndImages(classificationDocument);
         }
 
         return classificationDocument;

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

@@ -5,7 +5,7 @@ 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 com.knecon.fforesight.service.viewerdoc.service.PDFTronViewerDocumentService;
 
 import io.micrometer.observation.ObservationRegistry;
 
@@ -13,12 +13,11 @@ import io.micrometer.observation.ObservationRegistry;
 @ComponentScan
 public class LayoutParsingServiceProcessorConfiguration {
 
-
     @Bean
     @Autowired
-    public ViewerDocumentService viewerDocumentService(ObservationRegistry registry) {
+    public PDFTronViewerDocumentService viewerDocumentService(ObservationRegistry registry) {
 
-        return new ViewerDocumentService(registry);
+        return new PDFTronViewerDocumentService(registry);
     }
 
 }

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

@@ -1,9 +1,11 @@
 package com.knecon.fforesight.service.layoutparser.processor;
 
+import java.io.ByteArrayInputStream;
 import java.io.File;
 import java.io.FileInputStream;
 import java.io.IOException;
 import java.io.InputStream;
+import java.nio.charset.StandardCharsets;
 import java.nio.file.Files;
 import java.nio.file.Path;
 import java.nio.file.Paths;
@@ -21,6 +23,7 @@ import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsi
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.image.ImageServiceResponse;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.TableServiceResponse;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.VisualLayoutParsingResponse;
+import com.knecon.fforesight.service.viewerdoc.service.ViewerDocVersioningUtility;
 import com.knecon.fforesight.tenantcommons.TenantContext;
 
 import io.micrometer.observation.annotation.Observed;
@@ -36,6 +39,7 @@ public class LayoutParsingStorageService {
     private final StorageService storageService;
     private final ObjectMapper objectMapper;
 
+
     @Observed(name = "LayoutParsingStorageService", contextualName = "get-origin-file")
     public File getOriginFile(String storageId) throws IOException {
 
@@ -53,11 +57,18 @@ public class LayoutParsingStorageService {
         }
         File tempFile = createTempFile("viewerDocument", ".pdf");
         storageService.downloadTo(TenantContext.getTenantId(), storageId, tempFile);
+
+        if (!ViewerDocVersioningUtility.isCurrentVersion(tempFile)) {
+            assert tempFile.delete();
+            return Optional.empty();
+        }
+
         return Optional.of(tempFile);
     }
 
 
-    public ImageServiceResponse getImagesFile(String storageId) throws IOException {
+    @SneakyThrows
+    public ImageServiceResponse getImagesFile(String storageId) {
 
         try (InputStream inputStream = getObject(storageId)) {
 
@@ -68,7 +79,8 @@ public class LayoutParsingStorageService {
     }
 
 
-    public TableServiceResponse getTablesFile(String storageId) throws IOException {
+    @SneakyThrows
+    public TableServiceResponse getTablesFile(String storageId) {
 
         try (var tableClassificationStream = getObject(storageId)) {
 
@@ -78,11 +90,12 @@ public class LayoutParsingStorageService {
         }
     }
 
-    public VisualLayoutParsingResponse getVisualLayoutParsingFile(String storageId) throws IOException {
+
+@SneakyThrows
+    public VisualLayoutParsingResponse getVisualLayoutParsingFile(String storageId) {
 
         try (InputStream inputStream = getObject(storageId)) {
-            VisualLayoutParsingResponse visualLayoutParsingResponse = objectMapper.readValue(inputStream, VisualLayoutParsingResponse.class);
-            return visualLayoutParsingResponse;
+            return objectMapper.readValue(inputStream, VisualLayoutParsingResponse.class);
         }
     }
 
@@ -154,4 +167,16 @@ public class LayoutParsingStorageService {
         }
     }
 
+
+    @SneakyThrows
+    @Observed(name = "LayoutParsingStorageService", contextualName = "store-markdown-file")
+    public void storeMarkdownFile(String markdownFileStorageId, String markdownContent) {
+
+        try (InputStream inputStream = new ByteArrayInputStream(markdownContent.getBytes(StandardCharsets.UTF_8))) {
+
+            storageService.storeObject(TenantContext.getTenantId(), markdownFileStorageId, inputStream);
+        }
+
+    }
+
 }

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

@@ -17,4 +17,5 @@ public class LayoutparserSettings {
 
     boolean debug;
     LayoutParsingType layoutParsingTypeOverride;
+    String pdftronLicense;
 }

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

@@ -1,6 +1,7 @@
 package com.knecon.fforesight.service.layoutparser.processor.docstrum;
 
 import java.util.ArrayList;
+import java.util.EnumMap;
 import java.util.List;
 import java.util.stream.Collectors;
 
@@ -18,7 +19,7 @@ import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRul
 import com.knecon.fforesight.service.layoutparser.processor.model.text.RedTextPosition;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextDirection;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
-import com.knecon.fforesight.service.layoutparser.processor.visualization.LayoutparsingVisualizations;
+import com.knecon.fforesight.service.layoutparser.processor.visualization.LayoutDebugLayer;
 
 import lombok.RequiredArgsConstructor;
 
@@ -26,6 +27,7 @@ import lombok.RequiredArgsConstructor;
 @RequiredArgsConstructor
 public class DocstrumSegmentationService {
 
+    public static final double SAME_DIRECTION_THRESHOLD = 0.9;
     private final NearestNeighbourService nearestNeighbourService;
     private final SpacingService spacingService;
     private final LineBuilderService lineBuilderService;
@@ -33,19 +35,50 @@ public class DocstrumSegmentationService {
     private final ReadingOrderService readingOrderService;
 
 
-    public List<Zone> segmentPage(List<TextPositionSequence> textPositions, boolean xyOrder, CleanRulings usedRulings, LayoutparsingVisualizations visualizations) {
+    public List<Zone> segmentPage(List<TextPositionSequence> textPositions, boolean xyOrder, CleanRulings usedRulings, LayoutDebugLayer visualizations) {
 
-        List<Zone> zones = new ArrayList<>();
-        zones.addAll(computeZones(textPositions, usedRulings, visualizations, TextDirection.ZERO));
-        zones.addAll(computeZones(textPositions, usedRulings, visualizations, TextDirection.QUARTER_CIRCLE));
-        zones.addAll(computeZones(textPositions, usedRulings, visualizations, TextDirection.HALF_CIRCLE));
-        zones.addAll(computeZones(textPositions, usedRulings, visualizations, TextDirection.THREE_QUARTER_CIRCLE));
+        EnumMap<TextDirection, Integer> directionCounts = new EnumMap<>(TextDirection.class);
 
-        return readingOrderService.resolve(zones, xyOrder);
+        List<Zone> newZones = computeZones(textPositions, usedRulings, visualizations, TextDirection.ZERO);
+        directionCounts.put(TextDirection.ZERO, newZones.size());
+        List<Zone> zones = new ArrayList<>(newZones);
+
+        newZones = computeZones(textPositions, usedRulings, visualizations, TextDirection.QUARTER_CIRCLE);
+        directionCounts.put(TextDirection.QUARTER_CIRCLE, newZones.size());
+        zones.addAll(newZones);
+
+        newZones = computeZones(textPositions, usedRulings, visualizations, TextDirection.HALF_CIRCLE);
+        directionCounts.put(TextDirection.HALF_CIRCLE, newZones.size());
+        zones.addAll(newZones);
+
+        newZones = computeZones(textPositions, usedRulings, visualizations, TextDirection.THREE_QUARTER_CIRCLE);
+        directionCounts.put(TextDirection.THREE_QUARTER_CIRCLE, newZones.size());
+        zones.addAll(newZones);
+
+        return readingOrderService.resolve(zones, xyOrder, mostSameDirection(directionCounts));
     }
 
 
-    private List<Zone> computeZones(List<TextPositionSequence> textPositions, CleanRulings rulings, LayoutparsingVisualizations visualizations, TextDirection direction) {
+    private boolean mostSameDirection(EnumMap<TextDirection, Integer> directionCounts) {
+
+        int total = directionCounts.values()
+                .stream()
+                .mapToInt(i -> i).sum();
+
+        if ((double) directionCounts.get(TextDirection.ZERO) / total > SAME_DIRECTION_THRESHOLD) {
+            return true;
+        } else if ((double) directionCounts.get(TextDirection.QUARTER_CIRCLE) / total > SAME_DIRECTION_THRESHOLD) {
+            return true;
+        } else if ((double) directionCounts.get(TextDirection.HALF_CIRCLE) / total > SAME_DIRECTION_THRESHOLD) {
+            return true;
+        } else if ((double) directionCounts.get(TextDirection.THREE_QUARTER_CIRCLE) / total > SAME_DIRECTION_THRESHOLD) {
+            return true;
+        }
+        return false;
+    }
+
+
+    private List<Zone> computeZones(List<TextPositionSequence> textPositions, CleanRulings rulings, LayoutDebugLayer visualizations, TextDirection direction) {
 
         List<RedTextPosition> positions = textPositions.stream()
                 .filter(t -> t.getDir() == direction)

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

@@ -15,10 +15,16 @@ public class AngleFilter {
 
     public boolean matches(Neighbor neighbor) {
 
+        return matches(neighbor.getAngle());
+    }
+
+
+    public boolean matches(double angle) {
+
         if (lowerAngle <= upperAngle) {
-            return lowerAngle <= neighbor.getAngle() && neighbor.getAngle() < upperAngle;
+            return lowerAngle <= angle && angle < upperAngle;
         } else {
-            return lowerAngle <= neighbor.getAngle() || neighbor.getAngle() < upperAngle;
+            return lowerAngle <= angle || angle < upperAngle;
         }
     }
 

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

@@ -7,8 +7,12 @@ import java.util.List;
 import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
 
 import lombok.Data;
+import lombok.NoArgsConstructor;
+import lombok.experimental.SuperBuilder;
 
 @Data
+@SuperBuilder
+@NoArgsConstructor
 public abstract class BoundingBox {
 
     // Java coordinate system: (0, 0) is always upper left, x is increasing left to right and y is increasing from top to bottom.
@@ -19,7 +23,7 @@ public abstract class BoundingBox {
     // This rotates completely in 90 degree steps with page rotation.
     // Needs to be used when writing to a PDF.
     // Also, these are definitely correct and should be used whenever possible.
-    protected Rectangle2D bBoxInitialUserSpace;
+    protected Rectangle2D bBoxPdf;
 
     protected static final float VERTICAL_COMPARISON_THRESHOLD = 0.4f;
 
@@ -50,25 +54,25 @@ public abstract class BoundingBox {
 
     public double getPdfMinX() {
 
-        return bBoxInitialUserSpace.getMinX();
+        return bBoxPdf.getMinX();
     }
 
 
     public double getPdfMaxX() {
 
-        return bBoxInitialUserSpace.getMaxX();
+        return bBoxPdf.getMaxX();
     }
 
 
     public double getPdfMinY() {
 
-        return bBoxInitialUserSpace.getMinY();
+        return bBoxPdf.getMinY();
     }
 
 
     public double getPdfMaxY() {
 
-        return bBoxInitialUserSpace.getMaxY();
+        return bBoxPdf.getMaxY();
     }
 
 
@@ -129,13 +133,31 @@ public abstract class BoundingBox {
     }
 
 
-    public boolean intersectsY(BoundingBox other) {
+    public boolean intersectsX(BoundingBox other, float threshold) {
+
+        return this.getX() - threshold <= other.getMaxX() && this.getMaxX() + threshold >= other.getX();
+    }
+
+
+    public boolean intersectsPdf(BoundingBox other) {
+
+        return this.intersectsXPdf(other) && this.intersectsYPdf(other);
+    }
+
+
+    public boolean intersectsPdf(BoundingBox other, float yThreshold, float xThreshold) {
+
+        return this.intersectsXPdf(other, xThreshold) && this.intersectsYPdf(other, yThreshold);
+    }
+
+
+    public boolean intersectsYPdf(BoundingBox other) {
 
         return this.getPdfMinY() <= other.getPdfMaxY() && this.getPdfMaxY() >= other.getPdfMinY();
     }
 
 
-    public boolean intersectsYJava(BoundingBox other) {
+    public boolean intersectsY(BoundingBox other) {
 
         return this.getY() <= other.getMaxY() && this.getMaxY() >= other.getY();
     }
@@ -143,25 +165,31 @@ public abstract class BoundingBox {
 
     public boolean intersectsY(BoundingBox other, float threshold) {
 
+        return this.getY() - threshold <= other.getMaxY() && this.getMaxY() + threshold >= other.getY();
+    }
+
+
+    public boolean intersectsYPdf(BoundingBox other, float threshold) {
+
         return this.getPdfMinY() - threshold <= other.getPdfMaxY() && this.getPdfMaxY() + threshold >= other.getPdfMinY();
     }
 
 
-    public boolean intersectsX(BoundingBox other) {
+    public boolean intersectsXPdf(BoundingBox other) {
 
         return this.getPdfMinX() <= other.getPdfMaxX() && this.getPdfMaxX() >= other.getPdfMinX();
     }
 
 
-    public boolean intersectsXJava(BoundingBox other) {
+    public boolean intersectsX(BoundingBox other) {
 
         return this.getX() <= other.getMaxX() && this.getMaxX() >= other.getMinX();
     }
 
 
-    public boolean intersectsX(BoundingBox other, float threshold) {
+    public boolean intersectsXPdf(BoundingBox other, float threshold) {
 
-        return this.getPdfMinX() - threshold <= other.getMaxX() && this.getMaxX() + threshold >= other.getPdfMinX();
+        return this.getPdfMinX() - threshold <= other.getPdfMaxX() && this.getMaxX() + threshold >= other.getPdfMinX();
     }
 
 
@@ -170,8 +198,8 @@ public abstract class BoundingBox {
         this.bBox = components.stream()
                 .map(BoundingBox::getBBox)
                 .collect(RectangleTransformations.collectBBox());
-        this.bBoxInitialUserSpace = components.stream()
-                .map(BoundingBox::getBBoxInitialUserSpace)
+        this.bBoxPdf = components.stream()
+                .map(BoundingBox::getBBoxPdf)
                 .collect(RectangleTransformations.collectBBox());
     }
 
@@ -229,25 +257,25 @@ public abstract class BoundingBox {
 
     public boolean rightOf(BoundingBox other) {
 
-        return this.intersectsYJava(other) && other.getMaxX() <= this.getMinX();
+        return this.intersectsY(other) && other.getMaxX() <= this.getMinX();
     }
 
 
     public boolean leftOf(BoundingBox other) {
 
-        return this.intersectsYJava(other) && other.getMinX() >= this.getMaxX();
+        return this.intersectsY(other) && other.getMinX() >= this.getMaxX();
     }
 
 
     public boolean isAbove(BoundingBox other) {
 
-        return this.intersectsXJava(other) && other.getMinY() >= this.getMaxY();
+        return this.intersectsX(other) && other.getMinY() >= this.getMaxY();
     }
 
 
     public boolean isBelow(BoundingBox other) {
 
-        return this.intersectsXJava(other) && this.getMinY() >= other.getMaxY();
+        return this.intersectsX(other) && this.getMinY() >= other.getMaxY();
     }
 
 }

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

@@ -35,7 +35,7 @@ public class Character {
 
     public double getHeight() {
 
-        return textPosition.getHeightDir();
+        return textPosition.getHeightDirAdj();
     }
 
 
@@ -65,9 +65,9 @@ public class Character {
         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[1] = c * (x + textPosition.getWidthDirAdj()) - s * (y + textPosition.getHeightDirAdj());
         xs[2] = c * other.x - s * other.y;
-        xs[3] = c * (other.x + other.textPosition.getWidthDirAdj()) - s * (other.y + other.textPosition.getHeightDir());
+        xs[3] = c * (other.x + other.textPosition.getWidthDirAdj()) - s * (other.y + other.textPosition.getHeightDirAdj());
         boolean overlapping = xs[1] >= xs[2] && xs[3] >= xs[0];
         Arrays.sort(xs);
         return Math.abs(xs[2] - xs[1]) * (overlapping ? 1 : -1);

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

@@ -1,18 +1,28 @@
 package com.knecon.fforesight.service.layoutparser.processor.docstrum.model;
 
+import static com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence.BOLD;
+import static com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence.BOLD_ITALIC;
+import static com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence.ITALIC;
+import static com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence.STANDARD;
+
 import java.util.ArrayList;
 import java.util.Arrays;
+import java.util.Comparator;
+import java.util.EnumMap;
 import java.util.List;
+import java.util.Map;
+import java.util.concurrent.atomic.AtomicInteger;
 
 import com.knecon.fforesight.service.layoutparser.processor.docstrum.utils.FastAtan2;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.FontStyle;
 
 import lombok.Data;
 import lombok.EqualsAndHashCode;
 
 @Data
-@EqualsAndHashCode(onlyExplicitlyIncluded = true)
-public class Line extends BoundingBox {
+@EqualsAndHashCode(onlyExplicitlyIncluded = true, callSuper = false)
+public class Line extends TextBoundingBox {
 
     private static final double WORD_DISTANCE_MULTIPLIER = 0.18;
 
@@ -28,6 +38,8 @@ public class Line extends BoundingBox {
 
     private final double height;
 
+    private FontStyle fontStyle;
+
     private final List<Character> characters;
     private final List<TextPositionSequence> words = new ArrayList<>();
 
@@ -67,6 +79,29 @@ public class Line extends BoundingBox {
         height = computeHeight();
         computeWords(wordSpacing * WORD_DISTANCE_MULTIPLIER);
         buildBBox();
+        computeFontStyle();
+    }
+
+
+    private void computeFontStyle() {
+
+        EnumMap<FontStyle, AtomicInteger> fontStyleCounter = new EnumMap<>(FontStyle.class);
+        for (FontStyle fontStyle : FontStyle.values()) {
+            fontStyleCounter.put(fontStyle, new AtomicInteger(0));
+        }
+        for (TextPositionSequence word : words) {
+            switch (word.getFontStyle()) {
+                case STANDARD -> fontStyleCounter.get(FontStyle.REGULAR).getAndIncrement();
+                case BOLD -> fontStyleCounter.get(FontStyle.BOLD).getAndIncrement();
+                case ITALIC -> fontStyleCounter.get(FontStyle.ITALIC).getAndIncrement();
+                case BOLD_ITALIC -> fontStyleCounter.get(FontStyle.BOLD_ITALIC).getAndIncrement();
+            }
+        }
+        fontStyle = fontStyleCounter.entrySet()
+                .stream()
+                .max(Comparator.comparing(entry -> entry.getValue().get()))
+                .map(Map.Entry::getKey)
+                .orElse(FontStyle.REGULAR);
     }
 
 
@@ -144,8 +179,8 @@ public class Line extends BoundingBox {
     private void buildBBox() {
 
         this.setToBBoxOfComponents(characters.stream()
-                                       .map(Character::getTextPosition)
-                                       .toList());
+                                           .map(Character::getTextPosition)
+                                           .toList());
     }
 
 

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

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

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

@@ -6,9 +6,11 @@ import java.util.List;
 import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
 
 import lombok.Data;
+import lombok.EqualsAndHashCode;
 
 @Data
-public class Zone extends BoundingBox {
+@EqualsAndHashCode(callSuper = false)
+public class Zone extends TextBoundingBox {
 
     private List<Line> lines;
 

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

@@ -1,15 +1,17 @@
 package com.knecon.fforesight.service.layoutparser.processor.docstrum.service;
 
+import java.awt.geom.Rectangle2D;
 import java.util.ArrayList;
 import java.util.Comparator;
 import java.util.HashMap;
 import java.util.List;
 import java.util.ListIterator;
 import java.util.Map;
+import java.util.stream.Collectors;
 
 import org.springframework.stereotype.Service;
 
-import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.BoundingBox;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.TextBoundingBox;
 import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Zone;
 import com.knecon.fforesight.service.layoutparser.processor.docstrum.utils.DoubleUtils;
 
@@ -19,21 +21,30 @@ public class ReadingOrderService {
     private static final double THRESHOLD = 5;
     public static final double MULTI_COLUMN_DETECTION_THRESHOLD = 1.5;
 
+    private static final Comparator<TextBoundingBox> COMPARATOR = //
+            Comparator.comparing(TextBoundingBox::getY, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD))
+                    .thenComparing(TextBoundingBox::getX, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD));
 
-    public List<Zone> resolve(List<Zone> zones, boolean xyReadingOrder) {
+    private static final Comparator<TextBoundingBox> COMPARATOR_DIR_ADJ = //
+            Comparator.comparing(TextBoundingBox::getYDirAdj, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD))
+                    .thenComparing(TextBoundingBox::getXDirAdj, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD));
+
+
+    public List<Zone> resolve(List<Zone> zones, boolean xyReadingOrder, boolean useDirAdjCoords) {
 
         if (zones.isEmpty() || zones.size() == 1) {
             return zones;
         }
 
         if (xyReadingOrder) {
-            return resolveSingleColumnReadingOrder(zones);
+            return resolveSingleColumnReadingOrder(zones, useDirAdjCoords);
         }
 
         Map<Long, Integer> histogram = new HashMap<>();
         for (Zone zone : zones) {
-            long minY = Math.round(zone.getBBox().getMinY());
-            long maxY = Math.round(zone.getBBox().getMaxY());
+            Rectangle2D bbox = useDirAdjCoords ? zone.getBBoxDirAdj() : zone.getBBox();
+            long minY = Math.round(bbox.getMinY());
+            long maxY = Math.round(bbox.getMaxY());
             for (long i = minY; i <= maxY; i++) {
                 histogram.put(i, histogram.getOrDefault(i, 0) + 1);
             }
@@ -43,24 +54,32 @@ public class ReadingOrderService {
                     .stream()
                     .mapToInt(Integer::intValue).average()
                     .orElse(1) < MULTI_COLUMN_DETECTION_THRESHOLD) {
-            return resolveSingleColumnReadingOrder(zones);
+            return resolveSingleColumnReadingOrder(zones, useDirAdjCoords);
         } else {
 
-            return resolveMultiColumnReadingOder(zones);
+            return resolveMultiColumnReadingOder(zones, useDirAdjCoords);
         }
 
     }
 
 
-    private static List<Zone> resolveSingleColumnReadingOrder(List<Zone> zones) {
+    private static List<Zone> resolveSingleColumnReadingOrder(List<Zone> zones, boolean useDirAdjCoords) {
 
-        zones.sort(Comparator.comparing(BoundingBox::getY, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD))
-                           .thenComparing(BoundingBox::getX, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD)));
+        if (useDirAdjCoords) {
+            return zones.stream()
+                    .collect(Collectors.groupingBy(TextBoundingBox::getDir)).values()
+                    .stream()
+                    .flatMap(words -> words.stream()
+                            .sorted(COMPARATOR_DIR_ADJ))
+                    .toList();
+        }
+
+        zones.sort(COMPARATOR);
         return zones;
     }
 
 
-    private List<Zone> resolveMultiColumnReadingOder(List<Zone> zones) {
+    private List<Zone> resolveMultiColumnReadingOder(List<Zone> zones, boolean useDirAdjCoords) {
 
         // Simple reading order resolver for multi column page layout as described here : https://pub.towardsai.net/advanced-rag-02-unveiling-pdf-parsing-b84ae866344e
         // TODO implement a more fancy reading order resolver see https://github.com/BobLd/DocumentLayoutAnalysis/blob/master/README.md#reading-order
@@ -69,11 +88,12 @@ public class ReadingOrderService {
         double maxX = Double.NEGATIVE_INFINITY;
 
         for (Zone zone : zones) {
-            if (zone.getX() < minX) {
-                minX = zone.getX();
+            Rectangle2D bbox = useDirAdjCoords ? zone.getBBoxDirAdj() : zone.getBBox();
+            if (bbox.getX() < minX) {
+                minX = zone.getXDirAdj();
             }
-            if (zone.getX() + zone.getWidth() > maxX) {
-                maxX = zone.getX() + zone.getWidth();
+            if (bbox.getMaxX() > maxX) {
+                maxX = zone.getMaxXDirAdj();
             }
         }
 
@@ -82,24 +102,27 @@ public class ReadingOrderService {
         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) {
+            Rectangle2D bbox = useDirAdjCoords ? zone.getBBoxDirAdj() : zone.getBBox();
+            if (bbox.getX() < midLineXCoordinate && bbox.getX() + bbox.getWidth() < midLineXCoordinate) {
                 leftOf.add(zone);
-            } else if (zone.getX() > midLineXCoordinate && zone.getX() + zone.getWidth() > midLineXCoordinate) {
+            } else if (bbox.getX() > midLineXCoordinate && bbox.getX() + bbox.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)));
+        if (useDirAdjCoords) {
+            leftOf.sort(COMPARATOR_DIR_ADJ);
+            rightOf.sort(COMPARATOR_DIR_ADJ);
+            middle.sort(COMPARATOR_DIR_ADJ);
+        } else {
+            leftOf.sort(COMPARATOR);
+            rightOf.sort(COMPARATOR);
+            middle.sort(COMPARATOR);
+        }
 /*
         List<Zone> leftNotIntersecting = new ArrayList<>();
         for (Zone leftZone : leftOf) {
@@ -151,8 +174,9 @@ public class ReadingOrderService {
 
         while (itty.hasNext()) {
             Zone current = itty.next();
+            Rectangle2D bbox = useDirAdjCoords ? current.getBBoxDirAdj() : current.getBBox();
             for (int i = 0; i < sortedZones.size(); i++) {
-                if (current.getY() < sortedZones.get(i).getY()) {
+                if (bbox.getY() < sortedZones.get(i).getY()) {
                     sortedZones.add(i, current);
                     itty.remove();
                     break;

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

@@ -1,5 +1,7 @@
 package com.knecon.fforesight.service.layoutparser.processor.docstrum.service;
 
+import static com.knecon.fforesight.service.layoutparser.processor.model.SectionIdentifier.numericalIdentifierPattern;
+
 import java.util.ArrayList;
 import java.util.Comparator;
 import java.util.HashSet;
@@ -21,7 +23,7 @@ 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 MIN_HORIZONTAL_MERGE_DISTANCE_MULTIPLIER = -7;
 
     private static final double MAX_VERTICAL_MERGE_DISTANCE_MULTIPLIER = 0.5;
 
