Merge branch 'CLARI-003' into 'main'

CLARI-003: add treeId to StructureObject See merge request fforesight/layout-parser!176
CLARI-003: add treeId to StructureObject
2024-07-02 11:37:22 +02:00 · 2024-07-01 13:56:16 +02:00 · 2024-07-01 11:13:27 +02:00 · 2024-07-01 11:13:26 +02:00 · 2024-06-26 13:42:42 +02:00 · 2024-06-26 09:10:59 +02:00
52 changed files with 1596 additions and 199 deletions
@@ -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));
    }


@@ -2,5 +2,6 @@ package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;

 public enum LayoutEngine {
    ALGORITHM,
-    AI
+    AI,
+    OUTLINE
 }
@@ -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;

@@ -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.")
@@ -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,
@@ -51,8 +51,8 @@ import com.knecon.fforesight.service.layoutparser.processor.services.RulingClean
 import com.knecon.fforesight.service.layoutparser.processor.services.SectionsBuilderService;
 import com.knecon.fforesight.service.layoutparser.processor.services.SimplifiedSectionTextService;
 import com.knecon.fforesight.service.layoutparser.processor.services.TableExtractionService;
-import com.knecon.fforesight.service.layoutparser.processor.services.blockification.BlockificationPostprocessingService;
 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;
@@ -155,7 +155,12 @@ public class LayoutParsingPipeline {

        log.info("Creating viewer document for {}", layoutParsingRequest.identifier());

-        layoutGridService.addLayoutGrid(viewerDocumentFile, documentGraph, viewerDocumentFile, false, layoutParsingRequest.visualLayoutParsingFileId().isPresent());
+        layoutGridService.addLayoutGrid(viewerDocumentFile,
+                                        documentGraph,
+                                        viewerDocumentFile,
+                                        false,
+                                        layoutParsingRequest.visualLayoutParsingFileId()
+                                                .isPresent());

        log.info("Storing resulting files for {}", layoutParsingRequest.identifier());

@@ -163,7 +168,7 @@ public class LayoutParsingPipeline {
        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);
@@ -254,7 +259,7 @@ public class LayoutParsingPipeline {
        OutlineObject lastProcessedOutlineObject = null;

        // parsing the structure elements could be useful as well
-        if(layoutParsingType != LayoutParsingType.REDACT_MANAGER_OLD && layoutParsingType != LayoutParsingType.DOCUMINE_OLD) {
+        if (layoutParsingType != LayoutParsingType.REDACT_MANAGER_OLD && layoutParsingType != LayoutParsingType.DOCUMINE_OLD) {
            classificationDocument.setOutlineObjectTree(outlineExtractorService.getOutlineObjectTree(originDocument));
        }

@@ -302,24 +307,20 @@ public class LayoutParsingPipeline {

            TextRulingsClassifier.classifyUnderlinedAndStrikethroughText(words, cleanRulings);

-            List<Box> graphics = graphicExtractorService.extractPathElementGraphics(originDocument,
-                                                                              pdPage,
-                                                                              pageNumber,
-                                                                              cleanRulings,
-                                                                              stripper.getTextPositionSequences(),
+            List<Box> graphics = graphicExtractorService.extractPathElementGraphics(originDocument, pdPage, pageNumber, cleanRulings, stripper.getTextPositionSequences(),

-                                                                              false);
+                                                                                    false);

            pdfImages.computeIfAbsent(pageNumber, x -> new ArrayList<>())
                    .addAll(graphics.stream()
-                                    .map(g -> new ClassifiedImage(new Rectangle2D.Double(g.x1, g.y1, g.width(), g.height()), ImageType.GRAPHIC, false, stripper.getPageNumber()))
+                                    .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());
                case DOCUMINE_OLD -> docuMineBlockificationService.blockify(words, cleanRulings);
-                case DOCUMINE, REDACT_MANAGER, REDACT_MANAGER_PARAGRAPH_DEBUG ->
+                case DOCUMINE, REDACT_MANAGER, REDACT_MANAGER_PARAGRAPH_DEBUG, REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH ->
                        docstrumBlockificationService.blockify(words, cleanRulings, true, classificationDocument.getVisualizations(), layoutParsingType);
                case CLARIFYND, CLARIFYND_PARAGRAPH_DEBUG ->
                        docstrumBlockificationService.blockify(words, cleanRulings, false, classificationDocument.getVisualizations(), layoutParsingType);
@@ -332,7 +333,7 @@ public class LayoutParsingPipeline {
            classificationPage.setPageWidth(cropbox.getWidth());
            classificationPage.setPageHeight(cropbox.getHeight());

-            if(layoutParsingType != LayoutParsingType.REDACT_MANAGER_OLD && layoutParsingType != LayoutParsingType.DOCUMINE_OLD) {
+            if (layoutParsingType != LayoutParsingType.REDACT_MANAGER_OLD && layoutParsingType != LayoutParsingType.DOCUMINE_OLD) {
                List<OutlineObject> outlineObjects = classificationDocument.getOutlineObjectTree().getOutlineObjectsPerPage().getOrDefault(pageNumber - 1, new ArrayList<>());

                OutlineObject notFoundOutlineObject = null;
@@ -381,7 +382,7 @@ public class LayoutParsingPipeline {
        }
        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);
@@ -5,6 +5,9 @@ import org.springframework.context.annotation.Bean;
 import org.springframework.context.annotation.ComponentScan;
 import org.springframework.context.annotation.Configuration;

+import com.google.common.base.Strings;
+import com.knecon.fforesight.service.viewerdoc.service.IViewerDocumentService;
+import com.knecon.fforesight.service.viewerdoc.service.pdftron.PDFTronViewerDocumentService;
 import com.knecon.fforesight.service.viewerdoc.service.ViewerDocumentService;

 import io.micrometer.observation.ObservationRegistry;
@@ -13,12 +16,16 @@ import io.micrometer.observation.ObservationRegistry;
@ComponentScan
 public class LayoutParsingServiceProcessorConfiguration {

-
    @Bean
    @Autowired
-    public ViewerDocumentService viewerDocumentService(ObservationRegistry registry) {
+    public IViewerDocumentService viewerDocumentService(ObservationRegistry registry, LayoutparserSettings settings) {
+
+        if (!Strings.isNullOrEmpty(settings.getPdftronLicense())) {
+            return new PDFTronViewerDocumentService(registry);
+        } else {
+            return new ViewerDocumentService(registry);
+        }

-        return new ViewerDocumentService(registry);
    }

 }
@@ -17,4 +17,5 @@ public class LayoutparserSettings {

    boolean debug;
    LayoutParsingType layoutParsingTypeOverride;
+    String pdftronLicense;
 }
@@ -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;

@@ -43,6 +43,11 @@ public class Document extends AbstractSemanticNode {
    }


+    /**
+     * Gets the sections of the document as a list.
+     *
+     * @return A list of all sections within the document.
+     */
    public List<Section> getAllSections() {

        return streamAllSubNodesOfType(NodeType.SECTION).map(node -> (Section) node)
@@ -50,6 +55,34 @@ public class Document extends AbstractSemanticNode {
    }