@@ -38,7 +40,7 @@ public class ZoneBuilderService {
 
         double minHorizontalDistance = characterSpacing * MIN_HORIZONTAL_DISTANCE_MULTIPLIER;
         double maxVerticalDistance = lineSpacing * MAX_VERTICAL_DISTANCE_MULTIPLIER;
-        double minHorizontalMergeDistance = characterSpacing * MIN_HORIZONTAL_MERGE_DISTANCE_MULTIPLIER;
+        double minHorizontalMergeDistance = lineSpacing * MIN_HORIZONTAL_MERGE_DISTANCE_MULTIPLIER;
         double maxVerticalMergeDistance = lineSpacing * MAX_VERTICAL_MERGE_DISTANCE_MULTIPLIER;
 
         UnionFind<Line> unionFind = new UnionFind<>(new HashSet<>(lines));
@@ -54,11 +56,26 @@ public class ZoneBuilderService {
                     return;
                 }
 
-                double scale = Math.min(outerLine.getHeight(), innerLine.getHeight()) / meanHeight;
-                scale = Math.max(MIN_LINE_SIZE_SCALE, Math.min(scale, MAX_LINE_SIZE_SCALE));
+//                if (!innerLine.getFontStyle().equals(outerLine.getFontStyle()) //
+//                    && !outerLine.intersectsY(innerLine, -2f)) {
+//                    return;
+//                }
 
-                double horizontalDistance = outerLine.horizontalDistance(innerLine) / scale;
-                double verticalDistance = outerLine.verticalDistance(innerLine) / scale;
+                double horizontalScale = Math.min(outerLine.getHeightDirAdj(), innerLine.getHeightDirAdj()) / meanHeight;
+                horizontalScale = Math.max(MIN_LINE_SIZE_SCALE, Math.min(horizontalScale, MAX_LINE_SIZE_SCALE));
+                double verticalScale = horizontalScale;
+
+//                if (innerLine.toString().endsWith(":")
+//                    || outerLine.toString().endsWith(":")
+//                    || numericalIdentifierPattern.matcher(innerLine.toString()).matches()
+//                    || numericalIdentifierPattern.matcher(outerLine.toString()).matches()) {
+//
+//                    horizontalScale *= 5;
+//                    verticalScale /= 10;
+//                }
+
+                double horizontalDistance = outerLine.horizontalDistance(innerLine) / horizontalScale;
+                double verticalDistance = outerLine.verticalDistance(innerLine) / verticalScale;
 
                 if ((!(minHorizontalDistance <= horizontalDistance) || !(verticalDistance <= maxVerticalDistance)) //
                     && (!(minHorizontalMergeDistance <= horizontalDistance) || !(verticalDistance <= maxVerticalMergeDistance))) {
@@ -87,7 +104,7 @@ public class ZoneBuilderService {
         double weights = 0.0;
         for (Line line : lines) {
             double weight = line.getLength();
-            meanHeight += line.getHeight() * weight;
+            meanHeight += line.getHeightDirAdj() * weight;
             weights += weight;
         }
         meanHeight /= weights;

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

@@ -7,9 +7,12 @@ public class DoubleUtils {
         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);
+
+        if (Math.abs(d1 - d2) < precision) {
+            return 0;
+        }
+
+        return Double.compare(d1, d2);
     }
 
 }

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

@@ -0,0 +1,326 @@
+package com.knecon.fforesight.service.layoutparser.processor.markdown;
+
+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 java.util.TreeMap;
+
+import org.commonmark.Extension;
+import org.commonmark.ext.gfm.tables.TableBlock;
+import org.commonmark.ext.gfm.tables.TableBody;
+import org.commonmark.ext.gfm.tables.TableCell;
+import org.commonmark.ext.gfm.tables.TableHead;
+import org.commonmark.ext.gfm.tables.TableRow;
+import org.commonmark.ext.gfm.tables.TablesExtension;
+import org.commonmark.node.Block;
+import org.commonmark.node.Document;
+import org.commonmark.node.Emphasis;
+import org.commonmark.node.HardLineBreak;
+import org.commonmark.node.Heading;
+import org.commonmark.node.Node;
+import org.commonmark.node.Paragraph;
+import org.commonmark.node.StrongEmphasis;
+import org.commonmark.node.Text;
+import org.commonmark.renderer.markdown.MarkdownRenderer;
+
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.AbstractNodeVisitor;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.TextRange;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Headline;
+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.textblock.TextBlock;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.FontStyle;
+
+public class MarkdownMapper extends AbstractNodeVisitor {
+
+    Document markdownDocument = new Document();
+
+
+    public String toMarkdownContent(com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document document) {
+
+        visit(document);
+
+        return buildRenderer().render(this.markdownDocument);
+    }
+
+
+    @Override
+    public void visit(Headline headline) {
+
+        markdownDocument.appendChild(parseHeadline(headline));
+    }
+
+
+    @Override
+    public void visit(com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Paragraph paragraph) {
+
+        markdownDocument.appendChild(parseParagraph(paragraph));
+    }
+
+
+    @Override
+    public void visit(Table table) {
+
+        markdownDocument.appendChild(parseTable(table));
+    }
+
+
+    private static MarkdownRenderer buildRenderer() {
+
+        List<Extension> extensions = List.of(TablesExtension.create());
+        return MarkdownRenderer.builder().extensions(extensions).build();
+    }
+
+
+    private Block parseTable(Table table) {
+
+//        if (table.getNumberOfRows() == 1 && table.getNumberOfCols() == 1) {
+//            org.commonmark.node.Paragraph markdownParagraph = new org.commonmark.node.Paragraph();
+//            parseTextBlock(table.getTextBlock(), true).forEach(markdownParagraph::appendChild);
+//            return markdownParagraph;
+//        }
+
+        TableBlock tableNode = new TableBlock();
+        TableHead head = new TableHead();
+        TableRow tableRow = createTableRow(table, 0);
+        head.appendChild(tableRow);
+        int row = 1;
+        tableNode.appendChild(head);
+        TableBody tableBody = new TableBody();
+        for (; row < table.getNumberOfRows(); row++) {
+            tableBody.appendChild(createTableRow(table, row));
+        }
+        tableNode.appendChild(tableBody);
+        return tableNode;
+    }
+
+
+    private TableRow createTableRow(Table table, int row) {
+
+        TableRow tableRow = new TableRow();
+        table.streamRow(row)
+                .map(this::createTableCell)
+                .forEach(tableRow::appendChild);
+        return tableRow;
+    }
+
+
+    private Node createTableCell(com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.TableCell tc) {
+
+        var cell = new TableCell();
+        List<SemanticNode> childNodes = tc.streamChildren()
+                .toList();
+        if (childNodes.isEmpty()) {
+            parseTextBlock(tc.getTextBlock(), false).forEach(cell::appendChild);
+        } else {
+            childNodes.forEach(semanticNode -> parseTextBlock(semanticNode.getTextBlock(), false).forEach(cell::appendChild));
+        }
+        return cell;
+    }
+
+
+    private Paragraph parseParagraph(com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Paragraph paragraph) {
+
+        org.commonmark.node.Paragraph markdownParagraph = new org.commonmark.node.Paragraph();
+        parseTextBlock(paragraph.getTextBlock(), true).forEach(markdownParagraph::appendChild);
+        return markdownParagraph;
+    }
+
+
+    private Heading parseHeadline(Headline headline) {
+
+        Heading heading = new Heading();
+        heading.setLevel(headline.getTreeId().size());
+        heading.appendChild(parseTextBlockAsText(headline.getTextBlock()));
+        return heading;
+
+    }
+
+
+    private Text parseTextBlockAsText(TextBlock textBlock) {
+
+        return new Text(textBlock.getSearchText());
+    }
+
+
+    private List<Node> parseTextBlock(TextBlock textBlock, boolean includeLineBreaks) {
+
+        LinkedList<Node> result = new LinkedList<>();
+        List<TextRangeWithTextType> textRanges = mergeTextStyles(textBlock);
+
+        String fullText = getText(textBlock, textBlock.getTextRange(), includeLineBreaks);
+        List<Integer> lineTextSizes = getLineTextSizes(fullText);
+        int idx = 0;
+        int charCount = 0;
+        for (TextRangeWithTextType textRange : textRanges) {
+            String text = getText(textBlock, textRange.textRange(), includeLineBreaks);
+            String[] lines = text.split("\n");
+            for (String line : lines) {
+                charCount += line.length();
+                switch (textRange.fontStyle()) {
+                    case REGULAR -> result.add(new Text(line));
+                    case BOLD -> {
+                        StrongEmphasis boldBlock = new StrongEmphasis();
+                        boldBlock.appendChild(new Text(line));
+                        result.add(boldBlock);
+                    }
+                    case ITALIC -> {
+                        Emphasis italicBlock = new Emphasis();
+                        italicBlock.appendChild(new Text(line));
+                        result.add(italicBlock);
+                    }
+                    case BOLD_ITALIC -> {
+                        Emphasis italicBlock = new Emphasis();
+
+                        StrongEmphasis boldBlock = new StrongEmphasis();
+                        boldBlock.appendChild(new Text(line));
+
+                        italicBlock.appendChild(boldBlock);
+                        result.add(italicBlock);
+                    }
+                }
+                if (includeLineBreaks && lineTextSizes.get(idx).equals(charCount)) {
+                    result.add(new HardLineBreak());
+                    idx++;
+                }
+
+            }
+        }
+        if (!result.isEmpty() && result.getLast() instanceof HardLineBreak) {
+            result.removeLast();
+        }
+        return result;
+    }
+
+
+    private static List<Integer> getLineTextSizes(String input) {
+
+        String[] parts = input.split("\n");
+        List<Integer> textSizes = new ArrayList<>();
+
+        int size = 0;
+        for (int i = 0; i < parts.length; i++) {
+            size += parts[i].length();
+            textSizes.add(size);
+        }
+
+        if (textSizes.isEmpty()) {
+            textSizes.add(0);
+        }
+
+        return textSizes;
+    }
+
+
+    private static String getText(TextBlock textBlock, TextRange textRange, boolean includeLineBreaks) {
+
+        return includeLineBreaks ? textBlock.subSequenceWithLineBreaks(textRange) : textBlock.subSequence(textRange).toString();
+    }
+
+
+    private List<TextRangeWithTextType> mergeTextStyles(TextBlock textBlock) {
+
+        List<TextRangeWithTextType> result = new ArrayList<>();
+
+        TreeMap<Integer, Set<FontStyleChange>> styleChanges = new TreeMap<>();
+
+        int start = textBlock.getTextRange().start();
+        int end = textBlock.getTextRange().end();
+
+        for (TextRange bold : textBlock.getBoldTextBoundaries()) {
+            styleChanges.computeIfAbsent(bold.start() + start, k -> new HashSet<>()).add(FontStyleChange.enter(FontStyle.BOLD));
+            styleChanges.computeIfAbsent(bold.end() + start, k -> new HashSet<>()).add(FontStyleChange.leave(FontStyle.BOLD));
+        }
+
+        for (TextRange italic : textBlock.getItalicTextBoundaries()) {
+            styleChanges.computeIfAbsent(italic.start() + start, k -> new HashSet<>()).add(FontStyleChange.enter(FontStyle.ITALIC));
+            styleChanges.computeIfAbsent(italic.end() + start, k -> new HashSet<>()).add(FontStyleChange.leave(FontStyle.ITALIC));
+        }
+
+        if (styleChanges.isEmpty()) {
+            result.add(new TextRangeWithTextType(new TextRange(start, end), FontStyle.REGULAR));
+            return result;
+        }
+
+        Set<FontStyle> currentStyles = new HashSet<>();
+        currentStyles.add(FontStyle.REGULAR);
+
+        for (Map.Entry<Integer, Set<FontStyleChange>> entry : styleChanges.entrySet()) {
+            int point = entry.getKey();
+            Set<FontStyleChange> changes = entry.getValue();
+
+            if (point > start) {
+                FontStyle style = determineFontStyle(currentStyles);
+                result.add(new TextRangeWithTextType(new TextRange(start, point), style));
+            }
+
+            changes.stream()
+                    .filter(FontStyleChange::leave)
+                    .map(FontStyleChange::style)
+                    .toList()
+                    .forEach(currentStyles::remove);
+
+            currentStyles.addAll(changes.stream()
+                                         .filter(FontStyleChange::enter)
+                                         .map(FontStyleChange::style)
+                                         .toList());
+
+            if (currentStyles.isEmpty()) {
+                currentStyles.add(FontStyle.REGULAR);
+            }
+
+            start = point;
+        }
+
+        if (start < end) {
+            FontStyle style = determineFontStyle(currentStyles);
+            result.add(new TextRangeWithTextType(new TextRange(start, textBlock.getTextRange().end()), style));
+        }
+
+        return result;
+    }
+
+
+    private FontStyle determineFontStyle(Set<FontStyle> styles) {
+
+        if (styles.contains(FontStyle.BOLD) && styles.contains(FontStyle.ITALIC)) {
+            return FontStyle.BOLD_ITALIC;
+        } else if (styles.contains(FontStyle.BOLD)) {
+            return FontStyle.BOLD;
+        } else if (styles.contains(FontStyle.ITALIC)) {
+            return FontStyle.ITALIC;
+        } else {
+            return FontStyle.REGULAR;
+        }
+    }
+
+
+
+    record FontStyleChange(boolean enter, FontStyle style) {
+
+        public static FontStyleChange enter(FontStyle style) {
+
+            return new FontStyleChange(true, style);
+        }
+
+
+        public static FontStyleChange leave(FontStyle style) {
+
+            return new FontStyleChange(false, style);
+        }
+
+
+        public boolean leave() {
+
+            return !enter;
+        }
+
+    }
+
+    record TextRangeWithTextType(TextRange textRange, FontStyle fontStyle) {
+
+    }
+}

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

@@ -1,9 +1,10 @@
 package com.knecon.fforesight.service.layoutparser.processor.model;
 
-import java.awt.geom.Rectangle2D;
+import java.util.HashSet;
+import java.util.Set;
 
 import com.fasterxml.jackson.annotation.JsonIgnore;
-import com.iqser.red.service.persistence.service.v1.api.shared.model.redactionlog.Rectangle;
+import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
 import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.BoundingBox;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
 
@@ -20,6 +21,9 @@ public abstract class AbstractPageBlock extends BoundingBox {
 
     @JsonIgnore
     protected PageBlockType classification;
+
+    Set<LayoutEngine> engines = new HashSet<>();
+
     @JsonIgnore
     protected int page;
 

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

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

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

@@ -8,21 +8,26 @@ import java.util.Map;
 
 import com.iqser.red.service.persistence.service.v1.api.shared.model.redactionlog.Rectangle;
 import com.knecon.fforesight.service.layoutparser.processor.model.image.ClassifiedImage;
+import com.knecon.fforesight.service.layoutparser.processor.model.outline.OutlineObject;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRulings;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.StringFrequencyCounter;
 
 import lombok.Data;
 import lombok.NonNull;
 import lombok.RequiredArgsConstructor;
-import org.apache.pdfbox.pdmodel.documentinterchange.markedcontent.PDMarkedContent;
 
 @Data
 @RequiredArgsConstructor
+
 public class ClassificationPage {
 
     @NonNull
     private List<AbstractPageBlock> textBlocks;
 
+    private List<OutlineObject> outlineObjects = new ArrayList<>();
+
+    private List<AbstractPageBlock> headlines = new ArrayList<>();
+
     private List<ClassifiedImage> images = new ArrayList<>();
 
     private Rectangle bodyTextFrame;
@@ -40,7 +45,7 @@ public class ClassificationPage {
     private float pageWidth;
     private float pageHeight;
 
-    CleanRulings cleanRulings;
+    private CleanRulings cleanRulings;
 
     private Map<String, List<Rectangle2D>> markedContentBboxPerType = new HashMap<>();
 

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

@@ -12,6 +12,7 @@ import lombok.NoArgsConstructor;
 
 @Data
 @NoArgsConstructor
+@Deprecated
 public class ClassificationSection {
 
     private List<AbstractPageBlock> pageBlocks = new ArrayList<>();

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

@@ -12,10 +12,10 @@ import lombok.Getter;
 @Getter
 public class FloatFrequencyCounter {
 
-    Map<Float, Integer> countPerValue = new HashMap<>();
+    Map<Double, Integer> countPerValue = new HashMap<>();
 
 
-    public void add(float value) {
+    public void add(double value) {
 
         if (!countPerValue.containsKey(value)) {
             countPerValue.put(value, 1);
@@ -25,9 +25,9 @@ public class FloatFrequencyCounter {
     }
 
 
-    public void addAll(Map<Float, Integer> otherCounter) {
+    public void addAll(Map<Double, Integer> otherCounter) {
 
-        for (Map.Entry<Float, Integer> entry : otherCounter.entrySet()) {
+        for (Map.Entry<Double, Integer> entry : otherCounter.entrySet()) {
             if (countPerValue.containsKey(entry.getKey())) {
                 countPerValue.put(entry.getKey(), countPerValue.get(entry.getKey()) + entry.getValue());
             } else {
@@ -37,10 +37,10 @@ public class FloatFrequencyCounter {
     }
 
 
-    public Float getMostPopular() {
+    public Double getMostPopular() {
 
-        Map.Entry<Float, Integer> mostPopular = null;
-        for (Map.Entry<Float, Integer> entry : countPerValue.entrySet()) {
+        Map.Entry<Double, Integer> mostPopular = null;
+        for (Map.Entry<Double, Integer> entry : countPerValue.entrySet()) {
             if (mostPopular == null || entry.getValue() >= mostPopular.getValue()) {
                 mostPopular = entry;
             }
@@ -49,11 +49,11 @@ public class FloatFrequencyCounter {
     }
 
 
-    public List<Float> getHighterThanMostPopular() {
+    public List<Double> getHigherThanMostPopular() {
 
-        Float mostPopular = getMostPopular();
-        List<Float> higher = new ArrayList<>();
-        for (Float value : countPerValue.keySet()) {
+        Double mostPopular = getMostPopular();
+        List<Double> higher = new ArrayList<>();
+        for (Double value : countPerValue.keySet()) {
             if (value > mostPopular) {
                 higher.add(value);
             }
@@ -63,10 +63,10 @@ public class FloatFrequencyCounter {
     }
 
 
-    public Float getHighest() {
+    public Double getHighest() {
 
-        Float highest = null;
-        for (Float value : countPerValue.keySet()) {
+        Double highest = null;
+        for (Double value : countPerValue.keySet()) {
             if (highest == null || value > highest) {
                 highest = value;
             }

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

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

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

@@ -8,21 +8,24 @@ import java.util.regex.Pattern;
 
 import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
+import lombok.Getter;
 import lombok.experimental.FieldDefaults;
 
 @AllArgsConstructor
 @FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
 public class SectionIdentifier {
 
-    static Pattern numericalIdentifierPattern = Pattern.compile("^[\\s]?(\\d+)[\\s.,;]?(\\d+)?[\\s.,;]?(\\d+)?[\\s.,;]?(\\d+)?");
+    public static Pattern numericalIdentifierPattern = Pattern.compile("^[\\s]?(\\d+)[\\s.,;]?(\\d+)?[\\s.,;]?(\\d+)?[\\s.,;]?(\\d+)?");
 
-    private enum Format {
+    public enum Format {
         EMPTY,
         NUMERICAL,
         DOCUMENT
     }
 
+    @Getter
     Format format;
+    @Getter
     String identifierString;
     List<Integer> identifiers;
     boolean asChild;

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

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

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

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

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

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

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

@@ -13,13 +13,13 @@ import lombok.Setter;
 @Setter
 @EqualsAndHashCode
 @SuppressWarnings("PMD.AvoidFieldNameMatchingMethodName")
-public class Boundary implements Comparable<Boundary> {
+public class TextRange implements Comparable<TextRange> {
 
     private int start;
     private int end;
 
 
-    public Boundary(int start, int end) {
+    public TextRange(int start, int end) {
 
         if (start > end) {
             throw new IllegalArgumentException(format("start: %d > end: %d", start, end));
@@ -47,15 +47,15 @@ public class Boundary implements Comparable<Boundary> {
     }
 
 
-    public boolean contains(Boundary boundary) {
+    public boolean contains(TextRange textRange) {
 
-        return start <= boundary.start() && boundary.end() <= end;
+        return start <= textRange.start() && textRange.end() <= end;
     }
 
 
-    public boolean containedBy(Boundary boundary) {
+    public boolean containedBy(TextRange textRange) {
 
-        return boundary.contains(this);
+        return textRange.contains(this);
     }
 
 
@@ -79,22 +79,33 @@ public class Boundary implements Comparable<Boundary> {
 
     public boolean contains(int index) {
 
+        return start <= index && index <= end;
+    }
+
+
+    public boolean containsExclusive(int index) {
+
         return start <= index && index < end;
     }
 
 
-    public boolean intersects(Boundary boundary) {
+    public boolean intersects(TextRange textRange) {
 
-        return boundary.start() < this.end && this.start < boundary.end();
+        return textRange.start() < this.end && this.start < textRange.end();
     }
 
 
-    public List<Boundary> split(List<Integer> splitIndices) {
+    public List<TextRange> split(List<Integer> splitIndices) {
 
-        if (splitIndices.stream().anyMatch(idx -> !this.contains(idx))) {
-            throw new IndexOutOfBoundsException(format("%s splitting indices are out of range for %s", splitIndices.stream().filter(idx -> !this.contains(idx)).toList(), this));
+        if (splitIndices.stream()
+                .anyMatch(idx -> !this.containsExclusive(idx))) {
+            throw new IndexOutOfBoundsException(format("%s splitting indices are out of range for %s",
+                                                       splitIndices.stream()
+                                                               .filter(idx -> !this.containsExclusive(idx))
+                                                               .toList(),
+                                                       this));
         }
-        List<Boundary> splitBoundaries = new LinkedList<>();
+        List<TextRange> splitBoundaries = new LinkedList<>();
         int previousIndex = start;
         for (int splitIndex : splitIndices) {
 
@@ -102,23 +113,31 @@ public class Boundary implements Comparable<Boundary> {
             if (splitIndex == previousIndex) {
                 continue;
             }
-            splitBoundaries.add(new Boundary(previousIndex, splitIndex));
+            splitBoundaries.add(new TextRange(previousIndex, splitIndex));
             previousIndex = splitIndex;
         }
-        splitBoundaries.add(new Boundary(previousIndex, end));
+        splitBoundaries.add(new TextRange(previousIndex, end));
         return splitBoundaries;
     }
 
+
     public IntStream intStream() {
 
         return IntStream.range(start, end);
     }
 
-    public static Boundary merge(Collection<Boundary> boundaries) {
 
-        int minStart = boundaries.stream().mapToInt(Boundary::start).min().orElseThrow(IllegalArgumentException::new);
-        int maxEnd = boundaries.stream().mapToInt(Boundary::end).max().orElseThrow(IllegalArgumentException::new);
-        return new Boundary(minStart, maxEnd);
+    public static TextRange merge(Collection<TextRange> boundaries) {
+
+        int minStart = boundaries.stream()
+                .mapToInt(TextRange::start)
+                .min()
+                .orElseThrow(IllegalArgumentException::new);
+        int maxEnd = boundaries.stream()
+                .mapToInt(TextRange::end)
+                .max()
+                .orElseThrow(IllegalArgumentException::new);
+        return new TextRange(minStart, maxEnd);
     }
 
 
@@ -130,12 +149,12 @@ public class Boundary implements Comparable<Boundary> {
 
 
     @Override
-    public int compareTo(Boundary boundary) {
+    public int compareTo(TextRange textRange) {
 
-        if (end < boundary.end() && start < boundary.start()) {
+        if (end < textRange.end() && start < textRange.start()) {
             return -1;
         }
-        if (start > boundary.start() && end > boundary.end()) {
+        if (start > textRange.start() && end > textRange.end()) {
             return 1;
         }
 

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

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

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

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

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

@@ -3,45 +3,38 @@ package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
 import java.awt.geom.Rectangle2D;
 import java.util.Collections;
 import java.util.HashMap;
-import java.util.HashSet;
 import java.util.List;
 import java.util.Map;
 import java.util.Set;
 import java.util.stream.Collectors;
 import java.util.stream.Stream;
 
-import com.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.visualization.LayoutparsingVisualizations;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
+import com.knecon.fforesight.service.layoutparser.processor.visualization.LayoutDebugLayer;
 
 import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
-import lombok.Builder;
 import lombok.Data;
+import lombok.EqualsAndHashCode;
 import lombok.NoArgsConstructor;
 import lombok.experimental.FieldDefaults;
+import lombok.experimental.SuperBuilder;
 
 @Data
-@Builder
+@SuperBuilder
 @AllArgsConstructor
 @NoArgsConstructor
+@EqualsAndHashCode(callSuper = true)
 @FieldDefaults(level = AccessLevel.PRIVATE)
-public class Document implements GenericSemanticNode {
-
-    @Builder.Default
-    Set<LayoutEngine> engines = new HashSet<>(Set.of(LayoutEngine.ALGORITHM));
+public class Document extends AbstractSemanticNode {
 
     Set<Page> pages;
-    DocumentTree documentTree;
     Integer numberOfPages;
-    TextBlock textBlock;
-    @Builder.Default
-    Set<RedactionEntity> entities = new HashSet<>();
 
-    LayoutparsingVisualizations visualizations;
+    LayoutDebugLayer layoutDebugLayer;
 
 
     @Override
@@ -51,15 +44,27 @@ public class Document implements GenericSemanticNode {
     }
 
 
-    public TextBlock getTextBlock() {
+    /**
+     * Gets the sections of the document as a list.
+     *
+     * @return A list of all sections within the document.
+     */
+    public List<Section> getAllSections() {
 
-        if (textBlock == null) {
-            textBlock = GenericSemanticNode.super.getTextBlock();
-        }
-        return textBlock;
+        return streamAllSubNodesOfType(NodeType.SECTION).map(node -> (Section) node)
+                .collect(Collectors.toList());
     }
 
 
+    /**
+     * Gets the main sections of the document as a list.
+     *
+     * @return A list of main sections within the document
+     * @deprecated This method is marked for removal.
+     * Use {@link #streamChildrenOfType(NodeType)} instead,
+     * or {@link #getChildrenOfTypeSectionOrSuperSection()} which returns children of type SECTION as well as SUPER_SECTION.
+     */
+    @Deprecated(forRemoval = true)
     public List<Section> getMainSections() {
 
         return streamChildrenOfType(NodeType.SECTION).map(node -> (Section) node)
@@ -67,6 +72,18 @@ public class Document implements GenericSemanticNode {
     }
 
 
+    /**
+     * Gets the direct children of type SECTION or SUPER_SECTION of the document as a list of SemanticNode objects.
+     *
+     * @return A list of all children of type SECTION or SUPER_SECTION.
+     */
+    public List<SemanticNode> getChildrenOfTypeSectionOrSuperSection() {
+
+        return streamChildren().filter(semanticNode -> semanticNode.getType().equals(NodeType.SECTION) || semanticNode.getType().equals(NodeType.SUPER_SECTION))
+                .toList();
+    }
+
+
     public List<Header> getHeaders() {
 
         return streamChildrenOfType(NodeType.HEADER).map(node -> (Header) node)
@@ -81,6 +98,14 @@ public class Document implements GenericSemanticNode {
     }
 
 
+    @Override
+    public Headline getHeadline() {
+
+        return streamAllSubNodesOfType(NodeType.HEADLINE).map(node -> (Headline) node)
+                .findFirst().orElse(Headline.builder().build());
+    }
+
+
     public Stream<TextBlock> streamTerminalTextBlocksInOrder() {
 
         return streamAllNodes().filter(SemanticNode::isLeaf)
@@ -102,18 +127,9 @@ public class Document implements GenericSemanticNode {
     }
 
 
-    @Override
-    public Headline getHeadline() {
-
-        return streamAllSubNodesOfType(NodeType.HEADLINE).map(node -> (Headline) node)
-                .findFirst()
-                .orElse(Headline.builder().build());
-    }
-
-
     private Stream<SemanticNode> streamAllNodes() {
 
-        return documentTree.allEntriesInOrder()
+        return getDocumentTree().allEntriesInOrder()
                 .map(DocumentTree.Entry::getNode);
     }
 
@@ -147,4 +163,11 @@ public class Document implements GenericSemanticNode {
         return bBox;
     }
 
+
+    @Override
+    public void accept(NodeVisitor visitor) {
+
+        visitor.visit(this);
+    }
+
 }

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

@@ -20,7 +20,8 @@ public class DuplicatedParagraph extends Paragraph {
     @Override
     public TextBlock getTextBlock() {
 
-        return Stream.of(leafTextBlock, unsortedLeafTextBlock).collect(new TextBlockCollector());
+        return Stream.of(super.getLeafTextBlock(), unsortedLeafTextBlock)
+                .collect(new TextBlockCollector());
 
     }
 

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

@@ -1,48 +1,25 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
 
-import java.awt.geom.Rectangle2D;
-import java.util.HashSet;
-import java.util.List;
-import java.util.Map;
-import java.util.Set;
-
-import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
 
 import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
-import lombok.Builder;
 import lombok.Data;
 import lombok.EqualsAndHashCode;
-import lombok.NoArgsConstructor;
 import lombok.experimental.FieldDefaults;
+import lombok.experimental.SuperBuilder;
 
 @Data
-@Builder
+@SuperBuilder
 @AllArgsConstructor
-@NoArgsConstructor
+@EqualsAndHashCode(callSuper = true)
 @FieldDefaults(level = AccessLevel.PRIVATE)
-public class Footer implements GenericSemanticNode {
+public class Footer extends AbstractSemanticNode {
 
-    @Builder.Default
-    Set<LayoutEngine> engines = new HashSet<>(Set.of(LayoutEngine.ALGORITHM));
-
-    List<Integer> treeId;
     TextBlock leafTextBlock;
 
-    @EqualsAndHashCode.Exclude
-    DocumentTree documentTree;
-
-    @Builder.Default
-    @EqualsAndHashCode.Exclude
-    Set<RedactionEntity> entities = new HashSet<>();
-
-    @EqualsAndHashCode.Exclude
-    Map<Page, Rectangle2D> bBoxCache;
-
 
     @Override
     public NodeType getType() {
@@ -58,6 +35,12 @@ public class Footer implements GenericSemanticNode {
     }
 
 
+    @Override
+    public void accept(NodeVisitor visitor) {
+        visitor.visit(this);
+    }
+
+
     @Override
     public TextBlock getTextBlock() {
 
@@ -68,17 +51,7 @@ public class Footer implements GenericSemanticNode {
     @Override
     public String toString() {
 
-        return treeId + ": " + NodeType.FOOTER + ": " + leafTextBlock.buildSummary();
-    }
-
-
-    @Override
-    public Map<Page, Rectangle2D> getBBox() {
-
-        if (bBoxCache == null) {
-            bBoxCache = GenericSemanticNode.super.getBBox();
-        }
-        return bBoxCache;
+        return getTreeId() + ": " + NodeType.FOOTER + ": " + leafTextBlock.buildSummary();
     }
 
 }

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

@@ -1,47 +1,25 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
 
-import java.awt.geom.Rectangle2D;
-import java.util.HashSet;
-import java.util.List;
-import java.util.Map;
-import java.util.Set;
-
-import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
 
 import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
-import lombok.Builder;
 import lombok.Data;
 import lombok.EqualsAndHashCode;
-import lombok.NoArgsConstructor;
 import lombok.experimental.FieldDefaults;
+import lombok.experimental.SuperBuilder;
 
 @Data
-@Builder
+@SuperBuilder
 @AllArgsConstructor
-@NoArgsConstructor
+@EqualsAndHashCode(callSuper = true)
 @FieldDefaults(level = AccessLevel.PRIVATE)
-public class Header implements GenericSemanticNode {
+public class Header extends AbstractSemanticNode {
 
-    @Builder.Default
-    Set<LayoutEngine> engines = new HashSet<>(Set.of(LayoutEngine.ALGORITHM));
-    List<Integer> treeId;
     TextBlock leafTextBlock;
 
-    @EqualsAndHashCode.Exclude
-    DocumentTree documentTree;
-
-    @Builder.Default
-    @EqualsAndHashCode.Exclude
-    Set<RedactionEntity> entities = new HashSet<>();
-
-    @EqualsAndHashCode.Exclude
-    Map<Page, Rectangle2D> bBoxCache;
-
 
     @Override
     public boolean isLeaf() {
@@ -50,6 +28,13 @@ public class Header implements GenericSemanticNode {
     }
 
 
+    @Override
+    public void accept(NodeVisitor visitor) {
+
+        visitor.visit(this);
+    }
+
+
     @Override
     public NodeType getType() {
 
@@ -67,17 +52,7 @@ public class Header implements GenericSemanticNode {
     @Override
     public String toString() {
 
-        return treeId + ": " + NodeType.HEADER + ": " + leafTextBlock.buildSummary();
-    }
-
-
-    @Override
-    public Map<Page, Rectangle2D> getBBox() {
-
-        if (bBoxCache == null) {
-            bBoxCache = GenericSemanticNode.super.getBBox();
-        }
-        return bBoxCache;
+        return getTreeId() + ": " + NodeType.HEADER + ": " + leafTextBlock.buildSummary();
     }
 
 }

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

@@ -1,47 +1,25 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
 
-import java.awt.geom.Rectangle2D;
-import java.util.HashSet;
-import java.util.List;
-import java.util.Map;
-import java.util.Set;
-
-import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
 
 import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
-import lombok.Builder;
 import lombok.Data;
 import lombok.EqualsAndHashCode;
-import lombok.NoArgsConstructor;
 import lombok.experimental.FieldDefaults;
+import lombok.experimental.SuperBuilder;
 
 @Data
-@Builder
+@SuperBuilder
 @AllArgsConstructor
-@NoArgsConstructor
+@EqualsAndHashCode(callSuper = true)
 @FieldDefaults(level = AccessLevel.PRIVATE)
-public class Headline implements GenericSemanticNode {
+public class Headline extends AbstractSemanticNode {
 
-    @Builder.Default
-    Set<LayoutEngine> engines = new HashSet<>(Set.of(LayoutEngine.ALGORITHM));
-    List<Integer> treeId;
     TextBlock leafTextBlock;
 
-    @EqualsAndHashCode.Exclude
-    DocumentTree documentTree;
-
-    @Builder.Default
-    @EqualsAndHashCode.Exclude
-    Set<RedactionEntity> entities = new HashSet<>();
-
-    @EqualsAndHashCode.Exclude
-    Map<Page, Rectangle2D> bBoxCache;
-
 
     @Override
     public NodeType getType() {
@@ -57,6 +35,13 @@ public class Headline implements GenericSemanticNode {
     }
 
 
+    @Override
+    public void accept(NodeVisitor visitor) {
+
+        visitor.visit(this);
+    }
+
+
     @Override
     public TextBlock getTextBlock() {
 
@@ -67,7 +52,7 @@ public class Headline implements GenericSemanticNode {
     @Override
     public String toString() {
 
-        return treeId + ": " + NodeType.HEADLINE + ": " + leafTextBlock.buildSummary();
+        return getTreeId() + ": " + NodeType.HEADLINE + ": " + leafTextBlock.buildSummary();
     }
 
 
@@ -77,14 +62,4 @@ public class Headline implements GenericSemanticNode {
         return this;
     }
 
-
-    @Override
-    public Map<Page, Rectangle2D> getBBox() {
-
-        if (bBoxCache == null) {
-            bBoxCache = GenericSemanticNode.super.getBBox();
-        }
-        return bBoxCache;
-    }
-
 }

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

@@ -3,15 +3,11 @@ package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
 import java.awt.geom.Rectangle2D;
 import java.util.Collections;
 import java.util.HashMap;
-import java.util.HashSet;
-import java.util.List;
 import java.util.Map;
 import java.util.Set;
 
-import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
 
 import lombok.AccessLevel;
@@ -21,20 +17,20 @@ import lombok.Data;
 import lombok.EqualsAndHashCode;
 import lombok.NoArgsConstructor;
 import lombok.experimental.FieldDefaults;
+import lombok.experimental.SuperBuilder;
 
 @Data
-@Builder
+@SuperBuilder
 @AllArgsConstructor
 @NoArgsConstructor
+@EqualsAndHashCode(callSuper = true)
 @FieldDefaults(level = AccessLevel.PRIVATE)
-public class Image implements GenericSemanticNode {
+public class Image extends AbstractSemanticNode {
 
-    @Builder.Default
-    Set<LayoutEngine> engines = new HashSet<>(Set.of(LayoutEngine.ALGORITHM));
-
-    List<Integer> treeId;
     String id;
 
+    String representationHash;
+
     ImageType imageType;
     boolean transparent;
     Rectangle2D position;
@@ -53,13 +49,6 @@ public class Image implements GenericSemanticNode {
     @EqualsAndHashCode.Exclude
     Page page;
 
-    @EqualsAndHashCode.Exclude
-    DocumentTree documentTree;
-
-    @Builder.Default
-    @EqualsAndHashCode.Exclude
-    Set<RedactionEntity> entities = new HashSet<>();
-
 
     @Override
     public NodeType getType() {
@@ -85,7 +74,7 @@ public class Image implements GenericSemanticNode {
     @Override
     public String toString() {
 
-        return treeId + ": " + NodeType.IMAGE + ": " + imageType.toString() + " " + position;
+        return getTreeId() + ": " + NodeType.IMAGE + ": " + imageType.toString() + " " + position;
     }
 
 
@@ -98,10 +87,29 @@ public class Image implements GenericSemanticNode {
     }
 
 
+    @Override
+    public void accept(NodeVisitor visitor) {
+
+        visitor.visit(this);
+    }
+
+
     @Override
     public boolean isLeaf() {
 
         return true;
     }
 
+
+    public double getArea() {
+
+        return position.getWidth() * position.getHeight();
+    }
+
+
+    public boolean isFullPageImage() {
+
+        return imageType.equals(ImageType.OCR) || getArea() >= 0.5 * page.getArea();
+    }
+
 }

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

@@ -6,7 +6,6 @@ public enum ImageType {
     LOGO,
     FORMULA,
     SIGNATURE,
-
     SIGNATURE_VISUAL,
     OTHER,
     OCR,

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

@@ -1,12 +1,15 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
 
+import java.util.Comparator;
 import java.util.HashSet;
 import java.util.LinkedList;
 import java.util.List;
 import java.util.Set;
+import java.util.stream.Stream;
 
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.TextEntity;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.AtomicTextBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlockCollector;
 
@@ -29,9 +32,8 @@ public class Page {
     Integer height;
     Integer width;
     Integer rotation;
-
     @EqualsAndHashCode.Exclude
-    List<SemanticNode> mainBody;
+    List<AtomicTextBlock> textBlocksOnPage;
     @EqualsAndHashCode.Exclude
     Header header;
     @EqualsAndHashCode.Exclude
@@ -39,7 +41,7 @@ public class Page {
 
     @Builder.Default
     @EqualsAndHashCode.Exclude
-    Set<RedactionEntity> entities = new HashSet<>();
+    Set<TextEntity> entities = new HashSet<>();
 
     @Builder.Default
     @EqualsAndHashCode.Exclude
@@ -53,14 +55,40 @@ public class Page {
                 .width((int) classificationPage.getPageWidth())
                 .number(classificationPage.getPageNumber())
                 .rotation(classificationPage.getRotation())
-                .mainBody(new LinkedList<>())
+                .textBlocksOnPage(new LinkedList<>())
                 .build();
     }
 
 
+    /**
+     * Constructs and returns a {@link TextBlock} representing the concatenated text of all leaf semantic nodes in the main body.
+     *
+     * @return The main body text block.
+     */
     public TextBlock getMainBodyTextBlock() {
 
-        return mainBody.stream().filter(SemanticNode::isLeaf).map(SemanticNode::getLeafTextBlock).collect(new TextBlockCollector());
+        return textBlocksOnPage.stream()
+                .collect(new TextBlockCollector());
+    }
+
+
+    public List<SemanticNode> getMainBody() {
+
+        return textBlocksOnPage.stream()
+                .map(AtomicTextBlock::getParent)
+                .map(this::getHighestParentOnPage)
+                .distinct()
+                .toList();
+    }
+
+
+    private SemanticNode getHighestParentOnPage(SemanticNode node) {
+
+        SemanticNode currentNode = node;
+        while (currentNode.hasParent() && currentNode.getParent().onlyOnPage(this)) {
+            currentNode = currentNode.getParent();
+        }
+        return currentNode;
     }
 
 
@@ -84,4 +112,10 @@ public class Page {
         return o instanceof Page && o.hashCode() == this.hashCode();
     }
 
+
+    public double getArea() {
+
+        return height * width;
+    }
+
 }

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

@@ -1,20 +1,11 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
 
-import java.awt.geom.Rectangle2D;
-import java.util.HashSet;
-import java.util.List;
-import java.util.Map;
-import java.util.Set;
-
-import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
 
 import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
-import lombok.Builder;
 import lombok.Data;
 import lombok.EqualsAndHashCode;
 import lombok.experimental.FieldDefaults;
@@ -23,25 +14,12 @@ import lombok.experimental.SuperBuilder;
 @Data
 @SuperBuilder
 @AllArgsConstructor
+@EqualsAndHashCode(callSuper = true)
 @FieldDefaults(level = AccessLevel.PROTECTED)
-public class Paragraph implements GenericSemanticNode {
+public class Paragraph extends AbstractSemanticNode {
 
-    @Builder.Default
-    Set<LayoutEngine> engines = new HashSet<>(Set.of(LayoutEngine.ALGORITHM));
-
-    List<Integer> treeId;
     TextBlock leafTextBlock;
 
-    @EqualsAndHashCode.Exclude
-    DocumentTree documentTree;
-
-    @Builder.Default
-    @EqualsAndHashCode.Exclude
-    Set<RedactionEntity> entities = new HashSet<>();
-
-    @EqualsAndHashCode.Exclude
-    Map<Page, Rectangle2D> bBoxCache;
-
 
     @Override
     public NodeType getType() {
@@ -57,27 +35,17 @@ public class Paragraph implements GenericSemanticNode {
     }
 
 
+    @Override
+    public void accept(NodeVisitor visitor) {
+
+        visitor.visit(this);
+    }
+
+
     @Override
     public TextBlock getTextBlock() {
 
         return leafTextBlock;
     }
 
-
-    @Override
-    public String toString() {
-
-        return treeId + ": " + NodeType.PARAGRAPH + ": " + leafTextBlock.buildSummary();
-    }
-
-
-    @Override
-    public Map<Page, Rectangle2D> getBBox() {
-
-        if (bBoxCache == null) {
-            bBoxCache = GenericSemanticNode.super.getBBox();
-        }
-        return bBoxCache;
-    }
-
 }

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

@@ -1,47 +1,21 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
 
-import java.awt.geom.Rectangle2D;
-import java.util.HashSet;
-import java.util.List;
-import java.util.Map;
-import java.util.Set;
-
-import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
 import com.knecon.fforesight.service.layoutparser.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.NodeVisitor;
 
 import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
-import lombok.Builder;
 import lombok.Data;
 import lombok.EqualsAndHashCode;
 import lombok.experimental.FieldDefaults;
-import lombok.extern.slf4j.Slf4j;
+import lombok.experimental.SuperBuilder;
 
-@Slf4j
 @Data
-@Builder
+@SuperBuilder
 @AllArgsConstructor
 @FieldDefaults(level = AccessLevel.PRIVATE)
-public class Section implements GenericSemanticNode {
-
-    @Builder.Default
-    Set<LayoutEngine> engines = new HashSet<>(Set.of(LayoutEngine.ALGORITHM));
-    List<Integer> treeId;
-
-    TextBlock textBlock;
-    @EqualsAndHashCode.Exclude
-    DocumentTree documentTree;
-
-    @Builder.Default
-    @EqualsAndHashCode.Exclude
-    Set<RedactionEntity> entities = new HashSet<>();
-
-    @EqualsAndHashCode.Exclude
-    Map<Page, Rectangle2D> bBoxCache;
-
+@EqualsAndHashCode(callSuper = true)
+public class Section extends AbstractSemanticNode {
 
     @Override
     public NodeType getType() {
@@ -50,46 +24,30 @@ public class Section implements GenericSemanticNode {
     }
 
 
-    public boolean hasTables() {
+    public Headline getHeadline() {
 
-        return streamAllSubNodesOfType(NodeType.TABLE).findAny()
-                .isPresent();
+        return streamChildrenOfType(NodeType.HEADLINE).map(node -> (Headline) node)
+                .findFirst().orElseGet(() -> getParent().getHeadline());
     }
 
 
     @Override
-    public TextBlock getTextBlock() {
+    public void accept(NodeVisitor visitor) {
 
-        if (textBlock == null) {
-            textBlock = GenericSemanticNode.super.getTextBlock();
-        }
-        return textBlock;
+        visitor.visit(this);
+    }
+
+
+    public boolean hasTables() {
+
+        return streamAllSubNodesOfType(NodeType.TABLE).findAny().isPresent();
     }
 
 
     @Override
     public String toString() {
 
-        return treeId.toString() + ": " + NodeType.SECTION + ": " + this.getTextBlock().buildSummary();
-    }
-
-
-    public Headline getHeadline() {
-
-        return streamChildrenOfType(NodeType.HEADLINE)//
-                .map(node -> (Headline) node)//
-                .findFirst()//
-                .orElseGet(() -> getParent().getHeadline());
-    }
-
-
-    @Override
-    public Map<Page, Rectangle2D> getBBox() {
-
-        if (bBoxCache == null) {
-            bBoxCache = GenericSemanticNode.super.getBBox();
-        }
-        return bBoxCache;
+        return getTreeId() + ": " + NodeType.SECTION + ": " + this.getTextBlock().buildSummary();
     }
 
 }

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

@@ -14,14 +14,16 @@ import java.util.stream.Stream;
 
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.Boundary;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.TextRange;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.EntityType;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.TextEntity;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.AtomicTextBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlockCollector;
+import com.knecon.fforesight.service.layoutparser.processor.utils.BBoxMergingUtility;
 import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
 
 public interface SemanticNode {
 
@@ -42,7 +44,9 @@ public interface SemanticNode {
      */
     default TextBlock getTextBlock() {
 
-        return streamAllSubNodes().filter(SemanticNode::isLeaf).map(SemanticNode::getTextBlock).collect(new TextBlockCollector());
+        return streamAllSubNodes().filter(SemanticNode::isLeaf)
+                .map(SemanticNode::getTextBlock)
+                .collect(new TextBlockCollector());
     }
 
 
@@ -52,7 +56,7 @@ public interface SemanticNode {
      *
      * @return Set of all Entities associated with this Node
      */
-    Set<RedactionEntity> getEntities();
+    Set<TextEntity> getEntities();
 
 
     /**
@@ -68,7 +72,10 @@ public interface SemanticNode {
 
     default Page getFirstPage() {
 
-        return getTextBlock().getPages().stream().min(Comparator.comparingInt(Page::getNumber)).orElseThrow(() -> new IllegalStateException("SemanticNode has no Page!"));
+        return getTextBlock().getPages()
+                .stream()
+                .min(Comparator.comparingInt(Page::getNumber))
+                .orElseThrow(() -> new IllegalStateException("SemanticNode has no Page!"));
     }
 
 
@@ -77,18 +84,19 @@ public interface SemanticNode {
      *
      * @return Set of PageNodes this node appears on.
      */
-    default Set<Page> getPages(Boundary boundary) {
+    default Set<Page> getPages(TextRange textRange) {
 
-        if (!getBoundary().contains(boundary)) {
-            throw new IllegalArgumentException(format("%s which was used to query for pages is not contained in the %s of this node!", boundary, getBoundary()));
+        if (!getBoundary().contains(textRange)) {
+            throw new IllegalArgumentException(format("%s which was used to query for pages is not contained in the %s of this node!", textRange, getBoundary()));
         }
-        return getTextBlock().getPages(boundary);
+        return getTextBlock().getPages(textRange);
     }
 
 
     default boolean isOnPage(int pageNumber) {
 
-        return getPages().stream().anyMatch(page -> page.getNumber() == pageNumber);
+        return getPages().stream()
+                .anyMatch(page -> page.getNumber() == pageNumber);
     }
 
 
@@ -203,7 +211,9 @@ public interface SemanticNode {
      */
     default boolean hasEntitiesOfType(String type) {
 
-        return getEntities().stream().filter(entity -> entity.getEntityType().equals(EntityType.ENTITY)).anyMatch(redactionEntity -> redactionEntity.getType().equals(type));
+        return getEntities().stream()
+                .filter(entity -> entity.getEntityType().equals(EntityType.ENTITY))
+                .anyMatch(redactionEntity -> redactionEntity.getType().equals(type));
     }
 
 
@@ -213,9 +223,11 @@ public interface SemanticNode {
      * @param type string representing the type of entities to return
      * @return List of RedactionEntities of any the type
      */
-    default List<RedactionEntity> getEntitiesOfType(String type) {
+    default List<TextEntity> getEntitiesOfType(String type) {
 
-        return getEntities().stream().filter(redactionEntity -> redactionEntity.getType().equals(type)).toList();
+        return getEntities().stream()
+                .filter(redactionEntity -> redactionEntity.getType().equals(type))
+                .toList();
     }
 
 
@@ -225,9 +237,11 @@ public interface SemanticNode {
      * @param types A list of strings representing the types of entities to return
      * @return List of RedactionEntities of any provided type
      */
-    default List<RedactionEntity> getEntitiesOfType(List<String> types) {
+    default List<TextEntity> getEntitiesOfType(List<String> types) {
 
-        return getEntities().stream().filter(redactionEntity -> redactionEntity.isAnyType(types)).toList();
+        return getEntities().stream()
+                .filter(redactionEntity -> redactionEntity.isAnyType(types))
+                .toList();
     }
 
 
@@ -279,7 +293,8 @@ public interface SemanticNode {
      */
     default boolean containsStrings(List<String> strings) {
 
-        return strings.stream().allMatch(this::containsString);
+        return strings.stream()
+                .allMatch(this::containsString);
     }
 
 
@@ -303,7 +318,8 @@ public interface SemanticNode {
      */
     default boolean containsAnyString(List<String> strings) {
 
-        return strings.stream().anyMatch(this::containsString);
+        return strings.stream()
+                .anyMatch(this::containsString);
     }
 
 
@@ -315,7 +331,8 @@ public interface SemanticNode {
      */
     default boolean containsAnyStringIgnoreCase(List<String> strings) {
 
-        return strings.stream().anyMatch(this::containsStringIgnoreCase);
+        return strings.stream()
+                .anyMatch(this::containsStringIgnoreCase);
     }
 
 
@@ -323,19 +340,19 @@ public interface SemanticNode {
      * This function is used during insertion of EntityNodes into the graph, it checks if the boundary of the RedactionEntity intersects or even contains the RedactionEntity.
      * It sets the fields accordingly and recursively calls this function on all its children.
      *
-     * @param redactionEntity RedactionEntity, which is being inserted into the graph
+     * @param textEntity RedactionEntity, which is being inserted into the graph
      */
-    default void addThisToEntityIfIntersects(RedactionEntity redactionEntity) {
+    default void addThisToEntityIfIntersects(TextEntity textEntity) {
 
         TextBlock textBlock = getTextBlock();
-        if (textBlock.getBoundary().intersects(redactionEntity.getBoundary())) {
-            if (textBlock.containsBoundary(redactionEntity.getBoundary())) {
-                redactionEntity.setDeepestFullyContainingNode(this);
+        if (textBlock.getTextRange().intersects(textEntity.getTextRange())) {
+            if (textBlock.containsBoundary(textEntity.getTextRange())) {
+                textEntity.setDeepestFullyContainingNode(this);
             }
 
-            redactionEntity.addIntersectingNode(this);
-            streamChildren().filter(semanticNode -> semanticNode.getBoundary().intersects(redactionEntity.getBoundary()))
-                    .forEach(node -> node.addThisToEntityIfIntersects(redactionEntity));
+            textEntity.addIntersectingNode(this);
+            streamChildren().filter(semanticNode -> semanticNode.getBoundary().intersects(textEntity.getTextRange()))
+                    .forEach(node -> node.addThisToEntityIfIntersects(textEntity));
         }
     }
 
@@ -386,7 +403,8 @@ public interface SemanticNode {
      */
     default Stream<SemanticNode> streamAllSubNodes() {
 
-        return getDocumentTree().allSubEntriesInOrder(getTreeId()).map(DocumentTree.Entry::getNode);
+        return getDocumentTree().allSubEntriesInOrder(getTreeId())
+                .map(DocumentTree.Entry::getNode);
     }
 
 
@@ -397,7 +415,9 @@ public interface SemanticNode {
      */
     default Stream<SemanticNode> streamAllSubNodesOfType(NodeType nodeType) {
 