+    /**
+     * Gets the main sections of the document as a list.
+     *
+     * @return A list of main sections within the document
+     * @deprecated This method is marked for removal.
+     *  Use {@link #streamChildrenOfType(NodeType)} instead,
+     *  or {@link #getChildrenOfTypeSectionOrSuperSection()} which returns children of type SECTION as well as SUPER_SECTION.
+     */
+    @Deprecated(forRemoval = true)
+    public List<Section> getMainSections() {
+
+        return streamChildrenOfType(NodeType.SECTION).map(node -> (Section) node)
+                .collect(Collectors.toList());
+    }
+
+
+    /**
+     * 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)
@@ -28,6 +28,8 @@ public class Image extends AbstractSemanticNode {

    String id;

+    String representationHash;
+
    ImageType imageType;
    boolean transparent;
    Rectangle2D position;
@@ -24,5 +24,7 @@ public class ClassifiedImage {
    private boolean hasTransparency;
    @NonNull
    private int page;
+    @NonNull
+    private String representation;

 }
@@ -29,7 +29,7 @@ public class TOCEnrichmentService {
        TableOfContents toc = document.getTableOfContents();
        Iterator<TableOfContentItem> iterator = toc.iterator();
        TableOfContentItem currentTOCItem = null;
-        if(iterator.hasNext()) {
+        if (iterator.hasNext()) {
            currentTOCItem = iterator.next();
        }
        List<AbstractPageBlock> startBlocks = new ArrayList<>();
@@ -79,7 +79,7 @@ public class TOCEnrichmentService {
                    currentTOCItem.getSectionBlocks().add(current);
                    currentPageTOCItems.add(currentTOCItem);

-                    if(iterator.hasNext()) {
+                    if (iterator.hasNext()) {
                        currentTOCItem = iterator.next();
                    }
                } else if (!foundFirstHeadline) {
@@ -149,9 +149,9 @@ public class TOCEnrichmentService {
                    }
                }
                if (!image.isAppendedToSection()) {
-                    log.debug("Image uses first paragraph");
+                    log.debug("Image uses last found section");
                    if (!lastFoundTOCItems.isEmpty()) {
-                        lastFoundTOCItems.get(0).getImages().add(image);
+                        lastFoundTOCItems.get(lastFoundTOCItems.size() - 1).getImages().add(image);
                    } else {
                        startImages.add(image);
                    }
@@ -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;
@@ -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);
            }
@@ -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;
@@ -42,7 +42,9 @@ public class SimplifiedSectionTextService {

        return SimplifiedSectionText.builder()
                .sectionNumber(section.getTreeId()
-                                       .get(0))
+                                       .stream()
+                                       .map(String::valueOf)
+                                       .collect(Collectors.joining(".")))
                .text(section.getTextBlock().getSearchText())
                .build();
    }
@@ -12,6 +12,7 @@ import java.util.function.Function;
 import org.apache.commons.lang3.StringUtils;
 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;
@@ -204,21 +205,28 @@ public class BlockificationPostprocessingService {
            return false;
        }
        if (minDistance == distanceToDirectMatch) {
-            directMatch.setClassification(headlineType);
+            setClassificationAndAddOutlineEngine(directMatch, headlineType);
        } else if (minDistance == distanceToSplitCandidate) {
            SplitBlockResult splitBlockResult = splitBlock(classificationPage, splitCandidate, context.sectionIdentifier, outlineObject.getTitle());
            if (splitBlockResult.modifiedBlockToSplit) {
-                splitCandidate.setClassification(headlineType);
+                setClassificationAndAddOutlineEngine(splitCandidate, headlineType);
            }
            splitBlockResult.otherBlocks.forEach(other -> other.setClassification(null));
        } else {
            var merged = mergeBlocks(classificationPage, bestMergeCandidateCombination);
-            merged.setClassification(headlineType);
+            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<>();
@@ -26,6 +26,7 @@ import lombok.RequiredArgsConstructor;
@RequiredArgsConstructor
 public class DocstrumBlockificationService {

+    public static final float Y_THRESHOLD = 5f;
    private final DocstrumSegmentationService docstrumSegmentationService;

    static final float THRESHOLD = 1f;
@@ -58,8 +59,10 @@ public class DocstrumBlockificationService {

        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) {
@@ -105,7 +108,7 @@ 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();
@@ -137,7 +140,8 @@ public class DocstrumBlockificationService {
                }

                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;
                }

@@ -160,7 +164,7 @@ public class DocstrumBlockificationService {
            previous = current;
        }

-        mergeIntersectingBlocks(page, usedRulings, 0, 6.5f);
+        mergeIntersectingBlocks(page, usedRulings, 0, Y_THRESHOLD);
    }


@@ -21,33 +21,31 @@ import lombok.extern.slf4j.Slf4j;
@RequiredArgsConstructor
 public class ClarifyndClassificationService {

-    private final HeadlineClassificationService headlineClassificationService;
-
    public void classifyDocument(ClassificationDocument document) {

        List<Float> headlineFontSizes = document.getFontSizeCounter().getHighterThanMostPopular();

        log.debug("Document FontSize counters are: {}", document.getFontSizeCounter().getCountPerValue());

-        headlineClassificationService.resetContext();
+        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<Float> 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<Float> headlineFontSizes) {

        var bodyTextFrame = page.getBodyTextFrame();

@@ -24,10 +24,9 @@ import lombok.extern.slf4j.Slf4j;
@RequiredArgsConstructor
 public class DocuMineClassificationService {

-    private final HeadlineClassificationService headlineClassificationService;
-    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) {
@@ -36,32 +35,39 @@ public class DocuMineClassificationService {

        log.debug("Document FontSize counters are: {}", document.getFontSizeCounter().getCountPerValue());

-        headlineClassificationService.resetContext();
+        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<Float> 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<Float> 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 (textBlock.getClassification() != null && textBlock.getClassification().isHeadline()) {
            headlineClassificationService.setLastHeadlineFromOutline(textBlock);
@@ -71,14 +77,10 @@ public class DocuMineClassificationService {
            textBlock.setClassification(PageBlockType.OTHER);
            return;
        }
-        if (MarkedContentUtils.intersects(textBlock, page.getMarkedContentBboxPerType(), MarkedContentUtils.HEADER)
-            || (PositionUtils.isOverBodyTextFrame(bodyTextFrame,
-                                                  textBlock,
-                                                  page.getRotation()) && (document.getFontSizeCounter().getMostPopular()
-                                                                          == null
-                                                                          || textBlock.getHighestFontSize()
-                                                                             <= document.getFontSizeCounter()
-                                                                                     .getMostPopular()))) {
+        if (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)
@@ -102,19 +104,25 @@ public class DocuMineClassificationService {
                       || 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("APPENDIX") //
+                       || textBlock.toString().startsWith("FIGURE") //
+                       || textBlock.toString().startsWith("Continued TABLE") //
                       || textBlock.toString().startsWith("TABLE"))
                   && !textBlock.toString().endsWith(":")
-                   && matcher2.reset().find()) {
+                   && 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()) {
+        } 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);
@@ -2,15 +2,12 @@ package com.knecon.fforesight.service.layoutparser.processor.services.classifica

 import static com.knecon.fforesight.service.layoutparser.processor.model.PageBlockType.getHeadlineNumber;