-        return getDocumentTree().allSubEntriesInOrder(getTreeId()).filter(entry -> entry.getType().equals(nodeType)).map(DocumentTree.Entry::getNode);
+        return getDocumentTree().allSubEntriesInOrder(getTreeId())
+                .filter(entry -> entry.getType().equals(nodeType))
+                .map(DocumentTree.Entry::getNode);
     }
 
 
@@ -406,9 +426,9 @@ public interface SemanticNode {
      *
      * @return Boundary of this Node's TextBlock
      */
-    default Boundary getBoundary() {
+    default TextRange getBoundary() {
 
-        return getTextBlock().getBoundary();
+        return getTextBlock().getTextRange();
     }
 
 
@@ -453,17 +473,19 @@ public interface SemanticNode {
      */
     private Map<Page, Rectangle2D> getBBoxFromChildren() {
 
-        Map<Page, Rectangle2D> bBoxPerPage = new HashMap<>();
-        List<Map<Page, Rectangle2D>> childrenBBoxes = streamChildren().map(SemanticNode::getBBox).toList();
-        Set<Page> pages = childrenBBoxes.stream().flatMap(map -> map.keySet().stream()).collect(Collectors.toSet());
-        for (Page page : pages) {
-            Rectangle2D bBoxOnPage = childrenBBoxes.stream()
-                    .filter(childBboxPerPage -> childBboxPerPage.containsKey(page))
-                    .map(childBboxPerPage -> childBboxPerPage.get(page))
-                    .collect(RectangleTransformations.collectBBox());
-            bBoxPerPage.put(page, bBoxOnPage);
+        List<Map<Page, Rectangle2D>> childrenBBoxes = streamChildren().filter(child -> !isFullPageImage(child))
+                .map(SemanticNode::getBBox)
+                .toList();
+        return BBoxMergingUtility.mergeBBoxes(childrenBBoxes);
+    }
+
+
+    private static boolean isFullPageImage(SemanticNode child) {
+
+        if (!child.getType().equals(NodeType.IMAGE)) {
+            return false;
         }
-        return bBoxPerPage;
+        return ((Image) child).isFullPageImage();
     }
 
 
@@ -473,9 +495,27 @@ public interface SemanticNode {
     private Map<Page, Rectangle2D> getBBoxFromLeafTextBlock() {
 
         Map<Page, Rectangle2D> bBoxPerPage = new HashMap<>();
-        Map<Page, List<AtomicTextBlock>> atomicTextBlockPerPage = getTextBlock().getAtomicTextBlocks().stream().collect(Collectors.groupingBy(AtomicTextBlock::getPage));
+        Map<Page, List<AtomicTextBlock>> atomicTextBlockPerPage = getTextBlock().getAtomicTextBlocks()
+                .stream()
+                .collect(Collectors.groupingBy(AtomicTextBlock::getPage));
         atomicTextBlockPerPage.forEach((page, atbs) -> bBoxPerPage.put(page, RectangleTransformations.bBoxUnionAtomicTextBlock(atbs)));
         return bBoxPerPage;
     }
 
+
+    void accept(NodeVisitor visitor);
+
+
+    /**
+     * Checks wether this SemanticNode appears on a single page only, and if that page is the provided one.
+     *
+     * @param page the page to check
+     * @return true, when SemanticNode is on a single page only and the page is the provided page. Otherwise, false.
+     */
+    default boolean onlyOnPage(Page page) {
+
+        Set<Page> pages = getPages();
+        return pages.size() == 1 && pages.contains(page);
+    }
+
 }

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

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

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

@@ -15,7 +15,8 @@ import java.util.stream.Stream;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.TextEntity;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlockCollector;
 
@@ -43,7 +44,7 @@ public class Table implements SemanticNode {
 
     @Builder.Default
     @EqualsAndHashCode.Exclude
-    Set<RedactionEntity> entities = new HashSet<>();
+    Set<TextEntity> entities = new HashSet<>();
 
     @EqualsAndHashCode.Exclude
     Map<Page, Rectangle2D> bBoxCache;
@@ -54,7 +55,7 @@ public class Table implements SemanticNode {
      * @param strings Strings to check whether a row contains them
      * @return Stream of all entities in this table, that appear in a row, which contains any of the provided strings
      */
-    public Stream<RedactionEntity> streamEntitiesWhereRowContainsStringsIgnoreCase(List<String> strings) {
+    public Stream<TextEntity> streamEntitiesWhereRowContainsStringsIgnoreCase(List<String> strings) {
 
         return IntStream.range(0, numberOfRows).boxed()
                 .filter(row -> rowContainsStringsIgnoreCase(row, strings))
@@ -88,7 +89,7 @@ public class Table implements SemanticNode {
      * @param value  the string which the table cell should contain
      * @return a stream of all entities, which appear in a row where at least one cell has the provided header and the provided value.
      */
-    public Stream<RedactionEntity> streamEntitiesWhereRowHasHeaderAndValue(String header, String value) {
+    public Stream<TextEntity> streamEntitiesWhereRowHasHeaderAndValue(String header, String value) {
 
         List<Integer> vertebrateStudyCols = streamHeaders().filter(headerNode -> headerNode.containsString(header))
                 .map(TableCell::getCol)
@@ -107,7 +108,7 @@ public class Table implements SemanticNode {
      * @param values the strings which the table cell should contain
      * @return a stream of all entities, which appear in a row where at least one cell has the provided header and any provided value.
      */
-    public Stream<RedactionEntity> streamEntitiesWhereRowHasHeaderAndAnyValue(String header, List<String> values) {
+    public Stream<TextEntity> streamEntitiesWhereRowHasHeaderAndAnyValue(String header, List<String> values) {
 
         List<Integer> colsWithHeader = streamHeaders().filter(headerNode -> headerNode.containsString(header))
                 .map(TableCell::getCol)
@@ -125,7 +126,7 @@ public class Table implements SemanticNode {
      * @param types type strings to check whether a row contains an entity like them
      * @return Stream of all entities in this table, that appear in a row, which contains at least one entity with any of the provided types.
      */
-    public Stream<RedactionEntity> streamEntitiesWhereRowContainsEntitiesOfType(List<String> types) {
+    public Stream<TextEntity> streamEntitiesWhereRowContainsEntitiesOfType(List<String> types) {
 
         List<Integer> rowsWithEntityOfType = IntStream.range(0, numberOfRows).boxed()
                 .filter(rowNumber -> streamEntityTypesInRow(rowNumber).anyMatch(existingType -> types.stream()
@@ -145,7 +146,7 @@ public class Table implements SemanticNode {
      * @param types type strings to check whether a row doesn't contain an entity like it
      * @return Stream of all entities in this table, that appear in a row, which does not contain any entity with any of the provided types.
      */
-    public Stream<RedactionEntity> streamEntitiesWhereRowContainsNoEntitiesOfType(List<String> types) {
+    public Stream<TextEntity> streamEntitiesWhereRowContainsNoEntitiesOfType(List<String> types) {
 
         List<Integer> rowsWithNoEntityOfType = IntStream.range(0, numberOfRows).boxed()
                 .filter(rowNumber -> streamEntityTypesInRow(rowNumber).noneMatch(existingType -> types.stream()
@@ -163,7 +164,7 @@ public class Table implements SemanticNode {
 
         return streamRow(rowNumber).map(TableCell::getEntities)
                 .flatMap(Collection::stream)
-                .map(RedactionEntity::getType)
+                .map(TextEntity::getType)
                 .distinct();
     }
 
@@ -304,12 +305,12 @@ public class Table implements SemanticNode {
      * Finds all entities of the provided type, which appear in the same row that the provided entity appears in.
      *
      * @param type            the type of entities to search for
-     * @param redactionEntity the entity, which appears in the row to search
+     * @param textEntity the entity, which appears in the row to search
      * @return List of all entities of the provided type, which appear in the same row that the provided entity appears in.
      */
-    public List<RedactionEntity> getEntitiesOfTypeInSameRow(String type, RedactionEntity redactionEntity) {
+    public List<TextEntity> getEntitiesOfTypeInSameRow(String type, TextEntity textEntity) {
 
-        return redactionEntity.getIntersectingNodes()
+        return textEntity.getIntersectingNodes()
                 .stream()
                 .filter(node -> node instanceof TableCell)
                 .map(node -> (TableCell) node)
@@ -353,4 +354,10 @@ public class Table implements SemanticNode {
         return bBoxCache;
     }
 
+
+    @Override
+    public void accept(NodeVisitor visitor) {
+        visitor.visit(this);
+    }
+
 }

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

@@ -2,34 +2,27 @@ package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
 
 import java.awt.geom.Rectangle2D;
 import java.util.HashMap;
-import java.util.HashSet;
-import java.util.List;
 import java.util.Map;
-import java.util.Set;
 
-import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlockCollector;
 
 import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
-import lombok.Builder;
 import lombok.Data;
 import lombok.EqualsAndHashCode;
 import lombok.experimental.FieldDefaults;
+import lombok.experimental.SuperBuilder;
 
 @Data
-@Builder
+@SuperBuilder
 @AllArgsConstructor
+@EqualsAndHashCode(callSuper = true)
 @FieldDefaults(level = AccessLevel.PRIVATE)
-public class TableCell implements GenericSemanticNode {
+public class TableCell extends AbstractSemanticNode {
 
-    @Builder.Default
-    Set<LayoutEngine> engines = new HashSet<>(Set.of(LayoutEngine.ALGORITHM));
-    List<Integer> treeId;
     int row;
     int col;
     boolean header;
@@ -40,13 +33,6 @@ public class TableCell implements GenericSemanticNode {
 
     TextBlock textBlock;
 
-    @EqualsAndHashCode.Exclude
-    DocumentTree documentTree;
-
-    @Builder.Default
-    @EqualsAndHashCode.Exclude
-    Set<RedactionEntity> entities = new HashSet<>();
-
 
     @Override
     public Map<Page, Rectangle2D> getBBox() {
@@ -57,6 +43,13 @@ public class TableCell implements GenericSemanticNode {
     }
 
 
+    @Override
+    public void accept(NodeVisitor visitor) {
+
+        visitor.visit(this);
+    }
+
+
     @Override
     public NodeType getType() {
 
@@ -96,7 +89,7 @@ public class TableCell implements GenericSemanticNode {
     @Override
     public String toString() {
 
-        return treeId + ": " + NodeType.TABLE_CELL + ": " + this.getTextBlock().buildSummary();
+        return getTreeId() + ": " + NodeType.TABLE_CELL + ": " + this.getTextBlock().buildSummary();
     }
 
 }

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

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

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

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

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

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

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

@@ -10,8 +10,8 @@ import lombok.NonNull;
 import lombok.RequiredArgsConstructor;
 
 @Data
-@RequiredArgsConstructor
 @AllArgsConstructor
+@RequiredArgsConstructor
 public class ClassifiedImage {
 
     @NonNull
@@ -20,9 +20,19 @@ public class ClassifiedImage {
     private ImageType imageType;
     private boolean sourceByAi;
     private boolean isAppendedToSection;
-    @NonNull
     private boolean hasTransparency;
-    @NonNull
     private int page;
+    @NonNull
+    private String representation;
+
+
+    public ClassifiedImage(@NonNull Rectangle2D position, @NonNull ImageType imageType, boolean hasTransparency, int page, @NonNull String representation) {
+
+        this.position = position;
+        this.imageType = imageType;
+        this.hasTransparency = hasTransparency;
+        this.page = page;
+        this.representation = representation;
+    }
 
 }

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

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

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

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

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

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

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

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

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

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

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

@@ -0,0 +1,259 @@
+package com.knecon.fforesight.service.layoutparser.processor.model.outline;
+
+import java.util.ArrayList;
+import java.util.Collections;
+import java.util.Iterator;
+import java.util.List;
+
+import org.springframework.stereotype.Service;
+
+import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
+import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationDocument;
+import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationFooter;
+import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationHeader;
+import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
+import com.knecon.fforesight.service.layoutparser.processor.model.PageBlockType;
+import com.knecon.fforesight.service.layoutparser.processor.model.image.ClassifiedImage;
+import com.knecon.fforesight.service.layoutparser.processor.model.table.Cell;
+import com.knecon.fforesight.service.layoutparser.processor.model.table.TablePageBlock;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
+
+import lombok.extern.slf4j.Slf4j;
+
+@Slf4j
+@Service
+public class TOCEnrichmentService {
+
+    public void assignSectionBlocksAndImages(ClassificationDocument document) {
+
+        TableOfContents toc = document.getTableOfContents();
+        Iterator<TableOfContentItem> iterator = toc.iterator();
+        TableOfContentItem currentTOCItem = null;
+        if (iterator.hasNext()) {
+            currentTOCItem = iterator.next();
+        }
+        List<AbstractPageBlock> startBlocks = new ArrayList<>();
+        List<ClassifiedImage> startImages = new ArrayList<>();
+        TableOfContentItem currentSection = null;
+        boolean foundFirstHeadline = false;
+
+        List<ClassificationHeader> headers = new ArrayList<>();
+        List<ClassificationFooter> footers = new ArrayList<>();
+        TablePageBlock previousTable = null;
+        List<TableOfContentItem> lastFoundTOCItems = new ArrayList<>();
+
+        for (ClassificationPage page : document.getPages()) {
+            List<TableOfContentItem> currentPageTOCItems = new ArrayList<>();
+            List<TextPageBlock> header = new ArrayList<>();
+            List<TextPageBlock> footer = new ArrayList<>();
+            for (AbstractPageBlock current : page.getTextBlocks()) {
+
+                if (current.getClassification() == null) {
+                    continue;
+                }
+
+                current.setPage(page.getPageNumber());
+
+                if (current.getClassification().equals(PageBlockType.HEADER)) {
+                    header.add((TextPageBlock) current);
+                    continue;
+                }
+
+                if (current.getClassification().equals(PageBlockType.FOOTER)) {
+                    footer.add((TextPageBlock) current);
+                    continue;
+                }
+
+                if (current instanceof TablePageBlock table) {
+                    if (previousTable != null) {
+                        mergeTableMetadata(table, previousTable);
+                    }
+                    previousTable = table;
+                }
+
+                if (current instanceof TextPageBlock && currentTOCItem != null && currentTOCItem.getHeadline().getText().equals(current.getText())) {
+                    if (!foundFirstHeadline) {
+                        foundFirstHeadline = true;
+                    }
+                    currentSection = currentTOCItem;
+                    currentTOCItem.getSectionBlocks().add(current);
+                    currentPageTOCItems.add(currentTOCItem);
+
+                    if (iterator.hasNext()) {
+                        currentTOCItem = iterator.next();
+                    }
+                } else if (!foundFirstHeadline) {
+                    startBlocks.add(current);
+                } else {
+                    currentSection.getSectionBlocks().add(current);
+                }
+            }
+
+            if (!currentPageTOCItems.isEmpty()) {
+                lastFoundTOCItems = currentPageTOCItems;
+            }
+
+            for (ClassifiedImage image : page.getImages()) {
+
+                Double xMin = null;
+                Double yMin = null;
+                Double xMax = null;
+                Double yMax = null;
+
+                for (TableOfContentItem tocItem : lastFoundTOCItems) {
+                    var headline = tocItem.getHeadline();
+
+                    if (headline.getPage() != page.getPageNumber()) {
+                        continue;
+                    }
+
+                    if (headline.getMinX() < headline.getMaxX()) {
+                        if (xMin == null || headline.getMinX() < xMin) {
+                            xMin = headline.getMinX();
+                        }
+                        if (xMax == null || headline.getMaxX() > xMax) {
+                            xMax = headline.getMaxX();
+                        }
+                    } else {
+                        if (xMin == null || headline.getMaxX() < xMin) {
+                            xMin = headline.getMaxX();
+                        }
+                        if (xMax == null || headline.getMinX() > xMax) {
+                            xMax = headline.getMinX();
+                        }
+                    }
+
+                    if (headline.getMinY() < headline.getMaxY()) {
+                        if (yMin == null || headline.getMinY() < yMin) {
+                            yMin = headline.getMinY();
+                        }
+                        if (yMax == null || headline.getMaxY() > yMax) {
+                            yMax = headline.getMaxY();
+                        }
+                    } else {
+                        if (yMin == null || headline.getMaxY() < yMin) {
+                            yMin = headline.getMaxY();
+                        }
+                        if (yMax == null || headline.getMinY() > yMax) {
+                            yMax = headline.getMinY();
+                        }
+                    }
+
+                    log.debug("Image position x: {}, y: {}", image.getPosition().getX(), image.getPosition().getY());
+                    log.debug("Headline position xMin: {}, xMax: {}, yMin: {}, yMax: {}", xMin, xMax, yMin, yMax);
+
+                    if (image.getPosition().getX() >= xMin && image.getPosition().getX() <= xMax && image.getPosition().getY() >= yMin && image.getPosition().getY() <= yMax) {
+                        tocItem.getImages().add(image);
+                        image.setAppendedToSection(true);
+                        break;
+                    }
+                }
+                if (!image.isAppendedToSection()) {
+                    log.debug("Image uses last found section");
+                    if (!lastFoundTOCItems.isEmpty()) {
+                        lastFoundTOCItems.get(lastFoundTOCItems.size() - 1).getImages().add(image);
+                    } else {
+                        startImages.add(image);
+                    }
+                    image.setAppendedToSection(true);
+                }
+            }
+
+            if (!header.isEmpty()) {
+                headers.add(new ClassificationHeader(header));
+            }
+            if (!footer.isEmpty()) {
+                footers.add(new ClassificationFooter(footer));
+            }
+        }
+
+        if (!startBlocks.isEmpty()) {
+            TableOfContentItem unassigned = new TableOfContentItem(null);
+            unassigned.setSectionBlocks(startBlocks);
+            unassigned.setImages(startImages);
+            document.getTableOfContents().getMainSections().add(0, unassigned);
+        }
+        document.setHeaders(headers);
+        document.setFooters(footers);
+    }
+
+
+    private void mergeTableMetadata(TablePageBlock currentTable, TablePageBlock previousTable) {
+
+        // Distribute header information for subsequent tables
+        if (previousTable != null && hasInvalidHeaderInformation(currentTable) && hasValidHeaderInformation(previousTable)) {
+            List<Cell> previousTableNonHeaderRow = getRowWithNonHeaderCells(previousTable);
+            List<Cell> tableNonHeaderRow = getRowWithNonHeaderCells(currentTable);
+            // Allow merging of tables if header row is separated from first logical non-header row
+            if (previousTableNonHeaderRow.isEmpty()
+                && previousTable.getRowCount() == 1
+                && previousTable.getRows()
+                           .get(0).size() == tableNonHeaderRow.size()) {
+                previousTableNonHeaderRow = previousTable.getRows()
+                        .get(0)
+                        .stream()
+                        .map(cell -> {
+                            Cell fakeCell = Cell.copy(cell);
+                            fakeCell.setHeaderCells(Collections.singletonList(cell));
+                            return fakeCell;
+                        })
+                        .toList();
+            }
+            if (previousTableNonHeaderRow.size() == tableNonHeaderRow.size()) {
+                for (int i = currentTable.getRowCount() - 1; i >= 0; i--) { // Non header rows are most likely at bottom of table
+                    List<Cell> row = currentTable.getRows()
+                            .get(i);
+                    if (row.size() == tableNonHeaderRow.size() && row.stream()
+                            .allMatch(cell -> cell.getHeaderCells().isEmpty())) {
+                        for (int j = 0; j < row.size(); j++) {
+                            row.get(j).setHeaderCells(previousTableNonHeaderRow.get(j).getHeaderCells());
+                        }
+                    }
+                }
+            }
+        }
+    }
+
+
+    private boolean hasValidHeaderInformation(TablePageBlock table) {
+
+        return !hasInvalidHeaderInformation(table);
+    }
+
+
+    private boolean hasInvalidHeaderInformation(TablePageBlock table) {
+
+        return table.getRows()
+                .stream()
+                .flatMap(row -> row.stream()
+                        .filter(cell -> !cell.getHeaderCells().isEmpty()))
+                .findAny().isEmpty();
+
+    }
+
+
+    private List<Cell> getRowWithNonHeaderCells(TablePageBlock table) {
+
+        for (int i = table.getRowCount() - 1; i >= 0; i--) { // Non header rows are most likely at bottom of table
+            List<Cell> row = table.getRows()
+                    .get(i);
+            if (row.size() == 1) {
+                continue;
+            }
+            boolean allNonHeader = true;
+            for (Cell cell : row) {
+                if (cell.isHeaderCell()) {
+                    allNonHeader = false;
+                    break;
+                }
+            }
+            if (allNonHeader) {
+                return row;
+            }
+        }
+
+        return Collections.emptyList();
+
+    }
+
+}

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

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

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

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

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

@@ -35,14 +35,14 @@ public class Cell extends BoundingBox {
 
     public Cell(Point2D topLeft, Point2D bottomRight) {
 
-        this.bBoxInitialUserSpace = new Rectangle2D.Double(topLeft.getX(), topLeft.getY(), (bottomRight.getX() - topLeft.getX()), (bottomRight.getY() - topLeft.getY()));
-        this.bBox = bBoxInitialUserSpace;
+        this.bBoxPdf = new Rectangle2D.Double(topLeft.getX(), topLeft.getY(), (bottomRight.getX() - topLeft.getX()), (bottomRight.getY() - topLeft.getY()));
+        this.bBox = bBoxPdf;
     }
 
 
     public Cell(Rectangle2D bBoxInitialUserSpace, AffineTransform initialUserSpaceToJava) {
 
-        this.bBoxInitialUserSpace = bBoxInitialUserSpace;
+        this.bBoxPdf = bBoxInitialUserSpace;
         this.bBox = initialUserSpaceToJava.createTransformedShape(bBoxInitialUserSpace).getBounds2D();
     }
 
@@ -50,7 +50,7 @@ public class Cell extends BoundingBox {
     public static Cell copy(Cell cell) {
 
         Cell copy = new Cell();
-        copy.bBoxInitialUserSpace = cell.bBoxInitialUserSpace;
+        copy.bBoxPdf = cell.bBoxPdf;
         copy.bBox = cell.bBox;
         return copy;
     }
@@ -87,7 +87,7 @@ public class Cell extends BoundingBox {
 
         }
 
-        return TextNormalizationUtilities.removeHyphenLineBreaks(sb.toString()).replaceAll("\n", " ").replaceAll(" {2}", " ");
+        return TextNormalizationUtilities.cleanString(sb.toString());
     }
 
 

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

@@ -70,7 +70,7 @@ public class CleanRulings {
 
     public boolean lineBetween(BoundingBox a, BoundingBox b) {
 
-        return lineBetween(a.getBBoxInitialUserSpace(), b.getBBoxInitialUserSpace());
+        return lineBetween(a.getBBoxPdf(), b.getBBoxPdf());
     }
 
 

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

@@ -263,8 +263,8 @@ public class TablePageBlock extends AbstractPageBlock {
 
                         cells.stream()
                                 .map(originalCell -> new CellWithIntersection(originalCell,
-                                                                              RectangleTransformations.calculateIntersectedArea(cellFromGridStructure.getBBoxInitialUserSpace(),
-                                                                                                                                originalCell.getBBoxInitialUserSpace())))
+                                                                              RectangleTransformations.calculateIntersectedArea(cellFromGridStructure.getBBoxPdf(),
+                                                                                                                                originalCell.getBBoxPdf())))
                                 .filter(cellWithIntersection -> cellWithIntersection.intersectedArea > 0)
                                 .filter(cellWithIntersection -> cellWithIntersection.originalCell.getArea() > cellWithIntersection.intersectedArea * CELL_AREA_CONTAINED_THRESHOLD)
                                 .max(Comparator.comparing(CellWithIntersection::intersectedArea))

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

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

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

@@ -5,86 +5,72 @@ import java.awt.geom.Rectangle2D;
 
 import org.apache.pdfbox.text.TextPosition;
 
-import com.fasterxml.jackson.annotation.JsonIgnore;
-import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.BoundingBox;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.TextBoundingBox;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.utils.FastAtan2;
 
+import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
-import lombok.Builder;
 import lombok.Data;
+import lombok.EqualsAndHashCode;
 import lombok.NoArgsConstructor;
 import lombok.SneakyThrows;
+import lombok.experimental.FieldDefaults;
+import lombok.experimental.SuperBuilder;
 
 @Data
-@Builder
+@SuperBuilder
 @NoArgsConstructor
 @AllArgsConstructor
-public class RedTextPosition extends BoundingBox {
+@EqualsAndHashCode(callSuper = true)
+@FieldDefaults(level = AccessLevel.PRIVATE)
+public class RedTextPosition extends TextBoundingBox {
 
     public final static int HEIGHT_PADDING = 2;
 
-    private Rectangle2D.Float bBoxDirAdj; // adjusted to text rotation
+    String unicode;
 
-    @JsonIgnore
-    private int rotation;
+    // estimated using the TextMatrix in radians
+    float exactDir;
 
-    @JsonIgnore
-    private float pageHeight;
+    float widthOfSpace;
 
-    @JsonIgnore
-    private float pageWidth;
+    float fontSizeInPt;
 
-    private String unicode;
-
-    @JsonIgnore
-    private float dir;
-
-    // not used in reanalysis
-    @JsonIgnore
-    private float widthOfSpace;
-
-    // not used in reanalysis
-    @JsonIgnore
-    private float fontSizeInPt;
-
-    // not used in reanalysis
-    @JsonIgnore
-    private String fontName;
+    String fontName;
 
 
     @SneakyThrows
     public static RedTextPosition fromTextPosition(TextPosition textPosition) {
 
         var pos = new RedTextPosition();
-        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());
+        pos.setExactDir((float) FastAtan2.fastAtan2(textPosition.getTextMatrix().getShearY(), textPosition.getTextMatrix().getScaleX()));
+        pos.setDir(TextDirection.fromDegrees(textPosition.getDir()));
 
         //TODO: There is a mismatch in the java coords of the text and the rulings,
         // I guess if we start with the initial user space positions and transform them the same way we do the rulings it would work.
         pos.setBBox(new Rectangle2D.Float(textPosition.getX(), textPosition.getY(), textPosition.getWidthDirAdj(), textPosition.getHeight()));
 