-import org.springframework.stereotype.Service;
-
 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;

-@Service
@Getter
@Setter
 public class HeadlineClassificationService {
@@ -19,13 +16,6 @@ public class HeadlineClassificationService {
    PageBlockType originalClassifiedBlockType;
    TextPageBlock lastHeadlineFromOutline;

-    public void resetContext() {
-        setLastHeadline(null);
-        setOriginalClassifiedBlockType(null);
-        setLastHeadlineFromOutline(null);
-    }
-
-
    public void setLastHeadlineFromOutline(TextPageBlock lastHeadlineFromOutline) {

        this.lastHeadlineFromOutline = lastHeadlineFromOutline;
@@ -22,8 +22,6 @@ import lombok.extern.slf4j.Slf4j;
@RequiredArgsConstructor
 public class RedactManagerClassificationService {

-    private final HeadlineClassificationService headlineClassificationService;
-

    public void classifyDocument(ClassificationDocument document) {

@@ -31,25 +29,25 @@ public class RedactManagerClassificationService {

        log.debug("Document FontSize counters are: {}", document.getFontSizeCounter().getCountPerValue());

-        headlineClassificationService.resetContext();
+        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<Float> 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<Float> headlineFontSizes) {

        var bodyTextFrame = page.getBodyTextFrame();

@@ -45,8 +45,10 @@ 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) {
@@ -122,6 +124,7 @@ public class DocumentGraphFactory {
        List<Integer> treeId = context.documentTree.createNewChildEntryAndReturnId(parentNode, node);
        node.setLeafTextBlock(textBlock);
        node.setTreeId(treeId);
+        node.getEngines().addAll(originalTextBlock.getEngines());
    }


@@ -144,6 +147,7 @@ public class DocumentGraphFactory {
                .position(position)
                .transparent(image.isHasTransparency())
                .page(page)
+                .representationHash(image.getRepresentation())
                .documentTree(context.getDocumentTree())
                .build();
        page.getMainBody().add(imageNode);
@@ -14,11 +14,11 @@ 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.AbstractSemanticNode;
 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;
@@ -55,7 +55,8 @@ public class SectionNodeFactory {
                .collect(groupingBy(AbstractPageBlock::getPage));

        AbstractSemanticNode section;
-        if (isLeaf) {
+        boolean containsTablesAndTextBlocks = containsTablesAndTextBlocks(pageBlocks);
+        if (isLeaf && !containsTablesAndTextBlocks) {
            section = Section.builder().documentTree(context.getDocumentTree()).build();
        } else {
            section = SuperSection.builder().documentTree(context.getDocumentTree()).build();
@@ -68,7 +69,7 @@ public class SectionNodeFactory {
        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,
@@ -132,7 +133,7 @@ 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);
@@ -24,6 +24,7 @@ public class PropertiesMapper {
        properties.put(DocumentStructure.ImageProperties.TRANSPARENT, String.valueOf(image.isTransparent()));
        properties.put(DocumentStructure.ImageProperties.POSITION, toString(image.getPosition()));
        properties.put(DocumentStructure.ImageProperties.ID, image.getId());
+        properties.put(DocumentStructure.ImageProperties.REPRESENTATION_HASH, image.getRepresentationHash());
        return properties;
    }

@@ -99,6 +99,7 @@ public class TaasDocumentDataMapper {
        Page page = semanticNode.getFirstPage();
        Rectangle2D bBox = semanticNode.getBBox().get(page);
        return StructureObject.builder()
+                .treeId(semanticNode.getTreeId())
                .structureObjectNumber(structureObjectNumber)
                .boundingBox(toFloatArray(bBox))
                .stringOffset(semanticNode.getBoundary().start())
@@ -114,6 +115,7 @@ public class TaasDocumentDataMapper {
        Page page = table.getFirstPage();
        Rectangle2D bBox = table.getBBox().get(page);
        return StructureObject.builder()
+                .treeId(table.getTreeId())
                .structureObjectNumber(structureObjectNumber)
                .boundingBox(toFloatArray(bBox))
                .stringOffset(table.getBoundary().start())
@@ -9,6 +9,7 @@ import java.io.File;
 import java.util.ArrayList;
 import java.util.Collections;
 import java.util.Comparator;
+import java.util.HashMap;
 import java.util.List;
 import java.util.Map;
 import java.util.Optional;
@@ -23,6 +24,7 @@ import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Do
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.DuplicatedParagraph;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Page;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.SemanticNode;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.SuperSection;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Table;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.TableCell;
 import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
@@ -33,7 +35,7 @@ import com.knecon.fforesight.service.viewerdoc.model.LayoutGrid;
 import com.knecon.fforesight.service.viewerdoc.model.PlacedText;
 import com.knecon.fforesight.service.viewerdoc.model.Standard14EmbeddableFont;
 import com.knecon.fforesight.service.viewerdoc.model.Visualizations;
-import com.knecon.fforesight.service.viewerdoc.service.ViewerDocumentService;
+import com.knecon.fforesight.service.viewerdoc.service.IViewerDocumentService;

 import io.micrometer.observation.annotation.Observed;
 import lombok.AccessLevel;
@@ -46,7 +48,7 @@ import lombok.experimental.FieldDefaults;
@FieldDefaults(level = AccessLevel.PRIVATE, makeFinal = true)
 public class LayoutGridService {

-    ViewerDocumentService viewerDocumentService;
+    IViewerDocumentService viewerDocumentService;

    static float FONT_SIZE = 10f;
    static float LINE_WIDTH = 1f;
@@ -62,6 +64,12 @@ public class LayoutGridService {
    static Color HEADER_COLOR = new Color(171, 131, 6);
    static Color IMAGE_COLOR = new Color(253, 63, 146);

+    private record RectangleIdentifier(List<Integer> treeId, Integer pageNumber) {
+
+    }
+
+    HashMap<RectangleIdentifier, Rectangle2D> rectangleMap = new HashMap<>();
+

    @SneakyThrows
    @Observed(name = "ViewerDocumentService", contextualName = "create-viewer-document")
@@ -165,8 +173,10 @@ public class LayoutGridService {
            List<Double> ys = yStream.collect(Collectors.toList());
            ys.remove(0);