-        float textHeight = textPosition.getHeight() + HEIGHT_PADDING;
+        float textHeight = textPosition.getHeight() + 2 * HEIGHT_PADDING;
         Rectangle2D.Float dirAdjPosition = new Rectangle2D.Float(textPosition.getXDirAdj(),
                                                                  textPosition.getYDirAdj() - textHeight,
                                                                  textPosition.getWidthDirAdj(),
                                                                  textHeight + HEIGHT_PADDING);
+
         pos.setBBoxDirAdj(dirAdjPosition);
 
         AffineTransform affineTransform = getRotationMatrix(TextDirection.fromDegrees(textPosition.getDir()), textPosition.getPageWidth(), textPosition.getPageHeight());
         Rectangle2D bBoxInitialUserSpace = affineTransform.createTransformedShape(dirAdjPosition).getBounds2D();
 
-        pos.setBBoxInitialUserSpace(bBoxInitialUserSpace); // These are definitely correct
+        pos.setBBoxPdf(bBoxInitialUserSpace); // These are definitely correct
 
         return pos;
     }
 
 
-
     private static AffineTransform getRotationMatrix(TextDirection textDirection, float pageWidth, float pageHeight) {
 
         AffineTransform transform = new AffineTransform();
@@ -103,32 +89,4 @@ public class RedTextPosition extends BoundingBox {
         return transform;
     }
 
-
-    @JsonIgnore
-    public float getXDirAdj() {
-
-        return this.bBoxDirAdj.x;
-    }
-
-
-    @JsonIgnore
-    public float getYDirAdj() {
-
-        return this.bBoxDirAdj.y;
-    }
-
-
-    @JsonIgnore
-    public float getWidthDirAdj() {
-
-        return this.bBoxDirAdj.width;
-    }
-
-
-    @JsonIgnore
-    public float getHeightDir() {
-
-        return this.bBoxDirAdj.height;
-    }
-
 }

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

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

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

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

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

@@ -8,6 +8,7 @@ import com.fasterxml.jackson.annotation.JsonIgnore;
 import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.FloatFrequencyCounter;
 import com.knecon.fforesight.service.layoutparser.processor.model.PageBlockType;
+import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
 import com.knecon.fforesight.service.layoutparser.processor.utils.TextNormalizationUtilities;
 
 import lombok.AllArgsConstructor;
@@ -26,17 +27,19 @@ public class TextPageBlock extends AbstractPageBlock {
     @Builder.Default
     private List<TextPositionSequence> sequences = new ArrayList<>();
 
+    private Rectangle2D bBoxDirAdj;
+
     private String mostPopularWordFont;
 
     private String mostPopularWordStyle;
 
-    private float mostPopularWordFontSize;
+    private double mostPopularWordFontSize;
 
-    private float mostPopularWordHeight;
+    private double mostPopularWordHeight;
 
-    private float mostPopularWordSpaceWidth;
+    private double mostPopularWordSpaceWidth;
 
-    private float highestFontSize;
+    private double highestFontSize;
 
     private PageBlockType classification;
 
@@ -46,12 +49,13 @@ public class TextPageBlock extends AbstractPageBlock {
     public TextPageBlock(List<TextPositionSequence> sequences) {
 
         this.sequences = sequences;
-        calculateFrequencyCounters();
+        if (!sequences.isEmpty()) {
+            calculateFrequencyCounters();
+        }
         calculateBBox();
     }
 
 
-    @JsonIgnore
     public TextDirection getDir() {
 
         return sequences.get(0).getDir();
@@ -62,13 +66,24 @@ public class TextPageBlock extends AbstractPageBlock {
 
         if (sequences == null) {
             this.bBox = new Rectangle2D.Double();
-            this.bBoxInitialUserSpace = new Rectangle2D.Double();
+            this.bBoxPdf = new Rectangle2D.Double();
+            this.bBoxDirAdj = new Rectangle2D.Double();
             return;
         }
+        this.bBoxDirAdj = sequences.stream()
+                .map(TextPositionSequence::getBBoxDirAdj)
+                .collect(RectangleTransformations.collectBBox());
+
         setToBBoxOfComponents(sequences);
     }
 
 
+    public void recalculateBBox() {
+
+        calculateBBox();
+    }
+
+
     public static TextPageBlock merge(List<TextPageBlock> textBlocksToMerge) {
 
         if (textBlocksToMerge.isEmpty()) {
@@ -198,7 +213,7 @@ public class TextPageBlock extends AbstractPageBlock {
             previous = word;
         }
 
-        return TextNormalizationUtilities.removeHyphenLineBreaks(sb.toString());
+        return TextNormalizationUtilities.removeHyphenLinebreaks(sb.toString());
     }
 
 

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

@@ -1,6 +1,7 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.text;
 
-import java.awt.geom.Rectangle2D;
+import static com.knecon.fforesight.service.layoutparser.processor.model.text.RedTextPosition.HEIGHT_PADDING;
+
 import java.util.ArrayList;
 import java.util.List;
 import java.util.Locale;
@@ -8,8 +9,7 @@ import java.util.stream.Collectors;
 
 import org.apache.pdfbox.text.TextPosition;
 
-import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.BoundingBox;
-import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.TextBoundingBox;
 
 import lombok.AllArgsConstructor;
 import lombok.Builder;
@@ -23,23 +23,21 @@ import lombok.extern.slf4j.Slf4j;
 @Builder
 @NoArgsConstructor
 @AllArgsConstructor
-@EqualsAndHashCode(onlyExplicitlyIncluded = true, callSuper = false)
-public class TextPositionSequence extends BoundingBox implements CharSequence {
+@EqualsAndHashCode(onlyExplicitlyIncluded = true, callSuper = true) // needs the bbox to be unique
+public class TextPositionSequence extends TextBoundingBox implements CharSequence {
 
-    public static final int HEIGHT_PADDING = 2;
+    public static final String STANDARD = "standard";
+    public static final String BOLD_ITALIC = "bold, italic";
+    public static final String BOLD = "bold";
+    public static final String ITALIC = "italic";
 
     @EqualsAndHashCode.Include
     private int page;
 
     @EqualsAndHashCode.Include
+    @Builder.Default
     private List<RedTextPosition> textPositions = new ArrayList<>();
 
-    private Rectangle2D bBoxDirAdj;
-    @EqualsAndHashCode.Include
-    private TextDirection dir;
-    private int rotation;
-    private float pageHeight;
-    private float pageWidth;
     private boolean isParagraphStart;
     private boolean strikethrough;
     private boolean underline;
@@ -51,10 +49,6 @@ public class TextPositionSequence extends BoundingBox implements CharSequence {
                 .map(RedTextPosition::fromTextPosition)
                 .collect(Collectors.toList());
         this.page = pageNumber;
-        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();
         this.isParagraphStart = isParagraphStart;
         calculateBBox();
     }
@@ -62,9 +56,6 @@ public class TextPositionSequence extends BoundingBox implements CharSequence {
 
     private void calculateBBox() {
 
-        this.bBoxDirAdj = textPositions.stream()
-                .map(RedTextPosition::getBBoxDirAdj)
-                .collect(RectangleTransformations.collectBBox());
         setToBBoxOfComponents(getTextPositions());
     }
 
@@ -73,10 +64,6 @@ public class TextPositionSequence extends BoundingBox implements CharSequence {
 
         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();
         calculateBBox();
     }
 
@@ -112,9 +99,6 @@ public class TextPositionSequence extends BoundingBox implements CharSequence {
         textPositionSequence.textPositions = textPositions.subList(start, end);
         textPositionSequence.page = page;
         textPositionSequence.dir = dir;
-        textPositionSequence.rotation = rotation;
-        textPositionSequence.pageHeight = pageHeight;
-        textPositionSequence.pageWidth = pageWidth;
         textPositionSequence.setToBBoxOfComponents(getTextPositions());
         return textPositionSequence;
     }
@@ -141,10 +125,6 @@ public class TextPositionSequence extends BoundingBox implements CharSequence {
 
         this.textPositions.add(textPosition);
         this.page = textPositionSequence.getPage();
-        this.dir = textPositionSequence.getDir();
-        this.rotation = textPositionSequence.getRotation();
-        this.pageHeight = textPositionSequence.getPageHeight();
-        this.pageWidth = textPositionSequence.getPageWidth();
         calculateBBox();
     }
 
@@ -152,79 +132,18 @@ public class TextPositionSequence extends BoundingBox implements CharSequence {
     public void add(TextPosition textPosition) {
 
         this.textPositions.add(RedTextPosition.fromTextPosition(textPosition));
-        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();
         calculateBBox();
     }
 
+    public double getTextHeightNoPadding() {
 
-    /**
-     * This value is adjusted so that 0,0 is upper left and it is adjusted based on the text direction.
-     * This method ignores the page rotation but takes the text rotation and adjusts the coordinates to awt.
-     *
-     * @return the text direction adjusted minX value
-     */
-
-    public float getMinXDirAdj() {
-
-        return textPositions.get(0).getXDirAdj();
-
+        return textPositions.get(0).getHeightDirAdj();
     }
 
 
-    /**
-     * This value is adjusted so that 0,0 is upper left and it is adjusted based on the text direction.
-     * This method ignores the page rotation but takes the text rotation and adjusts the coordinates to awt.
-     *
-     * @return the text direction adjusted maxX value
-     */
+    public double getTextHeight() {
 
-    public float getMaxXDirAdj() {
-
-        return textPositions.get(textPositions.size() - 1).getXDirAdj() + textPositions.get(textPositions.size() - 1).getWidthDirAdj() + HEIGHT_PADDING;
-
-    }
-
-
-    /**
-     * This value is adjusted so that 0,0 is upper left and it is adjusted based on the text direction.
-     * This method ignores the page rotation but takes the text rotation and adjusts the coordinates to awt.
-     *
-     * @return the text direction adjusted minY value. The upper border of the bounding box of the word.
-     */
-
-    public float getMinYDirAdj() {
-
-        return textPositions.get(0).getYDirAdj() - getTextHeight();
-
-    }
-
-
-    /**
-     * This value is adjusted so that 0,0 is upper left and it is adjusted based on the text direction.
-     * This method ignores the page rotation but takes the text rotation and adjusts the coordinates to awt.
-     *
-     * @return the text direction adjusted maxY value. The lower border of the bounding box of the word.
-     */
-
-    public float getMaxYDirAdj() {
-
-        return textPositions.get(0).getYDirAdj();
-
-    }
-
-
-    public float getTextHeightNoPadding() {
-
-        return textPositions.get(0).getHeightDir();
-    }
-
-
-    public float getTextHeight() {
-
-        return textPositions.get(0).getHeightDir() + HEIGHT_PADDING;
+        return textPositions.get(0).getHeightDirAdj() + HEIGHT_PADDING;
     }
 
 
@@ -240,18 +159,18 @@ public class TextPositionSequence extends BoundingBox implements CharSequence {
     public String getFontStyle() {
 
         if (textPositions.get(0).getFontName() == null) {
-            return "standard";
+            return STANDARD;
         }
         String lowercaseFontName = textPositions.get(0).getFontName().toLowerCase(Locale.ROOT);
 
-        if (lowercaseFontName.contains("bold") && lowercaseFontName.contains("italic")) {
-            return "bold, italic";
-        } else if (lowercaseFontName.contains("bold")) {
-            return "bold";
-        } else if (lowercaseFontName.contains("italic")) {
-            return "italic";
+        if (lowercaseFontName.contains(BOLD) && lowercaseFontName.contains(ITALIC)) {
+            return BOLD_ITALIC;
+        } else if (lowercaseFontName.contains(BOLD)) {
+            return BOLD;
+        } else if (lowercaseFontName.contains(ITALIC)) {
+            return ITALIC;
         } else {
-            return "standard";
+            return STANDARD;
         }
     }
 

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

@@ -32,7 +32,7 @@ public class ImageServiceResponseAdapter {
                     .add(new ClassifiedImage(new Rectangle2D.Double(imageMetadata.getPosition().getX1(),
                             imageMetadata.getPosition().getY1(),
                             imageMetadata.getGeometry().getWidth(),
-                            imageMetadata.getGeometry().getHeight()), classification, imageMetadata.isAlpha(), imageMetadata.getPosition().getPageNumber()));
+                            imageMetadata.getGeometry().getHeight()), classification, imageMetadata.isAlpha(), imageMetadata.getPosition().getPageNumber(),imageMetadata.getRepresentation()));
         });
 
         // Currently This is a copy but, it will be changed later because i don' t think that we should unclassified images.
@@ -44,7 +44,7 @@ public class ImageServiceResponseAdapter {
                     .add(new ClassifiedImage(new Rectangle2D.Double(imageMetadata.getPosition().getX1(),
                             imageMetadata.getPosition().getY1(),
                             imageMetadata.getGeometry().getWidth(),
-                            imageMetadata.getGeometry().getHeight()), classification, imageMetadata.isAlpha(), imageMetadata.getPosition().getPageNumber()));
+                            imageMetadata.getGeometry().getHeight()), classification, imageMetadata.isAlpha(), imageMetadata.getPosition().getPageNumber(),imageMetadata.getRepresentation()));
         });
 
         return images;
@@ -56,7 +56,7 @@ public class ImageServiceResponseAdapter {
         classificationPage.getImages().forEach(image -> {
             if (image.getImageType().equals(ImageType.OTHER)) {
                 for (AbstractPageBlock textblock : classificationPage.getTextBlocks()) {
-                    if (image.getPosition().contains(textblock.getBBoxInitialUserSpace())) {
+                    if (image.getPosition().contains(textblock.getBBoxPdf())) {
                         image.setImageType(ImageType.OCR);
                         return;
                     }

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

@@ -79,7 +79,7 @@ public class VisualLayoutParsingAdapter {
                 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);
+                        t.getBox().getY2() - t.getBox().getY1()), ImageType.SIGNATURE, true, false, false, pageNumber,"");
 
                 signatures.add(signature);
             }

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

@@ -14,6 +14,7 @@ import lombok.NoArgsConstructor;
 @AllArgsConstructor
 public class Classification {
 
+    @Builder.Default
     private Map<String, Float> probabilities = new HashMap<>();
     private String label;
 

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

@@ -12,6 +12,7 @@ import lombok.NoArgsConstructor;
 public class ImageMetadata {
 
     private Classification classification;
+    private String representation;
     private Position position;
     private Geometry geometry;
     private Filters filters;

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

@@ -22,8 +22,10 @@ public class ImageServiceResponse {
 
     @JsonProperty(value = "imageMetadata")
     @JsonAlias("data")
+    @Builder.Default
     private List<ImageMetadata> data = new ArrayList<>();
 
+    @Builder.Default
     private List<ImageMetadata> dataCV = new ArrayList<>();
 
 

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

@@ -15,6 +15,7 @@ import lombok.NoArgsConstructor;
 public class TableData {
 
     private PageInfo pageInfo;
+    @Builder.Default
     private List<TableCells> tableCells = new ArrayList<>();
 
 }

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

@@ -19,7 +19,7 @@ public class TableServiceResponse {
     private String operation;
     private String targetFileExtension;
     private String responseFileExtension;
-
+    @Builder.Default
     private List<TableData> data = new ArrayList<>();
 
 }

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

@@ -33,7 +33,7 @@ public class BodyTextFrameService {
         for (ClassificationPage page : classificationDocument.getPages()) {
             var updatedBodyTextFrame = getBodyTextFrameFromRulings(page, bodyTextFrame, landscapeBodyTextFrame);
             setBodyTextFrameAdjustedToPage(page, bodyTextFrame, landscapeBodyTextFrame);
-            classificationDocument.getVisualizations().addMainBodyVisualization(page.isLandscape() ? landscapeBodyTextFrame : bodyTextFrame, page.getPageNumber());
+            classificationDocument.getLayoutDebugLayer().addMainBodyVisualization(page.isLandscape() ? landscapeBodyTextFrame : bodyTextFrame, page.getPageNumber());
         }
     }
 

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

@@ -6,7 +6,6 @@ import java.util.List;
 
 import com.knecon.fforesight.service.layoutparser.processor.model.GapInformation;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
-import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
 
 import lombok.AllArgsConstructor;
 import lombok.experimental.UtilityClass;
@@ -38,7 +37,7 @@ public class GapDetectionService {
         for (TextPositionSequence currentTextPosition : sortedTextPositionSequences.subList(1, sortedTextPositionSequences.size())) {
 
             double yDifference = Math.abs(currentTextPosition.getMaxYDirAdj() - previousTextPosition.getMaxYDirAdj());
-            double xGap = Math.abs(previousTextPosition.getMaxXDirAdj() - currentTextPosition.getMinXDirAdj());
+            double xGap = Math.abs(previousTextPosition.getMaxXDirAdj() - currentTextPosition.getXDirAdj());
             Rectangle2D previousTextPositionBBox = toRectangle2D(previousTextPosition);
             Rectangle2D currentTextPositionBBox = toRectangle2D(currentTextPosition);
 

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

@@ -71,7 +71,7 @@ public class LineDetectionService {
 
     private static boolean isXGap(TextPositionSequence currentTextPosition, TextPositionSequence previousTextPosition, double avgTextPositionHeight) {
 
-        return Math.abs(previousTextPosition.getMaxXDirAdj() - currentTextPosition.getMinXDirAdj()) > (avgTextPositionHeight * X_GAP_FACTOR);
+        return Math.abs(previousTextPosition.getMaxXDirAdj() - currentTextPosition.getXDirAdj()) > (avgTextPositionHeight * X_GAP_FACTOR);
     }
 
 
@@ -83,7 +83,7 @@ public class LineDetectionService {
 
     private static boolean isNewLine(TextPositionSequence currentTextPosition, TextPositionSequence previousTextPosition, double avgTextPositionHeight) {
 
-        return Math.abs(previousTextPosition.getMinYDirAdj() - currentTextPosition.getMinYDirAdj()) > avgTextPositionHeight;
+        return Math.abs(previousTextPosition.getYDirAdj() - currentTextPosition.getYDirAdj()) > avgTextPositionHeight;
     }
 
 

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

@@ -27,8 +27,10 @@ import lombok.extern.slf4j.Slf4j;
 
 @Slf4j
 @Service
+@Deprecated
 public class SectionsBuilderService {
 
+
     public void buildSections(ClassificationDocument document) {
 
         List<AbstractPageBlock> chunkWords = new ArrayList<>();

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

@@ -9,7 +9,6 @@ import org.springframework.stereotype.Service;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.SimplifiedSectionText;
 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;
@@ -20,7 +19,7 @@ public class SimplifiedSectionTextService {
 
     public SimplifiedText toSimplifiedText(Document document) {
 
-        List<SimplifiedSectionText> simplifiedMainSectionsList = document.getMainSections()
+        List<SimplifiedSectionText> simplifiedMainSectionsList = document.getAllSections()
                 .stream()
                 .map(this::toSimplifiedSectionText)
                 .toList();
@@ -35,7 +34,22 @@ public class SimplifiedSectionTextService {
         List<SimplifiedSectionText> simplifiedText = Stream.of(simplifiedMainSectionsList, simplifiedHeadersList, simplifiedFootersList)
                 .flatMap(List::stream)
                 .collect(Collectors.toList());
-        return SimplifiedText.builder().numberOfPages(document.getNumberOfPages()).sectionTexts(simplifiedText).build();
+        return SimplifiedText.builder()
+                .numberOfPages(document.getNumberOfPages())
+                .sectionTexts(simplifiedText)
+                .mainSectionNumbers(document.getAllSections()
+                                            .stream()
+                                            .map(this::getSectionNumber)
+                                            .toList())
+                .headerSectionNumbers(document.getHeaders()
+                                              .stream()
+                                              .map(this::getSectionNumber)
+                                              .toList())
+                .footerSectionNumbers(document.getFooters()
+                                              .stream()
+                                              .map(this::getSectionNumber)
+                                              .toList())
+                .build();
     }
 
 
@@ -43,9 +57,20 @@ public class SimplifiedSectionTextService {
 
         return SimplifiedSectionText.builder()
                 .sectionNumber(section.getTreeId()
-                                       .get(0))
+                                       .stream()
+                                       .map(String::valueOf)
+                                       .collect(Collectors.joining(".")))
                 .text(section.getTextBlock().getSearchText())
                 .build();
     }
 
+
+    private String getSectionNumber(SemanticNode semanticNode) {
+
+        return semanticNode.getTreeId()
+                .stream()
+                .map(String::valueOf)
+                .collect(Collectors.joining("."));
+    }
+
 }

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

@@ -78,7 +78,7 @@ public class TableExtractionService {
 
             List<Cell> containedCells = new ArrayList<>();
             for (Cell c : cells) {
-                if (c.hasMinimumSize() && area.contains(c.getBBoxInitialUserSpace())) {
+                if (c.hasMinimumSize() && area.contains(c.getBBoxPdf())) {
                     containedCells.add(c);
                 }
             }

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

@@ -31,13 +31,13 @@ public class TextRulingsClassifier {
 
     private static void handleVerticalText(CleanRulings cleanRulings, TextPositionSequence word) {
 
-        float lowerY = (float) (word.getBBoxInitialUserSpace().getMinY() + TEXT_BBOX_THRESHOLD_FACTOR * word.getWidth());
-        float upperY = (float) (word.getBBoxInitialUserSpace().getMaxY() - TEXT_BBOX_THRESHOLD_FACTOR * word.getWidth());
+        float lowerY = (float) (word.getBBoxPdf().getMinY() + TEXT_BBOX_THRESHOLD_FACTOR * word.getWidth());
+        float upperY = (float) (word.getBBoxPdf().getMaxY() - TEXT_BBOX_THRESHOLD_FACTOR * word.getWidth());
 
-        float strikethroughCenterX = (float) word.getBBoxInitialUserSpace().getCenterX();
+        float strikethroughCenterX = (float) word.getBBoxPdf().getCenterX();
         float strikethroughBoxHeight = (float) ((word.getBBoxDirAdj().getHeight() * STRIKETHROUGH_ZONE) / 2);
 
-        float underlineCenterX = (float) (word.getDir().equals(TextDirection.QUARTER_CIRCLE) ? word.getBBoxInitialUserSpace().getMaxX() : word.getBBoxInitialUserSpace().getMinX());
+        float underlineCenterX = (float) (word.getDir().equals(TextDirection.QUARTER_CIRCLE) ? word.getBBoxPdf().getMaxX() : word.getBBoxPdf().getMinX());
         float underlineBoxHeight = (float) ((word.getBBoxDirAdj().getHeight() * UNDERLINE_ZONE) / 2);
 
         float leftX = Math.min(underlineCenterX - underlineBoxHeight, strikethroughCenterX - strikethroughBoxHeight);
@@ -65,13 +65,13 @@ public class TextRulingsClassifier {
 
     private static void handleHorizontalText(CleanRulings cleanRulings, TextPositionSequence word) {
 
-        float leftX = (float) (word.getBBoxInitialUserSpace().getMinX() + TEXT_BBOX_THRESHOLD_FACTOR * word.getWidth());
-        float rightX = (float) (word.getBBoxInitialUserSpace().getMaxX() - TEXT_BBOX_THRESHOLD_FACTOR * word.getWidth());
+        float leftX = (float) (word.getBBoxPdf().getMinX() + TEXT_BBOX_THRESHOLD_FACTOR * word.getWidth());
+        float rightX = (float) (word.getBBoxPdf().getMaxX() - TEXT_BBOX_THRESHOLD_FACTOR * word.getWidth());
 
-        float strikethroughCenterY = (float) word.getBBoxInitialUserSpace().getCenterY();
+        float strikethroughCenterY = (float) word.getBBoxPdf().getCenterY();
         float strikethroughBoxHeight = (float) ((word.getBBoxDirAdj().getHeight() * STRIKETHROUGH_ZONE) / 2);
 
-        float underlineCenterY = (float) (word.getDir().equals(TextDirection.ZERO) ? word.getBBoxInitialUserSpace().getMinY() : word.getBBoxInitialUserSpace().getMaxY());
+        float underlineCenterY = (float) (word.getDir().equals(TextDirection.ZERO) ? word.getBBoxPdf().getMinY() : word.getBBoxPdf().getMaxY());
         float underlineBoxHeight = (float) ((word.getBBoxDirAdj().getHeight() * UNDERLINE_ZONE) / 2);
 
         float lowerY = Math.min(underlineCenterY - underlineBoxHeight, strikethroughCenterY - strikethroughBoxHeight);