-            Rectangle2D tableBBox = table.getBBox().get(page);
-            List<ColoredLine> coloredLines = layoutGrid.getVisualizationsPerPages().get(page.getNumber() - 1).getColoredLines();
+            Rectangle2D tableBBox = table.getBBox()
+                    .get(page);
+            List<ColoredLine> coloredLines = layoutGrid.getVisualizationsPerPages()
+                    .get(page.getNumber() - 1).getColoredLines();
            xs.forEach(x -> {
                Line2D line = new Line2D.Double(new Point2D.Double(x, tableBBox.getMaxY()), new Point2D.Double(x, tableBBox.getMinY()));
                coloredLines.add(new ColoredLine(line, INNER_LINES_COLOR, LINE_WIDTH));
@@ -190,25 +200,20 @@ public class LayoutGridService {
    private void addSection(SemanticNode semanticNode, LayoutGrid layoutGrid, Color color) {

        Map<Page, Rectangle2D> bBoxMap = semanticNode.getBBox();
+
        List<SemanticNode> subSections = semanticNode.streamAllSubNodesOfType(NodeType.SECTION)
                .toList();
+        Integer maxChildDepth = subSections.stream()
+                .map(node -> node.getTreeId().size())
+                .max(Integer::compareTo)
+                .orElse(semanticNode.getTreeId().size());
+        int ownDepth = semanticNode.getTreeId().size();
+
        Page firstPage = semanticNode.getFirstPage();
        String treeIdString = buildTreeIdString(semanticNode);
-        if (!subSections.isEmpty()) {
-            addPlacedText(firstPage, bBoxMap.get(firstPage), treeIdString, layoutGrid);
-        } else {
-            bBoxMap.forEach(((page, textBBox) -> addPlacedText(page, textBBox, treeIdString, layoutGrid)));
-        }
+
        if (bBoxMap.values().size() == 1) {
-            Rectangle2D r = RectangleTransformations.pad(bBoxMap.get(firstPage), LINE_WIDTH, LINE_WIDTH);
-            List<ColoredLine> coloredLines = layoutGrid.getVisualizationsPerPages().get(firstPage.getNumber() - 1).getColoredLines();
-            List<Line2D> lines = createLinesFromRectangle(r, firstPage.getRotation());
-            // add string to top line
-            var firstLine = lines.remove(0);
-            coloredLines.add(new ColoredLine(firstLine, color, LINE_WIDTH));
-            for (Line2D line : lines) {
-                coloredLines.add(new ColoredLine(line, color, LINE_WIDTH));
-            }
+            handleSinglePage(semanticNode, layoutGrid, color, firstPage, bBoxMap.get(firstPage), treeIdString, maxChildDepth, ownDepth);
            return;
        }
        List<Page> pagesInOrder = bBoxMap.keySet()
@@ -216,99 +221,203 @@ public class LayoutGridService {
                .sorted(Comparator.comparingInt(Page::getNumber))
                .collect(Collectors.toList());
        pagesInOrder.remove(0);
-        addLinesForFirstPageOfSection(semanticNode, color, firstPage, layoutGrid);
-        var lastPage = pagesInOrder.remove(pagesInOrder.size() - 1);
-        addLinesForLastPageOfSection(semanticNode, color, lastPage, layoutGrid);
-        for (Page middlePage : pagesInOrder) {
-            addLinesForMiddlePageOfSection(semanticNode, color, middlePage, layoutGrid);
+        handleFirstPageOfSection(semanticNode, color, firstPage, bBoxMap.get(firstPage), treeIdString, layoutGrid, maxChildDepth, ownDepth);
+        if (semanticNode instanceof SuperSection) {
+            return;
        }
+        for (Page middlePage : pagesInOrder.subList(0, pagesInOrder.size() - 1)) {
+            handleForMiddlePageOfSection(semanticNode, color, middlePage, bBoxMap.get(middlePage), treeIdString, layoutGrid, maxChildDepth, ownDepth);
+        }
+        var lastPage = pagesInOrder.remove(pagesInOrder.size() - 1);
+        handleLastPageOfSection(semanticNode, color, lastPage, bBoxMap.get(lastPage), treeIdString, layoutGrid, maxChildDepth, ownDepth);
    }


    @SneakyThrows
-    private void addPlacedText(Page page, Rectangle2D textBBox, String s, LayoutGrid layoutGrid) {
+    private void addPlacedText(Page page, Rectangle2D textBBox, Rectangle2D highestParentRect, String s, LayoutGrid layoutGrid, Integer maxChildDepth) {

        // translates text, such that its right edge is a bit to the left of the drawn box
-        float translationAmount = ((FONT.getStringWidth(s) / 1000) * FONT_SIZE + (2 * LINE_WIDTH) + 4);
+        float translationAmount = ((FONT.getStringWidth(s) / 1000) * FONT_SIZE + LINE_WIDTH + 2 * maxChildDepth);

        Point2D upperLeftCorner;
        Point2D translationVector;
        switch (page.getRotation()) {
            case 90 -> {
-                upperLeftCorner = new Point2D.Double(textBBox.getMinX(), textBBox.getMinY());
+                if (highestParentRect != null) {
+                    upperLeftCorner = new Point2D.Double(highestParentRect.getMinX(), textBBox.getMinY());
+                } else {
+                    upperLeftCorner = new Point2D.Double(textBBox.getMinX(), textBBox.getMinY());
+                }
                translationVector = new Point2D.Double(FONT_SIZE, -translationAmount);
            }
            case 180 -> {
-                upperLeftCorner = new Point2D.Double(textBBox.getMaxX(), textBBox.getMinY());
+                if (highestParentRect != null) {
+                    upperLeftCorner = new Point2D.Double(highestParentRect.getMaxX(), textBBox.getMinY());
+                } else {
+                    upperLeftCorner = new Point2D.Double(textBBox.getMaxX(), textBBox.getMinY());
+                }
                translationVector = new Point2D.Double(translationAmount, FONT_SIZE);
            }
            case 270 -> {
-                upperLeftCorner = new Point2D.Double(textBBox.getMaxX(), textBBox.getMaxY());
+
+                if (highestParentRect != null) {
+                    upperLeftCorner = new Point2D.Double(highestParentRect.getMaxX(), textBBox.getMaxY());
+                } else {
+                    upperLeftCorner = new Point2D.Double(textBBox.getMaxX(), textBBox.getMaxY());
+                }
                translationVector = new Point2D.Double(-FONT_SIZE, translationAmount);
            }
            default -> {
-                upperLeftCorner = new Point2D.Double(textBBox.getMinX(), textBBox.getMaxY());
+
+                if (highestParentRect != null) {
+                    upperLeftCorner = new Point2D.Double(highestParentRect.getMinX(), textBBox.getMaxY());
+                } else {
+                    upperLeftCorner = new Point2D.Double(textBBox.getMinX(), textBBox.getMaxY());
+                }
                translationVector = new Point2D.Double(-translationAmount, -FONT_SIZE);
            }
        }

        upperLeftCorner = add(upperLeftCorner, translationVector);