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

@@ -0,0 +1,527 @@
+package com.knecon.fforesight.service.layoutparser.processor.services.blockification;
+
+import static com.knecon.fforesight.service.layoutparser.processor.services.blockification.DocstrumBlockificationService.buildTextBlock;
+
+import java.util.ArrayList;
+import java.util.List;
+import java.util.ListIterator;
+import java.util.Locale;
+
+import org.apache.commons.lang3.StringUtils;
+import org.apache.commons.text.similarity.LevenshteinDistance;
+import org.springframework.stereotype.Service;
+
+import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
+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.SectionIdentifier;
+import com.knecon.fforesight.service.layoutparser.processor.model.outline.OutlineObject;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
+
+import lombok.Data;
+
+@Service
+public class BlockificationPostprocessingService {
+
+    private static final float STRING_SIMILARITY_THRESHOLD = 0.1f;
+
+
+    public OutlineObject sanitizeOutlineBlocks(ClassificationPage classificationPage, OutlineObject notFoundOutlineObject) {
+
+        List<OutlineObject> outlineObjects = classificationPage.getOutlineObjects();
+
+        if (getTextPageBlocks(classificationPage).isEmpty() || outlineObjects.isEmpty()) {
+            return null;
+        }
+
+        ListIterator<OutlineObject> outlineObjectListIterator = outlineObjects.listIterator();
+
+        if (notFoundOutlineObject != null) {
+            OutlineProcessionContext notFoundOutlineObjectProcessionContext = new OutlineProcessionContext(notFoundOutlineObject);
+            processTextBlocks(getTextPageBlocks(classificationPage), notFoundOutlineObjectProcessionContext);
+
+            OutlineObject firstOutlineObject = null;
+            OutlineProcessionContext firstOutlineObjectProcessionContext = null;
+            if (outlineObjectListIterator.hasNext()) {
+                firstOutlineObject = outlineObjectListIterator.next();
+                firstOutlineObjectProcessionContext = new OutlineProcessionContext(firstOutlineObject);
+                processTextBlocks(getTextPageBlocks(classificationPage), firstOutlineObjectProcessionContext);
+            }
+
+            if (!contextsOverlap(notFoundOutlineObjectProcessionContext, firstOutlineObjectProcessionContext)) {
+                notFoundOutlineObject.setFound(selectMatch(classificationPage, notFoundOutlineObjectProcessionContext));
+            }
+            if (firstOutlineObject != null) {
+                // re-create the context for the updated blocks
+                firstOutlineObjectProcessionContext = new OutlineProcessionContext(firstOutlineObject);
+                processTextBlocks(getTextPageBlocks(classificationPage), firstOutlineObjectProcessionContext);
+                firstOutlineObject.setFound(selectMatch(classificationPage, firstOutlineObjectProcessionContext));
+            }
+
+        }
+
+        outlineObjectListIterator.forEachRemaining(outlineObject -> {
+            OutlineProcessionContext outlineObjectProcessionContext = new OutlineProcessionContext(outlineObject);
+            processTextBlocks(getTextPageBlocks(classificationPage), outlineObjectProcessionContext);
+            outlineObject.setFound(selectMatch(classificationPage, outlineObjectProcessionContext));
+        });
+
+        if (!outlineObjects.isEmpty()) {
+            return outlineObjects.get(outlineObjects.size() - 1);
+        } else {
+            return notFoundOutlineObject;
+        }
+    }
+
+
+    private static List<TextPageBlock> getTextPageBlocks(ClassificationPage classificationPage) {
+
+        return classificationPage.getTextBlocks()
+                .stream()
+                .filter(block -> block instanceof TextPageBlock)
+                .map(block -> (TextPageBlock) block)
+                .toList();
+    }
+
+
+    private boolean contextsOverlap(OutlineProcessionContext notFoundOutlineObjectProcessionContext, OutlineProcessionContext firstOutlineObjectProcessionContext) {
+
+        if (firstOutlineObjectProcessionContext == null) {
+            return false;
+        }
+
+        String notFoundTitle = notFoundOutlineObjectProcessionContext.getOutlineObject().getTitle();
+        String firstTitle = firstOutlineObjectProcessionContext.getOutlineObject().getTitle();
+
+        if (!firstTitle.startsWith(notFoundTitle)) {
+            return false;
+        }
+
+        var blocksOfNotFoundOutline = getAllMatchingBlocks(notFoundOutlineObjectProcessionContext);
+        var blocksOfFirstOutline = getAllMatchingBlocks(firstOutlineObjectProcessionContext);
+
+        double maxYFirst = blocksOfFirstOutline.stream()
+                .mapToDouble(TextPageBlock::getPdfMaxY)
+                .max().orElse(Double.NEGATIVE_INFINITY);
+
+        return blocksOfNotFoundOutline.stream()
+                .mapToDouble(TextPageBlock::getPdfMaxY)
+                .anyMatch(y -> y >= maxYFirst);
+    }
+
+
+    private List<TextPageBlock> getAllMatchingBlocks(OutlineProcessionContext context) {
+
+        List<TextPageBlock> blocks = new ArrayList<>();
+        if (context.getDirectMatch() != null) {
+            blocks.add(context.getDirectMatch());
+        }
+        if (context.getSplitCandidate() != null) {
+            blocks.add(context.getSplitCandidate());
+        }
+        blocks.addAll(context.getMergeCandidates());
+        return blocks;
+    }
+
+
+    private void processTextBlocks(List<TextPageBlock> textBlocks, OutlineProcessionContext context) {
+
+        OutlineObject outlineObject = context.getOutlineObject();
+        ListIterator<TextPageBlock> iterator = textBlocks.listIterator();
+        while (iterator.hasNext()) {
+            TextPageBlock pageBlock = iterator.next();
+            if (outlineObject.isAbove(pageBlock)) {
+                break;
+            }
+        }
+        if (iterator.hasPrevious()) {
+            iterator.previous();
+        }
+        boolean earlyStop = false;
+        while (iterator.hasNext() && !earlyStop) {
+            TextPageBlock pageBlock = iterator.next();
+            earlyStop = processOutlineForTextBlock(pageBlock, context);
+        }
+    }
+
+
+    private boolean selectMatch(ClassificationPage classificationPage, OutlineProcessionContext context) {
+
+        OutlineObject outlineObject = context.outlineObject;
+        TextPageBlock directMatch = context.directMatch;
+        List<TextPageBlock> mergeCandidates = context.mergeCandidates;
+        TextPageBlock splitCandidate = context.splitCandidate;
+        PageBlockType headlineType = PageBlockType.getHeadlineType(outlineObject.getTreeDepth());
+
+        double distanceToDirectMatch = directMatch != null ? calculateDistance(outlineObject, directMatch) : Double.MAX_VALUE;
+        double distanceToSplitCandidate = splitCandidate != null ? calculateDistance(outlineObject, splitCandidate) : Double.MAX_VALUE;
+
+        double distanceToBestMergeCandidates = Double.MAX_VALUE;
+        List<TextPageBlock> bestMergeCandidateCombination = new ArrayList<>();
+        if (!mergeCandidates.isEmpty()) {
+
+            // with this code adjacent blocks to the first and last merge candidate get added, this could be useful for some edge cases:
+            //List<TextPageBlock> allMergeCandidates = new ArrayList<>(mergeCandidates);
+            //addNeighborsOfCandidate(kdTree, mergeCandidates.get(0), allMergeCandidates);
+            //if (mergeCandidates.size() > 1) {
+            //    addNeighborsOfCandidate(kdTree, mergeCandidates.get(mergeCandidates.size() - 1), allMergeCandidates);
+            //}
+            //allMergeCandidates = allMergeCandidates.stream()
+            //        .distinct()
+            //        .toList();
+
+            List<List<TextPageBlock>> combinations = findCombinations(outlineObject.getTitle(), mergeCandidates);
+
+            for (List<TextPageBlock> combination : combinations) {
+                double averageDistance = combination.stream()
+                        .map(block -> calculateDistance(outlineObject, block))
+                        .mapToDouble(Double::doubleValue).average().orElse(Double.MAX_VALUE);
+                if (distanceToBestMergeCandidates > averageDistance) {
+                    distanceToBestMergeCandidates = averageDistance;
+                    bestMergeCandidateCombination = combination;
+                }
+            }
+        }
+
+        double minDistance = Math.min(distanceToDirectMatch, Math.min(distanceToSplitCandidate, distanceToBestMergeCandidates));
+
+        if (minDistance == Double.MAX_VALUE) {
+            return false;
+        }
+        if (minDistance == distanceToDirectMatch) {
+            setClassificationAndAddOutlineEngine(directMatch, headlineType);
+        } else if (minDistance == distanceToSplitCandidate) {
+            SplitBlockResult splitBlockResult = splitBlock(classificationPage, splitCandidate, context.sectionIdentifier, outlineObject.getTitle());
+            if (splitBlockResult.modifiedBlockToSplit) {
+                setClassificationAndAddOutlineEngine(splitCandidate, headlineType);
+            }
+            splitBlockResult.otherBlocks.forEach(other -> other.setClassification(null));
+        } else {
+            var merged = mergeBlocks(classificationPage, bestMergeCandidateCombination);
+            setClassificationAndAddOutlineEngine(merged, headlineType);
+        }
+        return true;
+    }
+
+
+    private static void setClassificationAndAddOutlineEngine(TextPageBlock block, PageBlockType headlineType) {
+
+        block.setClassification(headlineType);
+        block.getEngines().add(LayoutEngine.OUTLINE);
+    }
+
+
+    private SplitBlockResult splitBlock(ClassificationPage classificationPage, TextPageBlock blockToSplit, SectionIdentifier sectionIdentifier, String title) {
+
+        List<TextPageBlock> otherBlocks = new ArrayList<>();
+        int blockToSplitIdx = classificationPage.getTextBlocks().indexOf(blockToSplit);
+
+        String headline = title;
+        if (!sectionIdentifier.getFormat().equals(SectionIdentifier.Format.EMPTY) && !title.startsWith(sectionIdentifier.getIdentifierString())) {
+            headline = sectionIdentifier + headline;
+        }
+
+        WordSequenceResult wordSequenceResult = findWordSequence(blockToSplit.getSequences(), headline);
+        if (wordSequenceResult.inSequence.isEmpty() && !headline.equals(title)) {
+            wordSequenceResult = findWordSequence(blockToSplit.getSequences(), title);
+        }
+
+        boolean modifiedBlockToSplit = false;
+        if (!wordSequenceResult.inSequence.isEmpty()) {
+            blockToSplit.setSequences(wordSequenceResult.inSequence);
+            blockToSplit.recalculateBBox();
+            modifiedBlockToSplit = true;
+        }
+
+        if (!wordSequenceResult.preSequence.isEmpty()) {
+            TextPageBlock block = buildTextBlock(wordSequenceResult.preSequence, 0);
+            classificationPage.getTextBlocks().add(blockToSplitIdx, block);
+            otherBlocks.add(block);
+            blockToSplitIdx++;
+        }
+        if (!wordSequenceResult.postSequence.isEmpty()) {
+            TextPageBlock block = buildTextBlock(wordSequenceResult.postSequence, 0);
+            classificationPage.getTextBlocks().add(blockToSplitIdx + 1, block);
+            otherBlocks.add(block);
+        }
+
+        return new SplitBlockResult(modifiedBlockToSplit, otherBlocks);
+    }
+
+
+    private static WordSequenceResult findWordSequence(List<TextPositionSequence> textPositionSequences, String text) {
+
+        String target = sanitizeString(text);
+        List<TextPositionSequence> inSequence = new ArrayList<>();
+        List<TextPositionSequence> preSequence = new ArrayList<>();
+        List<TextPositionSequence> postSequence = new ArrayList<>();
+        StringBuilder currentSequence = new StringBuilder();
+
+        if (target.isBlank()) {
+            return new WordSequenceResult();
+        }
+
+        for (TextPositionSequence sequence : textPositionSequences) {
+
+            currentSequence.append(sanitizeString(sequence.toString()));
+            inSequence.add(sequence);
+
+            if (currentSequence.length() >= target.length()) {
+
+                if (currentSequence.toString().endsWith(target)) {
+
+                    int index = 0;
+                    String toRemove = currentSequence.substring(0, currentSequence.length() - target.length());
+
+                    TextPositionSequence next = inSequence.get(index);
+                    while (currentSequence.length() - next.length() >= target.length()) {
+
+                        TextPositionSequence removed = inSequence.remove(index);
+                        currentSequence.delete(0, removed.toString().length());
+                        preSequence.add(removed);
+
+                        next = inSequence.get(index);
+                        toRemove = toRemove.substring(removed.length());
+                    }
+
+                    if (!toRemove.isEmpty()) {
+                        SplitSequenceResult splitSequenceResult = splitSequence(inSequence.remove(index), toRemove);
+
+                        currentSequence.delete(0, splitSequenceResult.out.length());
+                        preSequence.add(splitSequenceResult.out);
+                        inSequence.add(index, splitSequenceResult.in);
+                    }
+
+                } else if (currentSequence.toString().startsWith(target)) {
+
+                    int index = inSequence.size() - 1;
+                    String toRemove = currentSequence.substring(target.length());
+
+                    SplitSequenceResult splitSequenceResult = splitSequence(inSequence.remove(index), toRemove);
+                    currentSequence.delete(currentSequence.length() - splitSequenceResult.out.length(), currentSequence.length());
+
+                    inSequence.add(index, splitSequenceResult.in);
+                    postSequence.add(splitSequenceResult.out);
+                }
+
+                if (currentSequence.toString().equals(target)) {
+                    postSequence.addAll(textPositionSequences.subList(textPositionSequences.indexOf(sequence) + 1, textPositionSequences.size()));
+                    return new WordSequenceResult(inSequence, preSequence, postSequence);
+                }
+            }
+        }
+
+        return new WordSequenceResult();
+    }
+
+
+    private static SplitSequenceResult splitSequence(TextPositionSequence sequence, String toRemove) {
+
+        TextPositionSequence in = null;
+        TextPositionSequence out;
+
+        String currentSequence = sequence.toString().toLowerCase(Locale.ROOT);
+        int index = currentSequence.indexOf(toRemove);
+        int endIndex = index + toRemove.length();
+
+        out = createSubSequence(sequence, index, endIndex);
+
+        if (index > 0) {
+            in = createSubSequence(sequence, 0, index);
+        } else if (endIndex < sequence.getTextPositions().size()) {
+            in = createSubSequence(sequence, endIndex, sequence.getTextPositions().size());
+        }
+
+        return new SplitSequenceResult(in, out);
+    }
+
+
+    private static TextPositionSequence createSubSequence(TextPositionSequence sequence, int start, int end) {
+
+        TextPositionSequence newSeq = new TextPositionSequence(new ArrayList<>(sequence.getTextPositions().subList(start, end)), sequence.getPage());
+        newSeq.setParagraphStart(sequence.isParagraphStart());
+        return newSeq;
+    }
+
+
+    private TextPageBlock mergeBlocks(ClassificationPage classificationPage, List<TextPageBlock> blocksToMerge) {
+
+        TextPageBlock firstBlock = blocksToMerge.get(0);
+
+        if (blocksToMerge.size() > 1) {
+
+            List<TextPageBlock> mergedBlocks = new ArrayList<>();
+            for (TextPageBlock textPageBlock : blocksToMerge.subList(1, blocksToMerge.size())) {
+
+                if (firstBlock != null && !firstBlock.getSequences().isEmpty()) {
+
+                    if (textPageBlock.getDir() == firstBlock.getDir()) {
+                        firstBlock.getSequences().addAll(textPageBlock.getSequences());
+                        mergedBlocks.add(textPageBlock);
+                    }
+                }
+            }
+
+            assert firstBlock != null;
+            firstBlock.setToDuplicate(false);
+            firstBlock.recalculateBBox();
+            classificationPage.getTextBlocks().removeAll(mergedBlocks);
+        }
+
+        return firstBlock;
+    }
+
+
+    private static List<List<TextPageBlock>> findCombinations(String title, List<TextPageBlock> blocks) {
+
+        List<List<TextPageBlock>> combinations = new ArrayList<>();
+        findCombinations(title, blocks, new ArrayList<>(), combinations);
+        return combinations;
+    }
+
+
+    private static void findCombinations(String title, List<TextPageBlock> blocks, List<TextPageBlock> current, List<List<TextPageBlock>> combinations) {
+
+        String target = sanitizeString(title);
+        if (target.isEmpty()) {
+            combinations.add(new ArrayList<>(current));
+            return;
+        }
+
+        List<TextPageBlock> remaining = blocks.stream()
+                .filter(block -> !current.contains(block))
+                .toList();
+        for (TextPageBlock block : remaining) {
+            String prefix = sanitizeString(block.getText());
+            if (target.startsWith(prefix)) {
+                current.add(block);
+                findCombinations(target.substring(prefix.length()), blocks.subList(blocks.indexOf(block) + 1, blocks.size()), current, combinations);
+                current.remove(current.size() - 1);
+            }
+        }
+    }
+
+
+    private double calculateDistance(OutlineObject outlineObject, TextPageBlock pageBlock) {
+
+        return outlineObject.distance(pageBlock);
+    }
+
+
+    // currently only three cases are handled here:
+    // 1. equality
+    // 2. outline title contains block text
+    // 3. block text contains outline title
+    // another possible case is an intersection, meaning a title is split up between two different blocks
+    // this should not happen with how docstrum creates the blocks
+    // if it is indeed necessary, a splitting has to be done with a follow-up merge
+    private boolean processOutlineForTextBlock(TextPageBlock pageBlock, OutlineProcessionContext context) {
+
+        OutlineObject outlineObject = context.getOutlineObject();
+        String blockText = sanitizeString(pageBlock.getText());
+        String outlineTitle = sanitizeString(outlineObject.getTitle());
+
+        int threshold = (int) (Math.min(blockText.length(), outlineTitle.length()) * STRING_SIMILARITY_THRESHOLD) + 1;
+        int distance = new LevenshteinDistance(threshold).apply(blockText, outlineTitle);
+        if (distance >= 0 && distance < threshold) {
+            context.directMatch = pageBlock;
+            return true;
+        }
+
+        boolean blockTextContainsOutlineTitle = blockText.contains(outlineTitle);
+        boolean outlineTitleContainsBlockText = outlineTitle.contains(blockText);
+
+        if (!blockTextContainsOutlineTitle && !outlineTitleContainsBlockText) {
+            return false;
+        }
+
+        if (blockText.equals(outlineTitle) && context.directMatch == null) {
+            context.directMatch = pageBlock;
+            return true;
+        }
+
+        if (outlineTitleContainsBlockText) {
+            context.mergeCandidates.add(pageBlock);
+        }
+
+        if (blockTextContainsOutlineTitle) {
+            SectionIdentifier sectionIdentifier = SectionIdentifier.fromSearchText(blockText);
+
+            if (sectionIdentifier.getFormat() != SectionIdentifier.Format.EMPTY && !outlineTitle.startsWith(sectionIdentifier.getIdentifierString())) {
+
+                if (blockText.startsWith(sectionIdentifier.getIdentifierString()) && blockText.endsWith(outlineTitle) && context.directMatch == null) {
+                    context.directMatch = pageBlock;
+                    return true;
+                } else if (context.splitCandidate == null) {
+                    context.sectionIdentifier = sectionIdentifier;
+                }
+            }
+            if (context.splitCandidate == null) {
+                context.splitCandidate = pageBlock;
+            }
+        }
+        return false;
+    }
+
+
+    private static String sanitizeString(String text) {
+
+        return StringUtils.deleteWhitespace(text).toLowerCase(Locale.ROOT);
+    }
+
+
+    @Data
+    private static class OutlineProcessionContext {
+
+        private TextPageBlock directMatch;
+        private OutlineObject outlineObject;
+        private List<TextPageBlock> mergeCandidates;
+        private TextPageBlock splitCandidate;
+        private SectionIdentifier sectionIdentifier;
+
+
+        OutlineProcessionContext(OutlineObject outlineObject) {
+
+            this.outlineObject = outlineObject;
+            this.directMatch = null;
+            this.mergeCandidates = new ArrayList<>();
+            this.splitCandidate = null;
+            this.sectionIdentifier = SectionIdentifier.empty();
+        }
+
+    }
+
+    public static class WordSequenceResult {
+
+        public List<TextPositionSequence> inSequence;
+        public List<TextPositionSequence> preSequence;
+        public List<TextPositionSequence> postSequence;
+
+
+        public WordSequenceResult(List<TextPositionSequence> inSequence, List<TextPositionSequence> preSequence, List<TextPositionSequence> postSequence) {
+
+            this.inSequence = inSequence;
+            this.preSequence = preSequence;
+            this.postSequence = postSequence;
+        }
+
+
+        public WordSequenceResult() {
+
+            this.inSequence = new ArrayList<>();
+            this.preSequence = new ArrayList<>();
+            this.postSequence = new ArrayList<>();
+        }
+
+    }
+
+    public record SplitBlockResult(boolean modifiedBlockToSplit, List<TextPageBlock> otherBlocks) {
+
+    }
+
+    public record SplitSequenceResult(TextPositionSequence in, TextPositionSequence out) {
+
+    }
+
+}

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

@@ -1,7 +1,6 @@
 package com.knecon.fforesight.service.layoutparser.processor.services.blockification;
 
 import java.util.ArrayList;
-import java.util.Comparator;
 import java.util.List;
 import java.util.ListIterator;
 
@@ -10,14 +9,13 @@ import org.springframework.stereotype.Service;
 import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingType;
 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.table.CleanRulings;
 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.text.TextPositionSequence;
-import com.knecon.fforesight.service.layoutparser.processor.visualization.LayoutparsingVisualizations;
+import com.knecon.fforesight.service.layoutparser.processor.visualization.LayoutDebugLayer;
 
 import lombok.RequiredArgsConstructor;
 
@@ -26,6 +24,7 @@ import lombok.RequiredArgsConstructor;
 @RequiredArgsConstructor
 public class DocstrumBlockificationService {
 
+    public static final float Y_THRESHOLD = 5f;
     private final DocstrumSegmentationService docstrumSegmentationService;
 
     static final float THRESHOLD = 1f;
@@ -34,12 +33,12 @@ public class DocstrumBlockificationService {
     public ClassificationPage blockify(List<TextPositionSequence> textPositions,
                                        CleanRulings rulings,
                                        boolean xyOrder,
-                                       LayoutparsingVisualizations visualizations,
+                                       LayoutDebugLayer visualizations,
                                        LayoutParsingType layoutParsingType) {
 
         CleanRulings usedRulings = rulings.withoutTextRulings();
 
-        var zones = docstrumSegmentationService.segmentPage(textPositions, xyOrder, usedRulings, visualizations);
+        List<Zone> zones = docstrumSegmentationService.segmentPage(textPositions, xyOrder, usedRulings, visualizations);
 
         if (!textPositions.isEmpty()) {
             visualizations.addZoneVisualizations(zones, textPositions.get(0).getPage());
@@ -47,44 +46,28 @@ public class DocstrumBlockificationService {
             visualizations.addCharactersWithNeighbours(zones, textPositions.get(0).getPage());
         }
 
-        var pageBlocks = toAbstractPageBlocks(zones, xyOrder, usedRulings);
-
-        if (xyOrder) {
-            sortPageBlocksXThenY(pageBlocks);
-        }
+        var pageBlocks = toAbstractPageBlocks(zones);
 
         var classificationPage = new ClassificationPage(pageBlocks);
         classificationPage.setCleanRulings(rulings);
 
         mergeIntersectingBlocks(classificationPage, usedRulings, 0, 0);
 
-        if (layoutParsingType == LayoutParsingType.DOCUMINE || layoutParsingType == LayoutParsingType.REDACT_MANAGER) {
-            combineBlocks(classificationPage);
+        if (layoutParsingType == LayoutParsingType.DOCUMINE
+            || layoutParsingType == LayoutParsingType.REDACT_MANAGER
+            || layoutParsingType == LayoutParsingType.REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH) {
+            combineBlocks(classificationPage, layoutParsingType);
         }
 
         if (layoutParsingType == LayoutParsingType.CLARIFYND) {
-            mergeIntersectingBlocks(classificationPage, usedRulings, 0, 6.5f);
+            mergeIntersectingBlocks(classificationPage, usedRulings, 0, 0);
         }
 
         return classificationPage;
     }
 
 
-    private static void sortPageBlocksXThenY(List<AbstractPageBlock> pageBlocks) {
-
-        pageBlocks.sort(Comparator.comparing(AbstractPageBlock::getMinY, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD))
-                                .thenComparing(AbstractPageBlock::getMinX, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD)));
-        pageBlocks.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;
-            }
-        });
-    }
-
-
-    private List<AbstractPageBlock> toAbstractPageBlocks(List<Zone> zones, boolean xyOrder, CleanRulings usedRulings) {
+    private List<AbstractPageBlock> toAbstractPageBlocks(List<Zone> zones) {
 
         List<AbstractPageBlock> abstractPageBlocks = new ArrayList<>();
         zones.forEach(zone -> {
@@ -105,7 +88,7 @@ public class DocstrumBlockificationService {
     }
 
 
-    public void combineBlocks(ClassificationPage page) {
+    public void combineBlocks(ClassificationPage page, LayoutParsingType layoutParsingType) {
 
         TextPageBlock previous = new TextPageBlock();
         ListIterator<AbstractPageBlock> itty = page.getTextBlocks().listIterator();
@@ -126,8 +109,19 @@ public class DocstrumBlockificationService {
                     continue;
                 }
 
+                if (current.isHeadline() || previous.isHeadline()) {
+                    if (intersectsYWithPreviousHavingMaxOneLine(previous, current, page)) {
+                        previous = combineBlocksAndResetIterator(previous, current, itty, false);
+                    } else {
+                        previous = current;
+                    }
+
+                    continue;
+                }
+
                 if (areTheOnlyTwoBlocksOnHeightsWithBothMoreThanTwoLines(previous, current, page)) {
-                    previous = combineBlocksAndResetIterator(previous, current, itty, true);
+//                    previous = combineBlocksAndResetIterator(previous, current, itty, true);
+                    previous = combineBlocksAndResetIterator(previous, current, itty, layoutParsingType != LayoutParsingType.REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH);
                     continue;
                 }
 
@@ -150,7 +144,7 @@ public class DocstrumBlockificationService {
             previous = current;
         }
 
-        mergeIntersectingBlocks(page, usedRulings, 0, 6.5f);
+        mergeIntersectingBlocks(page, usedRulings, 0, Y_THRESHOLD);
     }
 
 
@@ -172,6 +166,12 @@ public class DocstrumBlockificationService {
     }
 
 
+    private boolean intersectsYWithPreviousHavingMaxOneLine(TextPageBlock previous, TextPageBlock current, ClassificationPage page) {
+
+        return previous.intersectsY(current) && (previous.getNumberOfLines() == 1 && current.getNumberOfLines() >= 1);
+    }
+
+
     private boolean areTheOnlyTwoBlocksOnHeightsWithBothMoreThanTwoLines(TextPageBlock previous, TextPageBlock current, ClassificationPage page) {
 
         return previous.getNumberOfLines() >= 2 && current.getNumberOfLines() >= 2 //
@@ -185,6 +185,9 @@ public class DocstrumBlockificationService {
         previous.getSequences().addAll(current.getSequences());
         previous = buildTextBlock(previous.getSequences(), 0);
         previous.setToDuplicate(toDuplicate);
+        if (current.getClassification() != null && previous.getClassification() == null) {
+            previous.setClassification(current.getClassification());
+        }
         itty.remove();
         itty.previous();
         itty.set(previous);
@@ -244,26 +247,39 @@ public class DocstrumBlockificationService {
                 continue;
             }
 
+            if (block.getClassification() != null && block.getClassification().isHeadline()) {
+                continue;
+            }
+
             TextPageBlock current = (TextPageBlock) block;
 
             for (int i = 0; i < blocks.size(); i++) {
 
-                if (blocks.get(i) == null) {
+                AbstractPageBlock abstractPageBlock = blocks.get(i);
+                if (abstractPageBlock == null) {
                     continue;
                 }
-                if (blocks.get(i) == current) {
+                if (abstractPageBlock == current) {
                     continue;
                 }
-                if (blocks.get(i) instanceof TablePageBlock) {
+                if (abstractPageBlock instanceof TablePageBlock) {
                     continue;
                 }
 
-                TextPageBlock inner = (TextPageBlock) blocks.get(i);
+                if (abstractPageBlock.getClassification() != null && abstractPageBlock.getClassification().isHeadline()) {
+                    continue;
+                }
+
+                TextPageBlock inner = (TextPageBlock) abstractPageBlock;
 
                 if (usedRulings.lineBetween(current, blocks.get(i))) {
                     continue;
                 }
 