-        var placedTexts = layoutGrid.getVisualizationsPerPages().get(page.getNumber() - 1).getPlacedTexts();
-        placedTexts.add(PlacedText.textFacingUp(s, upperLeftCorner, FONT_SIZE, Color.BLACK, FONT));
-    }
+        List<PlacedText> placedTexts = layoutGrid.getVisualizationsPerPages()
+                .get(page.getNumber() - 1).getPlacedTexts();

+        PlacedText newText = PlacedText.textFacingUp(s, upperLeftCorner, FONT_SIZE, Color.BLACK, FONT);

-    private void addLinesForMiddlePageOfSection(SemanticNode semanticNode, Color color, Page middlePage, LayoutGrid layoutGrid) {
+        Optional<PlacedText> conflictingText = placedTexts.stream()
+                .filter(pt -> Math.abs(pt.lineStart().getY() - newText.lineStart().getY()) <= FONT_SIZE)
+                .findFirst();

-        List<ColoredLine> coloredLines = layoutGrid.getVisualizationsPerPages().get(middlePage.getNumber() - 1).getColoredLines();
-        Rectangle2D r = RectangleTransformations.pad(semanticNode.getBBox().get(middlePage), LINE_WIDTH, LINE_WIDTH);
-        var midPageLines = createLinesFromRectangle(r, middlePage.getRotation());
-        // remove top line
-        midPageLines.remove(0);
-        // remove top line
-        midPageLines.remove(1);
-        // add string to left line
-        var leftLine = midPageLines.remove(1);
-        coloredLines.add(new ColoredLine(leftLine, color, LINE_WIDTH));
-        for (Line2D line : midPageLines) {
-            coloredLines.add(new ColoredLine(line, color, LINE_WIDTH));
+        if (conflictingText.isPresent()) {
+            PlacedText existingText = conflictingText.get();
+            if (newText.text().length() > existingText.text().length()) {
+                placedTexts.remove(existingText);
+                placedTexts.add(newText);
+            }
+        } else {
+            placedTexts.add(newText);
        }
+
    }


-    private void addLinesForLastPageOfSection(SemanticNode semanticNode, Color color, Page lastPage, LayoutGrid layoutGrid) {
+    private void handleSinglePage(SemanticNode semanticNode,
+                                  LayoutGrid layoutGrid,
+                                  Color color,
+                                  Page page,
+                                  Rectangle2D rectangle2D,
+                                  String treeIdString,
+                                  Integer maxChildDepth,
+                                  Integer ownDepth) {

-        List<ColoredLine> coloredLines = layoutGrid.getVisualizationsPerPages().get(lastPage.getNumber() - 1).getColoredLines();
-        Rectangle2D r = RectangleTransformations.pad(semanticNode.getBBox().get(lastPage), LINE_WIDTH, LINE_WIDTH);
-        var lastPageLines = createLinesFromRectangle(r, lastPage.getRotation());
-        // remove top line
-        lastPageLines.remove(0);
-        // add string to left line
-        var leftLine = lastPageLines.remove(2);
-        coloredLines.add(new ColoredLine(leftLine, color, LINE_WIDTH));
-        for (Line2D line : lastPageLines) {
-            coloredLines.add(new ColoredLine(line, color, LINE_WIDTH));
-        }
-    }
-
-
-    private void addLinesForFirstPageOfSection(SemanticNode semanticNode, Color color, Page firstPage, LayoutGrid layoutGrid) {
-
-        List<ColoredLine> coloredLines = layoutGrid.getVisualizationsPerPages().get(firstPage.getNumber() - 1).getColoredLines();
-        Rectangle2D r = RectangleTransformations.pad(semanticNode.getBBox().get(firstPage), LINE_WIDTH, LINE_WIDTH);
-        var firstPageLines = createLinesFromRectangle(r, firstPage.getRotation());
-        // remove bottom line
-        firstPageLines.remove(2);
+        RectangleAndLinesResult result = createLinesAndPlaceText(semanticNode, page, rectangle2D, treeIdString, layoutGrid, maxChildDepth, ownDepth);
        // add string to top line
-        var firstLine = firstPageLines.remove(0);
-        coloredLines.add(new ColoredLine(firstLine, color, LINE_WIDTH));
-        for (Line2D line : firstPageLines) {
-            coloredLines.add(new ColoredLine(line, color, LINE_WIDTH));
+        var firstLine = result.pageLines().remove(0);
+        result.coloredLines().add(new ColoredLine(firstLine, color, LINE_WIDTH));
+        for (Line2D line : result.pageLines()) {
+            result.coloredLines().add(new ColoredLine(line, color, LINE_WIDTH));
        }
    }


+    private void handleFirstPageOfSection(SemanticNode semanticNode,
+                                          Color color,
+                                          Page firstPage,
+                                          Rectangle2D rectangle2D,
+                                          String treeIdString,
+                                          LayoutGrid layoutGrid,
+                                          Integer maxChildDepth,
+                                          Integer ownDepth) {
+
+        RectangleAndLinesResult result = createLinesAndPlaceText(semanticNode, firstPage, rectangle2D, treeIdString, layoutGrid, maxChildDepth, ownDepth);
+        // remove bottom line
+        result.pageLines().remove(2);
+        // add string to top line
+        var firstLine = result.pageLines().remove(0);
+        result.coloredLines().add(new ColoredLine(firstLine, color, LINE_WIDTH));
+        for (Line2D line : result.pageLines()) {
+            result.coloredLines().add(new ColoredLine(line, color, LINE_WIDTH));
+        }
+    }
+
+
+    private void handleForMiddlePageOfSection(SemanticNode semanticNode,
+                                              Color color,
+                                              Page middlePage,
+                                              Rectangle2D rectangle2D,
+                                              String treeIdString,
+                                              LayoutGrid layoutGrid,
+                                              Integer maxChildDepth,
+                                              Integer ownDepth) {
+
+        RectangleAndLinesResult result = createLinesAndPlaceText(semanticNode, middlePage, rectangle2D, treeIdString, layoutGrid, maxChildDepth, ownDepth);
+        // remove top line
+        result.pageLines().remove(0);
+        // remove bottom line
+        result.pageLines().remove(1);
+        // add string to left line
+        var leftLine = result.pageLines().remove(1);
+        result.coloredLines().add(new ColoredLine(leftLine, color, LINE_WIDTH));
+        for (Line2D line : result.pageLines()) {
+            result.coloredLines().add(new ColoredLine(line, color, LINE_WIDTH));
+        }
+    }
+
+
+    private void handleLastPageOfSection(SemanticNode semanticNode,
+                                         Color color,
+                                         Page lastPage,
+                                         Rectangle2D rectangle2D,
+                                         String treeIdString,
+                                         LayoutGrid layoutGrid,
+                                         Integer maxChildDepth,
+                                         Integer ownDepth) {
+
+        RectangleAndLinesResult result = createLinesAndPlaceText(semanticNode, lastPage, rectangle2D, treeIdString, layoutGrid, maxChildDepth, ownDepth);
+        // remove top line
+        result.pageLines().remove(0);
+        // add string to left line
+        var leftLine = result.pageLines().remove(2);
+        result.coloredLines().add(new ColoredLine(leftLine, color, LINE_WIDTH));
+        for (Line2D line : result.pageLines()) {
+            result.coloredLines().add(new ColoredLine(line, color, LINE_WIDTH));
+        }
+    }
+
+
+    private RectangleAndLinesResult createLinesAndPlaceText(SemanticNode semanticNode,
+                                                            Page page,
+                                                            Rectangle2D rectangle2D,
+                                                            String treeIdString,
+                                                            LayoutGrid layoutGrid,
+                                                            Integer maxChildDepth,
+                                                            Integer ownDepth) {
+
+        List<ColoredLine> coloredLines = layoutGrid.getVisualizationsPerPages()
+                .get(page.getNumber() - 1).getColoredLines();
+        int lineWidthModifier = maxChildDepth - ownDepth;
+        Rectangle2D r = RectangleTransformations.pad(semanticNode.getBBox()
+                                                             .get(page), LINE_WIDTH * (1 + lineWidthModifier), LINE_WIDTH * (1 + lineWidthModifier));
+        var lastPageLines = createLinesFromRectangle(r, page.getRotation());
+
+        SemanticNode highestParent = semanticNode.getHighestParent();
+        Rectangle2D highestParentRect = rectangleMap.get(new RectangleIdentifier(highestParent.getTreeId(), page.getNumber()));
+        addPlacedText(page, rectangle2D, highestParentRect, treeIdString, layoutGrid, maxChildDepth);
+
+        if (semanticNode instanceof SuperSection) {
+            rectangleMap.put(new RectangleIdentifier(semanticNode.getTreeId(), page.getNumber()), r);
+        }
+
+        return new RectangleAndLinesResult(coloredLines, r, lastPageLines);
+    }
+
+
+    private record RectangleAndLinesResult(List<ColoredLine> coloredLines, Rectangle2D rectangle, List<Line2D> pageLines) {
+
+    }
+
+
    private String buildTreeIdString(SemanticNode semanticNode) {

        return semanticNode.getTreeId()
@@ -366,7 +475,10 @@ public class LayoutGridService {

    private static boolean isNotSectionOrTableCellOrDocument(SemanticNode semanticNode) {

-        return !(semanticNode.getType().equals(NodeType.DOCUMENT) || semanticNode.getType().equals(NodeType.SECTION) || semanticNode.getType().equals(NodeType.TABLE_CELL));
+        return !(semanticNode.getType().equals(NodeType.DOCUMENT)
+                 || semanticNode.getType().equals(NodeType.SECTION)
+                 || semanticNode.getType().equals(NodeType.SUPER_SECTION)
+                 || semanticNode.getType().equals(NodeType.TABLE_CELL));
    }


@@ -374,9 +486,7 @@ public class LayoutGridService {

        semanticNode.getBBox()
                .forEach((page, textBBox) -> layoutGrid.getVisualizationsPerPages()
-                        .get(page.getNumber() - 1)
-                        .getColoredRectangles()
-                        .add(new ColoredRectangle(textBBox, color, LINE_WIDTH)));
+                        .get(page.getNumber() - 1).getColoredRectangles().add(new ColoredRectangle(textBBox, color, LINE_WIDTH)));
    }


@@ -37,6 +37,7 @@ dependencies {
    implementation("com.amazonaws:aws-java-sdk-s3:1.12.536")
    implementation("org.springframework.cloud:spring-cloud-starter-openfeign:4.0.4")
    implementation("net.logstash.logback:logstash-logback-encoder:7.4")
+    implementation("com.pdftron:PDFNet:10.5.0")

    // for integration testing only
    testImplementation(project(":viewer-doc-processor"))
@@ -0,0 +1,45 @@
+package com.knecon.fforesight.service.layoutparser.server;
+
+import org.springframework.stereotype.Component;
+
+import com.google.common.base.Strings;
+import com.knecon.fforesight.service.layoutparser.processor.LayoutparserSettings;
+import com.pdftron.pdf.PDFNet;
+
+import jakarta.annotation.PostConstruct;
+import jakarta.annotation.PreDestroy;
+import lombok.RequiredArgsConstructor;
+import lombok.SneakyThrows;
+import lombok.extern.slf4j.Slf4j;
+
+@Slf4j
+@Component
+@RequiredArgsConstructor
+public class PDFNetInitializer {
+
+    private final LayoutparserSettings settings;
+
+
+    @SneakyThrows
+    @PostConstruct
+    // Do not change back to application runner, if it is application runner it takes messages from the queue before PDFNet is initialized, that leads to UnsatisfiedLinkError.
+    public void init() {
+
+        if (Strings.isNullOrEmpty(settings.getPdftronLicense())) {
+            return;
+        }
+        log.info("Initializing Native Libraries");
+        log.info("Setting pdftron license: {}", settings.getPdftronLicense());
+        PDFNet.setTempPath("/tmp/pdftron");
+        PDFNet.initialize(settings.getPdftronLicense());
+
+    }
+
+
+    @PreDestroy
+    public void terminate() {
+
+        PDFNet.terminate();
+    }
+
+}
@@ -70,7 +70,7 @@ public class LayoutparserEnd2EndTest extends AbstractTest {
            file = new File(filePath);
        }

-        LayoutParsingRequest layoutParsingRequest = buildDefaultLayoutParsingRequest(fileName, LayoutParsingType.REDACT_MANAGER, true);
+        LayoutParsingRequest layoutParsingRequest = buildDefaultLayoutParsingRequest(fileName, LayoutParsingType.REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH, true);
        prepareStorage(layoutParsingRequest, file);