+//                if (!current.getMostPopularWordStyle().equals(inner.getMostPopularWordStyle())) {
+//                    continue;
+//                }
+
                 if (current.getDir() == inner.getDir() && current.intersects(inner, yThreshold, xThreshold)) {
 
                     boolean toDuplicate = current.isToDuplicate() || inner.isToDuplicate();
@@ -285,7 +301,7 @@ public class DocstrumBlockificationService {
     }
 
 
-    private TextPageBlock buildTextBlock(List<TextPositionSequence> wordBlockList, int indexOnPage) {
+    public static TextPageBlock buildTextBlock(List<TextPositionSequence> wordBlockList, int indexOnPage) {
 
         return new TextPageBlock(wordBlockList);
     }

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

@@ -1,25 +1,24 @@
 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 java.util.regex.Matcher;
 import java.util.regex.Pattern;
 import java.util.stream.Collectors;
 
 import org.springframework.stereotype.Service;
 
+import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
 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.Orientation;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRulings;
-import com.knecon.fforesight.service.layoutparser.processor.model.text.StringFrequencyCounter;
+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.text.TextPositionSequence;
 
+@SuppressWarnings("all")
 @Service
 public class DocuMineBlockificationService {
 
@@ -44,31 +43,33 @@ public class DocuMineBlockificationService {
 
         CleanRulings usedRulings = cleanRulings.withoutTextRulings();
 
-        float minX = 1000;
-        float maxX = 0;
-        float minY = 1000;
-        float maxY = 0;
+        double minX = 1000;
+        double maxX = 0;
+        double minY = 1000;
+        double maxY = 0;
         TextPositionSequence prev = null;
 
         boolean wasSplitted = false;
-        Float splitX1 = null;
+        Double splitX1 = null;
         for (TextPositionSequence word : textPositions) {
 
-            boolean lineSeparation = prev != null && word.getMinYDirAdj() - maxY > Math.min(word.getHeight(), prev.getHeight()) * 1.25;
-            boolean startFromTop = prev != null && word.getMinYDirAdj() < prev.getMinYDirAdj() - prev.getTextHeight();
-            boolean splitByX = prev != null && maxX + 50 < word.getMinXDirAdj() && prev.getMinYDirAdj() == word.getMinYDirAdj();
-            boolean negativeXGap = prev != null && word.getMinXDirAdj() - minX < -5;
-            boolean newLineAfterSplit = prev != null && word.getMinYDirAdj() != prev.getMinYDirAdj() && wasSplitted && splitX1 != word.getMinXDirAdj();
+            boolean lineSeparation = prev != null && word.getYDirAdj() - prev.getMaxYDirAdj() > Math.min(word.getHeight(), prev.getHeight()) * 1.1;
+            boolean startFromTop = prev != null && word.getYDirAdj() < prev.getYDirAdj() - prev.getTextHeight();
+            boolean splitByX = prev != null && maxX + 50 < word.getXDirAdj() && prev.getYDirAdj() == word.getYDirAdj();
+            boolean negativeXGap = prev != null && word.getXDirAdj() - minX < -5;
+            boolean newLineAfterSplit = prev != null && word.getYDirAdj() != prev.getYDirAdj() && wasSplitted && splitX1 != word.getXDirAdj();
             boolean isSplitByRuling = prev != null && usedRulings.lineBetween(prev, word);
             boolean splitByDir = prev != null && !prev.getDir().equals(word.getDir());
-            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"));
+            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")
+                                                    || Math.abs(prev.getFontSize() - word.getFontSize()) >= 1
+                                                    || Math.abs(word.getTextHeight() - prev.getTextHeight()) > 0.8);
 
             Matcher matcher = pattern.matcher(chunkWords.stream()
                                                       .collect(Collectors.joining(" ")).toString());
-            boolean startsOnSameX = Math.abs(minX - word.getMinXDirAdj()) < 5 && matcher.matches();
+            boolean startsOnSameX = Math.abs(minX - word.getXDirAdj()) < 5 && matcher.matches();
 
             if (prev != null && (lineSeparation || startFromTop || splitByDir || isSplitByRuling || splitByOtherFontAndOtherY || negativeXGap || startsOnSameX)) {
 
@@ -84,7 +85,7 @@ public class DocuMineBlockificationService {
                 if (splitByX && !isSplitByRuling) {
                     wasSplitted = true;
                     cb1.setOrientation(Orientation.LEFT);
-                    splitX1 = word.getMinXDirAdj();
+                    splitX1 = word.getXDirAdj();
                 } else if (newLineAfterSplit && !isSplitByRuling) {
                     wasSplitted = false;
                     cb1.setOrientation(Orientation.RIGHT);
@@ -107,14 +108,14 @@ public class DocuMineBlockificationService {
             chunkWords.add(word);
 
             prev = word;
-            if (word.getMinXDirAdj() < minX) {
-                minX = word.getMinXDirAdj();
+            if (word.getXDirAdj() < minX) {
+                minX = word.getXDirAdj();
             }
             if (word.getMaxXDirAdj() > maxX) {
                 maxX = word.getMaxXDirAdj();
             }
-            if (word.getMinYDirAdj() < minY) {
-                minY = word.getMinYDirAdj();
+            if (word.getYDirAdj() < minY) {
+                minY = word.getYDirAdj();
             }
             if (word.getMaxYDirAdj() > maxY) {
                 maxY = word.getMaxYDirAdj();
@@ -127,6 +128,76 @@ public class DocuMineBlockificationService {
     }
 
 
+    public void mergeblocks(ClassificationPage page, CleanRulings usedRulings, float xThreshold, float yThreshold) {
+
+        var blocks = page.getTextBlocks();
+        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++) {
+
+                AbstractPageBlock abstractPageBlock = blocks.get(i);
+                if (abstractPageBlock == null) {
+                    continue;
+                }
+                if (abstractPageBlock == current) {
+                    continue;
+                }
+                if (abstractPageBlock instanceof TablePageBlock) {
+                    continue;
+                }
+
+                if (isHeadlineFromOutline(current) || isHeadlineFromOutline(abstractPageBlock)) {
+                    continue;
+                }
+
+                TextPageBlock inner = (TextPageBlock) abstractPageBlock;
+
+                if (usedRulings.lineBetween(current, blocks.get(i))) {
+                    continue;
+                }
+
+                if (current.getDir() == inner.getDir() && current.intersects(inner, yThreshold, xThreshold) && (current.getClassification() == null || current.getClassification()
+                        .equals(inner.getClassification()))) {
+
+                    boolean toDuplicate = current.isToDuplicate() || inner.isToDuplicate();
+                    current.getSequences().addAll(inner.getSequences());
+                    current = buildTextBlock(current.getSequences(), 0);
+                    current.setClassification(inner.getClassification());
+                    current.setToDuplicate(toDuplicate);
+                    blocks.set(i, null);
+                    itty.set(current);
+                }
+            }
+        }
+        var blocksIterator = blocks.iterator();
+        while (blocksIterator.hasNext()) {
+            if (blocksIterator.next() == null) {
+                blocksIterator.remove();
+            }
+        }
+    }
+
+
+    private boolean isHeadlineFromOutline(AbstractPageBlock abstractPageBlock) {
+
+        return abstractPageBlock.getEngines().contains(LayoutEngine.OUTLINE) && abstractPageBlock.getClassification() != null && abstractPageBlock.getClassification().isHeadline();
+    }
+
+
+    public static TextPageBlock buildTextBlock(List<TextPositionSequence> wordBlockList, int indexOnPage) {
+
+        return new TextPageBlock(wordBlockList);
+    }
 
 }
 

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

@@ -1,9 +1,6 @@
 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.Iterator;
 import java.util.List;
 
@@ -11,13 +8,11 @@ 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.FloatFrequencyCounter;
 import com.knecon.fforesight.service.layoutparser.processor.model.Orientation;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRulings;
-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.visualization.LayoutparsingVisualizations;
+import com.knecon.fforesight.service.layoutparser.processor.visualization.LayoutDebugLayer;
 
 @SuppressWarnings("all")
 @Service
@@ -35,7 +30,7 @@ public class RedactManagerBlockificationService {
      * @param visualizations
      * @return Page object that contains the Textblock and text statistics.
      */
-    public ClassificationPage blockify(List<TextPositionSequence> textPositions, CleanRulings cleanRulings, LayoutparsingVisualizations visualizations) {
+    public ClassificationPage blockify(List<TextPositionSequence> textPositions, CleanRulings cleanRulings, LayoutDebugLayer visualizations) {
 
         CleanRulings usedRulings = cleanRulings.withoutTextRulings();
 
@@ -43,18 +38,18 @@ public class RedactManagerBlockificationService {
         List<TextPositionSequence> chunkWords = new ArrayList<>();
         List<AbstractPageBlock> chunkBlockList = new ArrayList<>();
 
-        float minX = 1000, maxX = 0, minY = 1000, maxY = 0;
+        double minX = 1000, maxX = 0, minY = 1000, maxY = 0;
         TextPositionSequence prev = null;
 
         boolean wasSplitted = false;
-        Float splitX1 = null;
+        Double splitX1 = null;
         for (TextPositionSequence word : textPositions) {
 
-            boolean lineSeparation = word.getMinYDirAdj() - maxY > word.getHeight() * 1.25;
-            boolean startFromTop = prev != null && word.getMinYDirAdj() < prev.getMinYDirAdj() - prev.getTextHeight();
-            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 lineSeparation = word.getYDirAdj() - maxY > word.getHeight() * 1.25;
+            boolean startFromTop = prev != null && word.getYDirAdj() < prev.getYDirAdj() - prev.getTextHeight();
+            boolean splitByX = prev != null && maxX + 50 < word.getXDirAdj() && prev.getYDirAdj() == word.getYDirAdj();
+            boolean xIsBeforeFirstX = prev != null && word.getXDirAdj() < minX;
+            boolean newLineAfterSplit = prev != null && word.getYDirAdj() != prev.getYDirAdj() && wasSplitted && splitX1 != word.getXDirAdj();
             boolean isSplitByRuling = prev != null && usedRulings.lineBetween(prev, word);
             boolean splitByDir = prev != null && !prev.getDir().equals(word.getDir());
 
@@ -74,7 +69,7 @@ public class RedactManagerBlockificationService {
                 if (splitByX && !isSplitByRuling) {
                     wasSplitted = true;
                     cb1.setOrientation(Orientation.LEFT);
-                    splitX1 = word.getMinXDirAdj();
+                    splitX1 = word.getXDirAdj();
                 } else if (newLineAfterSplit && !isSplitByRuling) {
                     wasSplitted = false;
                     cb1.setOrientation(Orientation.RIGHT);
@@ -97,14 +92,14 @@ public class RedactManagerBlockificationService {
             chunkWords.add(word);
 
             prev = word;
-            if (word.getMinXDirAdj() < minX) {
-                minX = word.getMinXDirAdj();
+            if (word.getXDirAdj() < minX) {
+                minX = word.getXDirAdj();
             }
             if (word.getMaxXDirAdj() > maxX) {
                 maxX = word.getMaxXDirAdj();
             }
-            if (word.getMinYDirAdj() < minY) {
-                minY = word.getMinYDirAdj();
+            if (word.getYDirAdj() < minY) {
+                minY = word.getYDirAdj();
             }
             if (word.getMaxYDirAdj() > maxY) {
                 maxY = word.getMaxYDirAdj();

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

@@ -23,30 +23,36 @@ public class ClarifyndClassificationService {
 
     public void classifyDocument(ClassificationDocument document) {
 
-        List<Float> headlineFontSizes = document.getFontSizeCounter().getHighterThanMostPopular();
+        List<Double> headlineFontSizes = document.getFontSizeCounter().getHigherThanMostPopular();
 
         log.debug("Document FontSize counters are: {}", document.getFontSizeCounter().getCountPerValue());
 
+        HeadlineClassificationService headlineClassificationService = new HeadlineClassificationService();
+
         for (ClassificationPage page : document.getPages()) {
-            classifyPage(page, document, headlineFontSizes);
+            classifyPage(headlineClassificationService, page, document, headlineFontSizes);
         }
     }
 
 
-    private void classifyPage(ClassificationPage page, ClassificationDocument document, List<Float> headlineFontSizes) {
+    private void classifyPage(HeadlineClassificationService headlineClassificationService, ClassificationPage page, ClassificationDocument document, List<Double> headlineFontSizes) {
 
         for (AbstractPageBlock textBlock : page.getTextBlocks()) {
             if (textBlock instanceof TextPageBlock) {
-                classifyBlock((TextPageBlock) textBlock, page, document, headlineFontSizes);
+                classifyBlock(headlineClassificationService, (TextPageBlock) textBlock, page, document, headlineFontSizes);
             }
         }
     }
 
 
-    private void classifyBlock(TextPageBlock textBlock, ClassificationPage page, ClassificationDocument document, List<Float> headlineFontSizes) {
+    private void classifyBlock(HeadlineClassificationService headlineClassificationService, TextPageBlock textBlock, ClassificationPage page, ClassificationDocument document, List<Double> headlineFontSizes) {
 
         var bodyTextFrame = page.getBodyTextFrame();
 
+        if (textBlock.getClassification() != null && textBlock.getClassification().isHeadline()) {
+            headlineClassificationService.setLastHeadlineFromOutline(textBlock);
+            return;
+        }
         if (document.getFontSizeCounter().getMostPopular() == null) {
             textBlock.setClassification(PageBlockType.PARAGRAPH);
             return;
@@ -79,7 +85,8 @@ public class ClarifyndClassificationService {
 
             for (int i = 1; i <= headlineFontSizes.size(); i++) {
                 if (textBlock.getMostPopularWordFontSize() == headlineFontSizes.get(i - 1)) {
-                    textBlock.setClassification(PageBlockType.getHeadlineType(i));
+                    PageBlockType headlineType = PageBlockType.getHeadlineType(i);
+                    headlineClassificationService.classifyHeadline(textBlock, headlineType);
                     document.setHeadlines(true);
                 }
             }
@@ -89,7 +96,8 @@ public class ClarifyndClassificationService {
                 .getTextPositions()
                 .get(0)
                 .getFontSizeInPt() >= textBlock.getMostPopularWordFontSize()) {
-            textBlock.setClassification(PageBlockType.getHeadlineType(headlineFontSizes.size() + 1));
+            PageBlockType headlineType = PageBlockType.getHeadlineType(headlineFontSizes.size() + 1);
+            headlineClassificationService.classifyHeadline(textBlock, headlineType);
             document.setHeadlines(true);
         } else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock) && textBlock.getMostPopularWordFontSize() == document.getFontSizeCounter()
                 .getMostPopular() && textBlock.getMostPopularWordStyle().equals("bold") && !document.getFontStyleCounter().getMostPopular().equals("bold")) {

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

@@ -24,55 +24,63 @@ import lombok.extern.slf4j.Slf4j;
 @RequiredArgsConstructor
 public class DocuMineClassificationService {
 
-    private static final Pattern pattern = Pattern.compile("^([1-9]\\d?\\.){1,3}\\d{1,2}\\.?\\s[0-9A-Za-z \\[\\]]{2,50}", Pattern.CASE_INSENSITIVE);
-    private static final Pattern pattern2 = Pattern.compile("\\p{L}{3,}", Pattern.CASE_INSENSITIVE);
-    private static final Pattern pattern3 = Pattern.compile("^(\\d{1,1}\\.){1,3}\\d{1,2}\\.?\\s[a-z]{1,2}\\/[a-z]{1,2}.*");
+    private static final Pattern HEADLINE_WITH_IDENTIFER_PATTERN = Pattern.compile("^([1-9]\\d?\\.){1,3}\\d{1,2}\\.?\\s[0-9A-Za-z \\[\\]]{2,50}", Pattern.CASE_INSENSITIVE);
+    private static final Pattern AT_LEAST_3_PATTERN = Pattern.compile("\\p{L}{3,}", Pattern.CASE_INSENSITIVE);
+    private static final Pattern HEADLINE_PATTTERN_WITH_SLASHES = Pattern.compile("^(\\d{1,1}\\.){1,3}\\d{1,2}\\.?\\s[a-z]{1,2}\\/[a-z]{1,2}.*");
 
 
     public void classifyDocument(ClassificationDocument document) {
 
-        List<Float> headlineFontSizes = document.getFontSizeCounter().getHighterThanMostPopular();
+        List<Double> headlineFontSizes = document.getFontSizeCounter().getHigherThanMostPopular();
 
         log.debug("Document FontSize counters are: {}", document.getFontSizeCounter().getCountPerValue());
 
+        HeadlineClassificationService headlineClassificationService = new HeadlineClassificationService();
+
         for (ClassificationPage page : document.getPages()) {
-            classifyPage(page, document, headlineFontSizes);
+            classifyPage(headlineClassificationService, page, document, headlineFontSizes);
         }
     }
 
 
-    private void classifyPage(ClassificationPage page, ClassificationDocument document, List<Float> headlineFontSizes) {
+    private void classifyPage(HeadlineClassificationService headlineClassificationService,
+                              ClassificationPage page,
+                              ClassificationDocument document,
+                              List<Double> headlineFontSizes) {
 
         for (AbstractPageBlock textBlock : page.getTextBlocks()) {
             if (textBlock instanceof TextPageBlock) {
-                classifyBlock((TextPageBlock) textBlock, page, document, headlineFontSizes);
+                classifyBlock(headlineClassificationService, (TextPageBlock) textBlock, page, document, headlineFontSizes);
             }
         }
     }
 
 
-    private void classifyBlock(TextPageBlock textBlock, ClassificationPage page, ClassificationDocument document, List<Float> headlineFontSizes) {
+    private void classifyBlock(HeadlineClassificationService headlineClassificationService,
+                               TextPageBlock textBlock,
+                               ClassificationPage page,
+                               ClassificationDocument document,
+                               List<Double> headlineFontSizes) {
 
         log.debug("headlineFontSizes: {}", headlineFontSizes);
         var bodyTextFrame = page.getBodyTextFrame();
 
-        Matcher matcher = pattern.matcher(textBlock.toString());
-        Matcher matcher2 = pattern2.matcher(textBlock.toString());
-        Matcher matcher3 = pattern3.matcher(textBlock.toString());
+        Matcher headlineWithIdentifierMatcher = HEADLINE_WITH_IDENTIFER_PATTERN.matcher(textBlock.toString());
+        Matcher atLeast3Matcher = AT_LEAST_3_PATTERN.matcher(textBlock.toString());
+        Matcher headlineWithSlashesMatcher = HEADLINE_PATTTERN_WITH_SLASHES.matcher(textBlock.toString());
 
-        if (document.getFontSizeCounter().getMostPopular() == null) {
-            textBlock.setClassification(PageBlockType.OTHER);
+        if (textBlock.getClassification() != null && textBlock.getClassification().isHeadline()) {
+            headlineClassificationService.setLastHeadlineFromOutline(textBlock);
             return;
         }
-        if (MarkedContentUtils.intersects(textBlock, page.getMarkedContentBboxPerType(), MarkedContentUtils.HEADER)
-            || (PositionUtils.isOverBodyTextFrame(bodyTextFrame,
-                                                  textBlock,
-                                                  page.getRotation()) && (document.getFontSizeCounter().getMostPopular()
-                                                                          == null
-                                                                          || textBlock.getHighestFontSize()
-                                                                             <= document.getFontSizeCounter()
-                                                                                     .getMostPopular()))
-            || HeaderFooterDetection.isLikelyHeader(textBlock, document, page)) {
+        if (document.getFontSizeCounter().getMostPopular() == null) {
+            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)
@@ -95,19 +103,31 @@ public class DocuMineClassificationService {
                    && (textBlock.getMostPopularWordHeight() > document.getTextHeightCounter().getMostPopular()
                        || textBlock.getMostPopularWordFontSize() > document.getFontSizeCounter().getMostPopular())
                    && PositionUtils.getApproxLineCount(textBlock) < 5.9
-                   && (textBlock.getMostPopularWordStyle().contains("bold") && Character.isDigit(textBlock.toString().charAt(0)) && matcher2.reset().find() && !textBlock.toString()
-                .contains(":")
-                       || textBlock.toString().equals(textBlock.toString().toUpperCase(Locale.ROOT)) && matcher2.reset().find() && !textBlock.toString().contains(":")
-                       || textBlock.toString().startsWith("APPENDIX")
-                       || textBlock.toString().startsWith("FIGURE")
+
+                   && (textBlock.getMostPopularWordStyle().contains("bold")
+                       && Character.isDigit(textBlock.toString().charAt(0))
+                       && atLeast3Matcher.reset().find()
+                       && !textBlock.toString().contains(":") //
+                       || textBlock.toString().equals(textBlock.toString().toUpperCase(Locale.ROOT))
+                          && atLeast3Matcher.reset().find()
+                          && !textBlock.toString().contains(":")
+                          && !textBlock.toString().startsWith("(")//
+                       || textBlock.toString().startsWith("APPENDIX") //
+                       || textBlock.toString().startsWith("FIGURE") //
+                       || textBlock.toString().startsWith("Continued TABLE") //
                        || textBlock.toString().startsWith("TABLE"))
                    && !textBlock.toString().endsWith(":")
-                   && matcher2.reset().find()) {
-            textBlock.setClassification(PageBlockType.getHeadlineType(1));
+                   && atLeast3Matcher.reset().find()) {
+            PageBlockType headlineType = PageBlockType.getHeadlineType(1);
+            headlineClassificationService.classifyHeadline(textBlock, headlineType);
             document.setHeadlines(true);
 
-        } else if (matcher.reset().find() && PositionUtils.getApproxLineCount(textBlock) < 2.9 && matcher2.reset().find() && !matcher3.reset().matches()) {
-            textBlock.setClassification(PageBlockType.getHeadlineType(2));
+        } else if (headlineWithIdentifierMatcher.reset().find()
+                   && PositionUtils.getApproxLineCount(textBlock) < 2.9
+                   && atLeast3Matcher.reset().find()
+                   && !headlineWithSlashesMatcher.reset().matches()) {
+            PageBlockType headlineType = PageBlockType.getHeadlineType(2);
+            headlineClassificationService.classifyHeadline(textBlock, headlineType);
             document.setHeadlines(true);
         } else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock)
                    && textBlock.getMostPopularWordFontSize() == document.getFontSizeCounter().getMostPopular()
@@ -126,9 +146,9 @@ public class DocuMineClassificationService {
                    && PositionUtils.getApproxLineCount(textBlock) < 2.9) {
             textBlock.setClassification(PageBlockType.PARAGRAPH_ITALIC);
         } else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock)) {
-            textBlock.setClassification(PageBlockType.PARAGRAPH_UNKNOWN);
+            textBlock.setClassification(PageBlockType.PARAGRAPH);
         } 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/HeadlineClassificationService.java

@@ -0,0 +1,52 @@
+package com.knecon.fforesight.service.layoutparser.processor.services.classification;
+
+import static com.knecon.fforesight.service.layoutparser.processor.model.PageBlockType.getHeadlineNumber;
+
+import com.knecon.fforesight.service.layoutparser.processor.model.PageBlockType;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
+
+import lombok.Getter;
+import lombok.Setter;
+
+@Getter
+@Setter
+public class HeadlineClassificationService {
+
+    TextPageBlock lastHeadline;
+    PageBlockType originalClassifiedBlockType;
+    TextPageBlock lastHeadlineFromOutline;
+
+    public void setLastHeadlineFromOutline(TextPageBlock lastHeadlineFromOutline) {
+
+        this.lastHeadlineFromOutline = lastHeadlineFromOutline;
+        this.setLastHeadline(lastHeadlineFromOutline);
+    }
+
+
+    public void classifyHeadline(TextPageBlock textBlock, PageBlockType initialHeadlineType) {
+
+        TextPageBlock lastHeadline = getLastHeadline();
+        TextPageBlock lastHeadlineFromOutline = getLastHeadlineFromOutline();
+        PageBlockType originalClassifiedBlockType = getOriginalClassifiedBlockType();
+        PageBlockType finalHeadlineType = initialHeadlineType;
+
+        if (lastHeadline != null) {
+
+            if (lastHeadline.equals(lastHeadlineFromOutline)) {
+
+                finalHeadlineType = PageBlockType.getHeadlineType(getHeadlineNumber(lastHeadline.getClassification()) + 1);
+
+            } else if (originalClassifiedBlockType != null && lastHeadline.getClassification() != originalClassifiedBlockType) {
+
+                PageBlockType lastHeadlineType = lastHeadline.getClassification();
+                int difference = getHeadlineNumber(originalClassifiedBlockType) - getHeadlineNumber(lastHeadlineType);
+                finalHeadlineType = PageBlockType.getHeadlineType(getHeadlineNumber(initialHeadlineType) - difference);
+            }
+        }
+
+        setOriginalClassifiedBlockType(initialHeadlineType);
+        textBlock.setClassification(finalHeadlineType);
+        setLastHeadline(textBlock);
+    }
+
+}

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

@@ -22,34 +22,41 @@ import lombok.extern.slf4j.Slf4j;
 @RequiredArgsConstructor
 public class RedactManagerClassificationService {
 
+
     public void classifyDocument(ClassificationDocument document) {
 
-        List<Float> headlineFontSizes = document.getFontSizeCounter().getHighterThanMostPopular();
+        List<Double> headlineFontSizes = document.getFontSizeCounter().getHigherThanMostPopular();
 
         log.debug("Document FontSize counters are: {}", document.getFontSizeCounter().getCountPerValue());
 
+        HeadlineClassificationService headlineClassificationService = new HeadlineClassificationService();
+
         for (ClassificationPage page : document.getPages()) {
-            classifyPage(page, document, headlineFontSizes);
+            classifyPage(headlineClassificationService, page, document, headlineFontSizes);
         }
     }
 
 
-    private void classifyPage(ClassificationPage page, ClassificationDocument document, List<Float> headlineFontSizes) {
+    private void classifyPage(HeadlineClassificationService headlineClassificationService, ClassificationPage page, ClassificationDocument document, List<Double> headlineFontSizes) {
 
         for (AbstractPageBlock textBlock : page.getTextBlocks()) {
             if (textBlock instanceof TextPageBlock) {
-                classifyBlock((TextPageBlock) textBlock, page, document, headlineFontSizes);
+                classifyBlock(headlineClassificationService, (TextPageBlock) textBlock, page, document, headlineFontSizes);
             }
         }
     }
 