        LayoutParsingFinishedEvent finishedEvent = layoutParsingPipeline.parseLayoutAndSaveFilesToStorage(layoutParsingRequest);
@@ -0,0 +1,231 @@
+package com.knecon.fforesight.service.layoutparser.server;
+
+import static org.junit.jupiter.api.Assertions.assertEquals;
+import static org.junit.jupiter.api.Assertions.assertTrue;
+
+import java.io.File;
+import java.nio.file.Path;
+import java.util.Collection;
+import java.util.List;
+import java.util.Locale;
+import java.util.Map;
+import java.util.function.Predicate;
+
+import org.apache.commons.lang3.StringUtils;
+import org.junit.jupiter.api.Test;
+import org.springframework.beans.factory.annotation.Autowired;
+import org.springframework.core.io.ClassPathResource;
+
+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.LayoutParsingPipeline;
+import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationDocument;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Section;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.SemanticNode;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.SuperSection;
+import com.knecon.fforesight.service.layoutparser.processor.model.outline.OutlineObject;
+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.python_api.model.table.TableServiceResponse;
+import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.VisualLayoutParsingResponse;
+import com.knecon.fforesight.service.layoutparser.processor.services.factory.DocumentGraphFactory;
+import com.knecon.fforesight.service.layoutparser.processor.services.visualization.LayoutGridService;
+import com.knecon.fforesight.service.layoutparser.server.utils.AbstractTest;
+import com.knecon.fforesight.service.viewerdoc.service.ViewerDocumentService;
+
+import lombok.SneakyThrows;
+
+public class OutlineDetectionTest extends AbstractTest {
+
+    ViewerDocumentService viewerDocumentService = new ViewerDocumentService(null);
+    LayoutGridService layoutGridService = new LayoutGridService(viewerDocumentService);
+
+    @Autowired
+    protected LayoutParsingPipeline layoutParsingPipeline;
+
+
+    @Test
+    @SneakyThrows
+    public void testOutlinesToSections() {
+
+        String fileName = "files/new/crafted_outline_test_doc.pdf";
+
+        String tmpFileName = "/tmp/" + Path.of(fileName).getFileName() + "_VIEWER.pdf";
+
+        var documentFile = new ClassPathResource(fileName).getFile();
+
+        long start = System.currentTimeMillis();
+        ClassificationDocument classificationDocument = parseLayout(fileName, LayoutParsingType.REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH);
+
+        OutlineObjectTree outlineObjectTree = classificationDocument.getOutlineObjectTree();
+        assertEquals(outlineObjectTree.getRootNodes().size(), 8);
+        assertEquals(outlineObjectTree.getOutlineObjectsPerPage()
+                             .get(1).size(), 1);
+        assertEquals(outlineObjectTree.getOutlineObjectsPerPage()
+                             .get(3).size(), 1);
+        assertEquals(outlineObjectTree.getOutlineObjectsPerPage()
+                             .get(4).size(), 1);
+        assertEquals(outlineObjectTree.getOutlineObjectsPerPage()
+                             .get(5).size(), 1);
+        assertEquals(outlineObjectTree.getOutlineObjectsPerPage()
+                             .get(6).size(), 2);
+        assertEquals(outlineObjectTree.getOutlineObjectsPerPage()
+                             .get(7).size(), 3);
+        assertEquals(outlineObjectTree.getOutlineObjectsPerPage()
+                             .get(8).size(), 2);
+        assertEquals(outlineObjectTree.getOutlineObjectsPerPage()
+                             .get(10).size(), 1);
+        assertEquals(outlineObjectTree.getOutlineObjectsPerPage()
+                             .get(11).size(), 4);
+        assertEquals(outlineObjectTree.getOutlineObjectsPerPage()
+                             .get(12).size(), 1);
+        assertEquals(outlineObjectTree.getOutlineObjectsPerPage()
+                             .get(13).size(), 2);
+        assertTrue(outlineObjectTree.getOutlineObjectsPerPage().values()
+                           .stream()
+                           .flatMap(Collection::stream)
+                           .allMatch(OutlineObject::isFound));
+
+        TableOfContents tableOfContents = classificationDocument.getTableOfContents();
+
+        assertEquals(tableOfContents.getMainSections().size(), 9);
+        assertEquals(tableOfContents.getMainSections().subList(1, 9)
+                             .stream()
+                             .map(tableOfContentItem -> sanitizeString(tableOfContentItem.getHeadline().toString()))
+                             .toList(),
+                     outlineObjectTree.getRootNodes()
+                             .stream()
+                             .map(outlineObjectTreeNode -> sanitizeString(outlineObjectTreeNode.getOutlineObject().getTitle()))
+                             .toList());
+        assertEquals(tableOfContents.getMainSections()
+                             .get(5).getChildren().size(), 6);
+        assertEquals(tableOfContents.getMainSections()
+                             .get(7).getChildren().size(), 3);
+        assertEquals(tableOfContents.getMainSections()
+                             .get(8).getChildren().size(), 3);
+        assertEquals(tableOfContents.getMainSections()
+                             .get(8).getChildren()
+                             .get(2).getChildren().size(), 1);
+        assertEquals(tableOfContents.getMainSections()
+                             .get(8).getChildren()
+                             .get(2).getChildren()
+                             .get(0).getChildren().size(), 3);
+
+        assertEquals(tableOfContents.getMainSections()
+                             .get(0).getImages().size(), 1);
+        assertEquals(tableOfContents.getMainSections()
+                             .get(6).getImages().size(), 1);
+        assertEquals(tableOfContents.getMainSections()
+                             .get(8).getChildren()
+                             .get(2).getChildren()
+                             .get(0).getChildren()
+                             .get(2).getImages().size(), 1);
+
+        Document document = buildGraph(fileName, classificationDocument);
+
+        assertTrue(tableOfContents.getAllTableOfContentItems()
+                           .stream()
+                           .allMatch(tableOfContentItem -> tableOfContentItem.getSection() != null));
+        assertTrue(tableOfContents.getAllTableOfContentItems()
+                           .stream()
+                           .filter(tableOfContentItem -> tableOfContentItem.getChildren().isEmpty())
+                           .allMatch(tableOfContentItem -> tableOfContentItem.getSection() instanceof Section));
+        assertTrue(tableOfContents.getAllTableOfContentItems()
+                           .stream()
+                           .filter(tableOfContentItem -> !tableOfContentItem.getChildren().isEmpty())
+                           .allMatch(tableOfContentItem -> tableOfContentItem.getSection() instanceof SuperSection));
+
+        List<SemanticNode> childrenOfTypeSectionOrSuperSection = document.getChildrenOfTypeSectionOrSuperSection();
+
+        assertEquals(childrenOfTypeSectionOrSuperSection.size(), 9);
+        assertEquals(childrenOfTypeSectionOrSuperSection.subList(1, 9)
+                             .stream()
+                             .map(section -> sanitizeString(section.getHeadline().getLeafTextBlock().toString()))
+                             .toList(),
+                     outlineObjectTree.getRootNodes()
+                             .stream()
+                             .map(outlineObjectTreeNode -> sanitizeString(outlineObjectTreeNode.getOutlineObject().getTitle()))
+                             .toList());
+        Predicate<SemanticNode> isSectionOrSuperSection = semanticNode -> semanticNode instanceof Section || semanticNode instanceof SuperSection;
+        assertEquals(childrenOfTypeSectionOrSuperSection.get(5).streamChildren()
+                             .filter(isSectionOrSuperSection)
+                             .count(), 6 + 1); // 1 additional for main text of parent section
+        assertEquals(childrenOfTypeSectionOrSuperSection.get(7).streamChildren()
+                             .filter(isSectionOrSuperSection)
+                             .count(), 3 + 1);
+        assertEquals(childrenOfTypeSectionOrSuperSection.get(8).streamChildren()
+                             .filter(isSectionOrSuperSection)
+                             .count(), 3 + 1);
+        assertEquals(childrenOfTypeSectionOrSuperSection.get(8).streamChildren()
+                             .filter(isSectionOrSuperSection)
+                             .toList()
+                             .get(3).streamChildren()
+                             .filter(isSectionOrSuperSection)
+                             .count(), 1 + 1);
+        assertEquals(childrenOfTypeSectionOrSuperSection.get(8).streamChildren()
+                             .filter(isSectionOrSuperSection)
+                             .toList()
+                             .get(3).streamChildren()
+                             .filter(isSectionOrSuperSection)
+                             .toList()
+                             .get(1).streamChildren()
+                             .filter(isSectionOrSuperSection)
+                             .count(), 3 + 1);
+
+        List<List<Integer>> imageTreeIdList = document.streamAllImages()
+                .map(image -> image.getParent().getTreeId())
+                .toList();
+
+        assertEquals(imageTreeIdList.get(0), List.of(0));
+        assertEquals(imageTreeIdList.get(1), List.of(6));
+        assertEquals(imageTreeIdList.get(2), List.of(8, 4, 2, 4));
+
+        layoutGridService.addLayoutGrid(documentFile, document, new File(tmpFileName), true, false);
+        System.out.printf("Total time: %.2fs%n", ((float) (System.currentTimeMillis() - start)) / 1000);
+    }
+
+
+    private static String sanitizeString(String text) {
+
+        return StringUtils.deleteWhitespace(text).toLowerCase(Locale.ROOT);
+    }
+
+
+    @SneakyThrows
+    protected ClassificationDocument parseLayout(String filename, LayoutParsingType layoutParsingType) {
+
+        File fileResource = new ClassPathResource(filename).getFile();
+        prepareStorage(filename, "cv_table_parsing_response/empty.json", "image_service_response/crafted_outline_test_doc.IMAGE_INFO.json");
+        return layoutParsingPipeline.parseLayout(layoutParsingType,
+                                                 fileResource,
+                                                 layoutParsingStorageService.getImagesFile(IMAGE_FILE_ID),
+                                                 new TableServiceResponse(),
+                                                 new VisualLayoutParsingResponse(),
+                                                 Map.of("file", filename, "debug", "true"));
+    }
+
+
+    @SneakyThrows
+    protected Document buildGraph(String filename, ClassificationDocument classificationDocument) {
+
+        if (!filename.startsWith("files") && filename.startsWith("/")) {
+            LayoutParsingRequest layoutParsingRequest = buildDefaultLayoutParsingRequest(Path.of(filename).getFileName().toString(), LayoutParsingType.REDACT_MANAGER, true);
+            prepareStorage(layoutParsingRequest, new File(filename));
+            return DocumentGraphFactory.buildDocumentGraph(LayoutParsingType.REDACT_MANAGER,
+                                                           layoutParsingPipeline.parseLayout(LayoutParsingType.REDACT_MANAGER,
+                                                                                             new File(filename),
+                                                                                             layoutParsingStorageService.getImagesFile(layoutParsingRequest.imagesFileStorageId()
+                                                                                                                                               .get()),
+                                                                                             new TableServiceResponse(),
+                                                                                             new VisualLayoutParsingResponse(),
+                                                                                             layoutParsingRequest.identifier()));
+        } else {
+            prepareStorage(filename);
+            return DocumentGraphFactory.buildDocumentGraph(LayoutParsingType.REDACT_MANAGER, classificationDocument);
+        }
+
+    }
+
+}
+
@@ -64,8 +64,8 @@ public class SimplifiedTextServiceTest
    @SneakyThrows
    protected Document buildGraph(File file) {