 
-    private void classifyBlock(TextPageBlock textBlock, ClassificationPage page, ClassificationDocument document, List<Float> headlineFontSizes) {
+    private void classifyBlock(HeadlineClassificationService headlineClassificationService, TextPageBlock textBlock, ClassificationPage page, ClassificationDocument document, List<Double> headlineFontSizes) {
 
         var bodyTextFrame = page.getBodyTextFrame();
 
+        if (textBlock.getClassification() != null && textBlock.getClassification().isHeadline()) {
+            headlineClassificationService.setLastHeadlineFromOutline(textBlock);
+            return;
+        }
         if (document.getFontSizeCounter().getMostPopular() == null) {
-            textBlock.setClassification(PageBlockType.OTHER);
+            textBlock.setClassification(PageBlockType.PARAGRAPH);
             return;
         }
 
@@ -60,63 +67,69 @@ public class RedactManagerClassificationService {
             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()
+        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()
+        } 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()
-                .size() == 1)) {
+        } else if (page.getPageNumber() == 1 && (PositionUtils.getHeightDifferenceBetweenChunkWordAndDocumentWord(textBlock, document.getTextHeightCounter().getMostPopular()) > 2.5
+                                                 && textBlock.getHighestFontSize() > document.getFontSizeCounter().getMostPopular() || page.getTextBlocks().size() == 1)) {
             if (!Pattern.matches("[0-9]+", textBlock.toString())) {
                 textBlock.setClassification(PageBlockType.TITLE);
             }
-        } else if (textBlock.getMostPopularWordFontSize() > document.getFontSizeCounter()
-                .getMostPopular() && PositionUtils.getApproxLineCount(textBlock) < 4.9 && (textBlock.getMostPopularWordStyle().equals("bold") || !document.getFontStyleCounter()
-                .getCountPerValue()
-                .containsKey("bold") && textBlock.getMostPopularWordFontSize() > document.getFontSizeCounter().getMostPopular() + 1) && textBlock.getSequences()
-                .get(0)
-                .getTextPositions()
-                .get(0)
-                .getFontSizeInPt() >= textBlock.getMostPopularWordFontSize()) {
+        } else if (textBlock.getMostPopularWordFontSize() > document.getFontSizeCounter().getMostPopular()
+                   && PositionUtils.getApproxLineCount(textBlock) < 4.9
+                   && (textBlock.getMostPopularWordStyle().equals("bold")
+                       || !document.getFontStyleCounter().getCountPerValue().containsKey("bold")
+                          && textBlock.getMostPopularWordFontSize() > document.getFontSizeCounter().getMostPopular() + 1)
+                   && textBlock.getSequences()
+                              .get(0).getTextPositions()
+                              .get(0).getFontSizeInPt() >= textBlock.getMostPopularWordFontSize()) {
 
             for (int i = 1; i <= headlineFontSizes.size(); i++) {
                 if (textBlock.getMostPopularWordFontSize() == headlineFontSizes.get(i - 1)) {
-                    textBlock.setClassification(PageBlockType.getHeadlineType(i));
+                    PageBlockType headlineType = PageBlockType.getHeadlineType(i);
+                    headlineClassificationService.classifyHeadline(textBlock, headlineType);
                     document.setHeadlines(true);
                 }
             }
-        } else if (!textBlock.getText().startsWith("Figure ") && PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock) && textBlock.getMostPopularWordStyle()
-                .equals("bold") && !document.getFontStyleCounter().getMostPopular().equals("bold") && PositionUtils.getApproxLineCount(textBlock) < 2.9 && textBlock.getSequences()
-                .get(0)
-                .getTextPositions()
-                .get(0)
-                .getFontSizeInPt() >= textBlock.getMostPopularWordFontSize()) {
-            textBlock.setClassification(PageBlockType.getHeadlineType(headlineFontSizes.size() + 1));
+        } else if (!textBlock.getText().startsWith("Figure ")
+                   && PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock)
+                   && textBlock.getMostPopularWordStyle().equals("bold")
+                   && !document.getFontStyleCounter().getMostPopular().equals("bold")
+                   && PositionUtils.getApproxLineCount(textBlock) < 2.9
+                   && textBlock.getSequences()
+                              .get(0).getTextPositions()
+                              .get(0).getFontSizeInPt() >= textBlock.getMostPopularWordFontSize()) {
+            PageBlockType headlineType = PageBlockType.getHeadlineType(headlineFontSizes.size() + 1);
+            headlineClassificationService.classifyHeadline(textBlock, headlineType);
             document.setHeadlines(true);
-        } else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock) && textBlock.getMostPopularWordFontSize() == document.getFontSizeCounter()
-                .getMostPopular() && textBlock.getMostPopularWordStyle().equals("bold") && !document.getFontStyleCounter().getMostPopular().equals("bold")) {
+        } else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock)
+                   && textBlock.getMostPopularWordFontSize() == document.getFontSizeCounter().getMostPopular()
+                   && textBlock.getMostPopularWordStyle().equals("bold")
+                   && !document.getFontStyleCounter().getMostPopular().equals("bold")) {
             textBlock.setClassification(PageBlockType.PARAGRAPH_BOLD);
-        } else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock) && textBlock.getMostPopularWordFont()
-                .equals(document.getFontCounter().getMostPopular()) && textBlock.getMostPopularWordStyle()
-                .equals(document.getFontStyleCounter().getMostPopular()) && textBlock.getMostPopularWordFontSize() == document.getFontSizeCounter().getMostPopular()) {
+        } else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock)
+                   && textBlock.getMostPopularWordFont().equals(document.getFontCounter().getMostPopular())
+                   && textBlock.getMostPopularWordStyle().equals(document.getFontStyleCounter().getMostPopular())
+                   && textBlock.getMostPopularWordFontSize() == document.getFontSizeCounter().getMostPopular()) {
             textBlock.setClassification(PageBlockType.PARAGRAPH);
-        } else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock) && textBlock.getMostPopularWordFontSize() == document.getFontSizeCounter()
-                .getMostPopular() && textBlock.getMostPopularWordStyle().equals("italic") && !document.getFontStyleCounter()
-                .getMostPopular()
-                .equals("italic") && PositionUtils.getApproxLineCount(textBlock) < 2.9) {
+        } else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock)
+                   && textBlock.getMostPopularWordFontSize() == document.getFontSizeCounter().getMostPopular()
+                   && textBlock.getMostPopularWordStyle().equals("italic")
+                   && !document.getFontStyleCounter().getMostPopular().equals("italic")
+                   && PositionUtils.getApproxLineCount(textBlock) < 2.9) {
             textBlock.setClassification(PageBlockType.PARAGRAPH_ITALIC);
         } 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/factory/DocumentGraphFactory.java

@@ -6,14 +6,17 @@ import static java.util.stream.Collectors.toList;
 
 import java.awt.geom.Rectangle2D;
 import java.util.ArrayList;
+import java.util.Collection;
 import java.util.HashMap;
 import java.util.LinkedList;
 import java.util.List;
 import java.util.Map;
 import java.util.NoSuchElementException;
+import java.util.Optional;
 import java.util.Set;
 import java.util.stream.Collectors;
 
+import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
 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;
@@ -21,6 +24,7 @@ import com.knecon.fforesight.service.layoutparser.processor.model.Classification
 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.graph.DocumentTree;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.AbstractSemanticNode;
 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;
@@ -30,9 +34,11 @@ import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.He
 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.Paragraph;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Section;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.SemanticNode;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.AtomicTextBlock;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.image.ClassifiedImage;
+import com.knecon.fforesight.service.layoutparser.processor.model.outline.TableOfContentItem;
 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;
@@ -43,25 +49,22 @@ import lombok.Builder;
 import lombok.Getter;
 import lombok.experimental.FieldDefaults;
 import lombok.experimental.UtilityClass;
+import lombok.extern.slf4j.Slf4j;
 
 @UtilityClass
+@Slf4j
 public class DocumentGraphFactory {
 
     public Document buildDocumentGraph(LayoutParsingType layoutParsingType, ClassificationDocument document) {
 
         Document documentGraph = new Document();
 
-        documentGraph.setVisualizations(document.getVisualizations());
+        documentGraph.setLayoutDebugLayer(document.getLayoutDebugLayer());
 
         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(layoutParsingType, document, context, documentGraph);
         addHeaderAndFooterToEachPage(document, context);
 
@@ -69,14 +72,38 @@ public class DocumentGraphFactory {
         documentGraph.setPages(context.pages.keySet());
         documentGraph.setDocumentTree(context.documentTree);
         documentGraph.setTextBlock(documentGraph.getTextBlock());
+        addTextBlocksToPages(documentGraph);
+
         return documentGraph;
     }
 
 
+    private void addTextBlocksToPages(Document documentGraph) {
+
+        documentGraph.streamAllSubNodes()
+                .filter(SemanticNode::isLeaf)
+                .filter(node -> !node.getType().equals(NodeType.HEADER))
+                .filter(node -> !node.getType().equals(NodeType.FOOTER))
+                .map(SemanticNode::getTextBlock)
+                .map(TextBlock::getAtomicTextBlocks)
+                .flatMap(Collection::stream)
+                .forEach(atb -> atb.getPage().getTextBlocksOnPage().add(atb));
+    }
+
+
     private void addSections(LayoutParsingType layoutParsingType, ClassificationDocument classificationDocument, Context context, Document document) {
 
-        classificationDocument.getSections()
-                .forEach(section -> SectionNodeFactory.addSection(layoutParsingType, null, section.getNonEmptyPageBlocks(), section.getImages(), context, document));
+        for (TableOfContentItem tocItem : classificationDocument.getTableOfContents()) {
+            var parent = tocItem.getParent() == null ? null : tocItem.getParent().getSection();
+            Optional<AbstractSemanticNode> section = SectionNodeFactory.addSection(layoutParsingType,
+                                                                                   parent,
+                                                                                   tocItem.getChildren().isEmpty(),
+                                                                                   tocItem.getNonEmptySectionBlocks(),
+                                                                                   tocItem.getImages(),
+                                                                                   context,
+                                                                                   document);
+            tocItem.setSection(section.orElse(null));
+        }
     }
 
 
@@ -97,25 +124,24 @@ public class DocumentGraphFactory {
             node = Paragraph.builder().documentTree(context.getDocumentTree()).build();
         }
 
-        page.getMainBody().add(node);
-
         List<TextPageBlock> textBlocks = new ArrayList<>();
         textBlocks.add(originalTextBlock);
         textBlocks.addAll(textBlocksToMerge);
 
-        AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock(TextPositionOperations.mergeAndSortTextPositionSequenceByYThenX(textBlocks), node, context, page);
+        AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock2(TextPositionOperations.mergeAndSort(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);
+            AtomicTextBlock unsortedTextBlock = context.textBlockFactory.buildAtomicTextBlock2(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);
+        node.getEngines().addAll(originalTextBlock.getEngines());
     }
 
 
@@ -132,16 +158,15 @@ public class DocumentGraphFactory {
 
         Rectangle2D position = image.getPosition();
         Page page = context.getPage(image.getPage());
-        Image imageNode = Image.builder()
+        return Image.builder()
                 .id(IdBuilder.buildId(Set.of(page), List.of(position)))
                 .imageType(image.getImageType())
                 .position(position)
                 .transparent(image.isHasTransparency())
                 .page(page)
+                .representationHash(image.getRepresentation())
                 .documentTree(context.getDocumentTree())
                 .build();
-        page.getMainBody().add(imageNode);
-        return imageNode;
     }
 
 
@@ -181,10 +206,7 @@ public class DocumentGraphFactory {
 
         Page page = context.getPage(textBlocks.get(0).getPage());
         Footer footer = Footer.builder().documentTree(context.getDocumentTree()).build();
-        AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock(TextPositionOperations.mergeTextPositionSequence(textBlocks),
-                                                                                  footer,
-                                                                                  context,
-                                                                                  page);
+        AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock2(TextPositionOperations.merge(textBlocks), footer, context, page);
         List<Integer> tocId = context.getDocumentTree().createNewMainEntryAndReturnId(footer);
         footer.setTreeId(tocId);
         footer.setLeafTextBlock(textBlock);
@@ -196,7 +218,7 @@ public class DocumentGraphFactory {
 
         Page page = context.getPage(textBlocks.get(0).getPage());
         Header header = Header.builder().documentTree(context.getDocumentTree()).build();
-        AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock(TextPositionOperations.mergeTextPositionSequence(textBlocks), header, 0, page);
+        AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock(TextPositionOperations.merge(textBlocks), header, 0, page);
         List<Integer> tocId = context.getDocumentTree().createNewMainEntryAndReturnId(header);
         header.setTreeId(tocId);
         header.setLeafTextBlock(textBlock);
@@ -236,7 +258,7 @@ public class DocumentGraphFactory {
 
         DocumentTree documentTree;
         Map<Page, Integer> pages;
-        List<Section> sections;
+        List<AbstractSemanticNode> sections;
         List<ClassifiedImage> images;
         TextBlockFactory textBlockFactory;
 
@@ -273,7 +295,8 @@ public class DocumentGraphFactory {
             return pages.keySet()
                     .stream()
                     .filter(page -> page.getNumber() == pageIndex)
-                    .findFirst().orElseThrow(() -> new NoSuchElementException(format("ClassificationPage with number %d not found", pageIndex)));
+                    .findFirst()
+                    .orElseThrow(() -> new NoSuchElementException(format("ClassificationPage with number %d not found", pageIndex)));
         }
 
     }

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

@@ -4,7 +4,7 @@ import java.awt.geom.Rectangle2D;
 import java.util.Collections;
 import java.util.List;
 
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.Boundary;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.TextRange;
 
 import lombok.AccessLevel;
 import lombok.Builder;
@@ -19,8 +19,8 @@ public class SearchTextWithTextPositionDto {
     String searchText;
     List<Integer> lineBreaks;
     List<Integer> stringIdxToPositionIdx;
-    List<Boundary> boldTextBoundaries;
-    List<Boundary> italicTextBoundaries;
+    List<TextRange> boldTextBoundaries;
+    List<TextRange> italicTextBoundaries;
     List<Rectangle2D> positions;
 
 

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

@@ -1,6 +1,7 @@
 package com.knecon.fforesight.service.layoutparser.processor.services.factory;
 
-import java.awt.geom.AffineTransform;
+import static com.knecon.fforesight.service.layoutparser.processor.model.text.RedTextPosition.HEIGHT_PADDING;
+
 import java.awt.geom.Rectangle2D;
 import java.util.Collection;
 import java.util.Collections;
@@ -9,9 +10,8 @@ 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.graph.TextRange;
 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 lombok.experimental.UtilityClass;
@@ -19,13 +19,13 @@ import lombok.experimental.UtilityClass;
 @UtilityClass
 public class SearchTextWithTextPositionFactory {
 
-    public final int HEIGHT_PADDING = 2;
     // when checking for a hyphen linebreak, we need to check after a linebreak if the last hyphen was less than three symbols away.
     // We detect a linebreak as either a "\n" character or if two adjacent symbol's position differ in y-coordinates by at least one character height.
     // If there is a hyphen linebreak, the hyphen will be 1 position in front of a "\n" or 2 positions in front of the character which has a lower y-coordinate
     // This is why, we need to initialize this to < -2, otherwise, if the very first symbol is a \n we would detect a hyphen linebreak that isn't there.
     // Also, Integer.MIN_VALUE is a bad idea due to potential overflow during arithmetic operations. This is why the default should be -3.
     public final int MAX_HYPHEN_LINEBREAK_DISTANCE = 3;
+    public static final double LINEBREAK_DELTA_TOLERANCE = 1.5;
 
 
     public SearchTextWithTextPositionDto buildSearchTextToTextPositionDto(List<TextPositionSequence> sequences) {
@@ -37,15 +37,13 @@ public class SearchTextWithTextPositionFactory {
 
         Context context = new Context();
 
-        RedTextPosition currentTextPosition = sequences.get(0).getTextPositions()
-                .get(0);
+        RedTextPosition currentTextPosition = sequences.get(0).getTextPositions().get(0);
         RedTextPosition previousTextPosition = RedTextPosition.builder().unicode(" ").bBoxDirAdj(currentTextPosition.getBBoxDirAdj()).build();
 
         for (TextPositionSequence word : sequences) {
             for (int i = 0; i < word.getTextPositions().size(); ++i) {
 
-                currentTextPosition = word.getTextPositions()
-                        .get(i);
+                currentTextPosition = word.getTextPositions().get(i);
                 if (isLineBreak(currentTextPosition, previousTextPosition)) {
                     removeHyphenLinebreaks(context);
                     context.lineBreaksStringIdx.add(context.stringIdx);
@@ -67,11 +65,10 @@ public class SearchTextWithTextPositionFactory {
             ++context.stringIdx;
         }
 
-
         List<Rectangle2D> positions = sequences.stream()
                 .map(TextPositionSequence::getTextPositions)
                 .flatMap(Collection::stream)
-                .map(RedTextPosition::getBBoxInitialUserSpace)
+                .map(RedTextPosition::getBBoxPdf)
                 .toList();
 
         assert context.stringBuilder.length() == context.stringIdxToPositionIdx.size();
@@ -118,23 +115,23 @@ public class SearchTextWithTextPositionFactory {
     }
 
 
-    private static List<Boundary> mergeToBoundaries(List<Integer> integers) {
+    private static List<TextRange> mergeToBoundaries(List<Integer> integers) {
 
         if (integers.isEmpty()) {
             return Collections.emptyList();
         }
-        List<Boundary> boundaries = new LinkedList<>();
+        List<TextRange> boundaries = new LinkedList<>();
         int start = integers.get(0);
         int end = integers.get(0) + 1;
         for (int current : integers) {
             if (current > end + 1) {
-                boundaries.add(new Boundary(start, end));
+                boundaries.add(new TextRange(start, end));
                 start = current;
             }
             end = current + 1;
         }
         if (boundaries.isEmpty()) {
-            boundaries.add(new Boundary(start, end));
+            boundaries.add(new TextRange(start, end));
         }
         return boundaries;
     }
@@ -160,8 +157,8 @@ public class SearchTextWithTextPositionFactory {
             return false;
         }
 
-        double deltaY = Math.abs(currentPosition.getYDirAdj() - previousPosition.getYDirAdj());
-        return deltaY >= currentPosition.getHeightDir();
+        double deltaY = (Math.abs(currentPosition.getYDirAdj() - previousPosition.getYDirAdj()) * LINEBREAK_DELTA_TOLERANCE) + (2 * HEIGHT_PADDING);
+        return deltaY >= currentPosition.getHeightDirAdj() || deltaY >= previousPosition.getHeightDirAdj();
     }
 
 
@@ -187,32 +184,6 @@ public class SearchTextWithTextPositionFactory {
     }
 
 
-    public Rectangle2D mapRedTextPositionToInitialUserSpace(RedTextPosition textPosition, TextPositionSequence sequence) {
-
-        float textHeight = sequence.getTextHeight() + HEIGHT_PADDING;
-        Rectangle2D rectangle2D = new Rectangle2D.Double(textPosition.getXDirAdj(),
-                textPosition.getYDirAdj() - textHeight,
-                textPosition.getWidthDirAdj(),
-                textHeight + HEIGHT_PADDING);
-
-        AffineTransform transform = new AffineTransform();
-
-        if (sequence.getDir() == TextDirection.ZERO || sequence.getDir() == TextDirection.HALF_CIRCLE) {
-            transform.rotate(sequence.getDir().getRadians(), sequence.getPageWidth() / 2f, sequence.getPageHeight() / 2f);
-            transform.translate(0f, sequence.getPageHeight());
-        } else if (sequence.getDir() == TextDirection.QUARTER_CIRCLE) {
-            transform.rotate(sequence.getDir().getRadians(), sequence.getPageWidth() / 2f, sequence.getPageWidth() / 2f);
-            transform.translate(0f, sequence.getPageWidth());
-        } else {
-            transform.rotate(sequence.getDir().getRadians(), sequence.getPageHeight() / 2f, sequence.getPageHeight() / 2f);
-            transform.translate(0f, sequence.getPageWidth());
-        }
-        transform.scale(1., -1.);
-
-        return transform.createTransformedShape(rectangle2D).getBounds2D();
-    }
-
-
     private class Context {
 
         List<Integer> stringIdxToPositionIdx = new LinkedList<>();

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

@@ -9,14 +9,17 @@ import java.util.HashSet;
 import java.util.LinkedList;
 import java.util.List;
 import java.util.Map;
+import java.util.Optional;
 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.graph.nodes.AbstractSemanticNode;
 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.graph.nodes.SuperSection;
 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;
@@ -27,12 +30,13 @@ import lombok.experimental.UtilityClass;
 @UtilityClass
 public class SectionNodeFactory {
 
-    public void addSection(LayoutParsingType layoutParsingType,
-                           GenericSemanticNode parentNode,
-                           List<AbstractPageBlock> pageBlocks,
-                           List<ClassifiedImage> images,
-                           DocumentGraphFactory.Context context,
-                           Document document) {
+    public Optional<AbstractSemanticNode> addSection(LayoutParsingType layoutParsingType,
+                                                     GenericSemanticNode parentNode,
+                                                     boolean isLeaf,
+                                                     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.
@@ -40,31 +44,36 @@ public class SectionNodeFactory {
             images.stream()
                     .distinct()
                     .forEach(image -> DocumentGraphFactory.addImage(document, image, context));
-            return;
+            return Optional.empty();
         }
 
         if (pageBlocks.isEmpty()) {
-            return;
+            return Optional.empty();
         }
 
-        Map<Integer, List<AbstractPageBlock>> blocksPerPage = pageBlocks.stream()
-                .collect(groupingBy(AbstractPageBlock::getPage));
-        Section section = Section.builder().documentTree(context.getDocumentTree()).build();
+        AbstractSemanticNode section;
+        boolean containsTablesAndTextBlocks = containsTablesAndTextBlocks(pageBlocks);
+        if (isLeaf && !containsTablesAndTextBlocks) {
+            section = Section.builder().documentTree(context.getDocumentTree()).build();
+        } else {
+            section = SuperSection.builder().documentTree(context.getDocumentTree()).build();
+        }
 
         context.getSections().add(section);
-        blocksPerPage.keySet()
-                .forEach(pageNumber -> addSectionNodeToPageNode(context, section, pageNumber));
 
         section.setTreeId(getTreeId(parentNode, context, section));
 
         addFirstHeadlineDirectlyToSection(layoutParsingType, pageBlocks, context, section, document);
-        if (containsTablesAndTextBlocks(pageBlocks)) {
+        if (containsTablesAndTextBlocks) {
             splitPageBlocksIntoSubSections(pageBlocks).forEach(subSectionPageBlocks -> addSection(layoutParsingType,
                                                                                                   section,
+                                                                                                  true,
                                                                                                   subSectionPageBlocks,
                                                                                                   emptyList(),
                                                                                                   context,
                                                                                                   document));
+        } else if (!isLeaf) {
+            addSection(layoutParsingType, section, true, pageBlocks, emptyList(), context, document);
         } else {
             addTablesAndParagraphsAndHeadlinesToSection(layoutParsingType, pageBlocks, context, section, document);
         }
@@ -72,10 +81,12 @@ public class SectionNodeFactory {
         images.stream()
                 .distinct()
                 .forEach(image -> DocumentGraphFactory.addImage(section, image, context));
+
+        return Optional.of(section);
     }
 
 
-    private List<Integer> getTreeId(GenericSemanticNode parentNode, DocumentGraphFactory.Context context, Section section) {
+    private List<Integer> getTreeId(GenericSemanticNode parentNode, DocumentGraphFactory.Context context, AbstractSemanticNode section) {
 
         if (parentNode == null) {
             return context.getDocumentTree().createNewMainEntryAndReturnId(section);
@@ -88,7 +99,7 @@ public class SectionNodeFactory {
     private void addFirstHeadlineDirectlyToSection(LayoutParsingType layoutParsingType,
                                                    List<AbstractPageBlock> pageBlocks,
                                                    DocumentGraphFactory.Context context,
-                                                   Section section,
+                                                   AbstractSemanticNode section,
                                                    Document document) {
 
         if (pageBlocks.get(0).isHeadline()) {
@@ -101,7 +112,7 @@ public class SectionNodeFactory {
     private void addTablesAndParagraphsAndHeadlinesToSection(LayoutParsingType layoutParsingType,
                                                              List<AbstractPageBlock> pageBlocks,
                                                              DocumentGraphFactory.Context context,
-                                                             Section section,
+                                                             AbstractSemanticNode section,
                                                              Document document) {
 
         Set<AbstractPageBlock> alreadyMerged = new HashSet<>();
@@ -117,7 +128,7 @@ public class SectionNodeFactory {
             if (abstractPageBlock instanceof TextPageBlock) {
 
                 switch (layoutParsingType) {
-                    case REDACT_MANAGER, DOCUMINE, CLARIFYND, CLARIFYND_PARAGRAPH_DEBUG -> {
+                    case REDACT_MANAGER, DOCUMINE, CLARIFYND, CLARIFYND_PARAGRAPH_DEBUG, REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH -> {
                         alreadyMerged.add(abstractPageBlock);
                         remainingBlocks.remove(abstractPageBlock);
                         DocumentGraphFactory.addParagraphOrHeadline(section, (TextPageBlock) abstractPageBlock, context, new ArrayList<>(), layoutParsingType);
@@ -218,7 +229,7 @@ public class SectionNodeFactory {
                 .filter(abstractTextContainer -> !abstractTextContainer.equals(atc))
                 .filter(abstractTextContainer -> abstractTextContainer.getPage() == atc.getPage())
                 .filter(abstractTextContainer -> abstractTextContainer instanceof TextPageBlock)
-                .filter(abstractTextContainer -> abstractTextContainer.intersectsY(atc))
+                .filter(abstractTextContainer -> abstractTextContainer.intersectsYPdf(atc))
                 .map(abstractTextContainer -> (TextPageBlock) abstractTextContainer)
                 .filter(abstractTextContainer -> abstractTextContainer.getDir() == atc.getDir())
                 .filter(abstractTextContainer -> !abstractTextContainer.isToDuplicate())
@@ -226,10 +237,5 @@ public class SectionNodeFactory {
     }
 
 
-    private void addSectionNodeToPageNode(DocumentGraphFactory.Context context, Section section, Integer pageNumber) {
-
-        Page page = context.getPage(pageNumber);
-        page.getMainBody().add(section);
-    }
 
 }