-        return DocumentGraphFactory.buildDocumentGraph(LayoutParsingType.REDACT_MANAGER,
-                                                       layoutParsingPipeline.parseLayout(LayoutParsingType.REDACT_MANAGER,
+        return DocumentGraphFactory.buildDocumentGraph(LayoutParsingType.REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH,
+                                                       layoutParsingPipeline.parseLayout(LayoutParsingType.REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH,
                                                                                         file,
                                                                                         new ImageServiceResponse(),
                                                                                         new TableServiceResponse(),
@@ -56,8 +56,8 @@ public class DocumentGraphJsonWritingTest extends BuildDocumentTest {
    @SneakyThrows
    private void writeJsons(Path filename) {

-        Document documentGraph = DocumentGraphFactory.buildDocumentGraph(LayoutParsingType.REDACT_MANAGER_OLD,
-                layoutParsingPipeline.parseLayout(LayoutParsingType.REDACT_MANAGER_OLD,
+        Document documentGraph = DocumentGraphFactory.buildDocumentGraph(LayoutParsingType.REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH,
+                layoutParsingPipeline.parseLayout(LayoutParsingType.REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH,
                        filename.toFile(),
                        new ImageServiceResponse(),
                        new TableServiceResponse(),
@@ -31,14 +31,14 @@ public class ViewerDocumentTest extends BuildDocumentTest {
    @SneakyThrows
    public void testViewerDocument() {

-        String fileName = "files/syngenta/CustomerFiles/SinglePages/S4_S-Metolachlor_RAR_01_Volume_1_2018-09-06.pdf";
+        String fileName = "files/syngenta/CustomerFiles/S-Metolachlor_RAR_01_Volume_1_2018-09-06.pdf";

        String tmpFileName = "/tmp/" + Path.of(fileName).getFileName() + "_VIEWER.pdf";

        var documentFile = new ClassPathResource(fileName).getFile();

        long start = System.currentTimeMillis();
-        Document document = buildGraph(fileName, LayoutParsingType.REDACT_MANAGER);
+        Document document = buildGraph(fileName, LayoutParsingType.REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH);
        layoutGridService.addLayoutGrid(documentFile, document, new File(tmpFileName), true, false);
        System.out.printf("Total time: %.2fs%n", ((float) (System.currentTimeMillis() - start)) / 1000);
    }
@@ -54,7 +54,7 @@ public class PdfSegmentationServiceTest extends AbstractTest {
    @SneakyThrows
    public ClassificationDocument buildClassificationDocument(File originDocument, TableServiceResponse tableServiceResponse) {

-        return layoutParsingPipeline.parseLayout(LayoutParsingType.REDACT_MANAGER,
+        return layoutParsingPipeline.parseLayout(LayoutParsingType.REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH,
                                                 originDocument,
                                                 new ImageServiceResponse(),
                                                 tableServiceResponse,
@@ -114,7 +114,7 @@ public class PdfSegmentationServiceTest extends AbstractTest {
        ClassificationDocument classificationDocument = buildClassificationDocument(pdfFileResource.getFile());

        assertThat(classificationDocument.getHeaders()
-                           .get(0).getTextBlocks().size()).isEqualTo(2);
+                           .get(0).getTextBlocks().size()).isEqualTo(3);
        assertThat(classificationDocument.getHeaders()
                           .get(0).getTextBlocks()
                           .get(0).getSequences().size()).isEqualTo(8);
@@ -122,7 +122,7 @@ public class PdfSegmentationServiceTest extends AbstractTest {
                           .get(0).getTextBlocks()
                           .get(0).toString()).contains(textToSearch);

-        Document document = DocumentGraphFactory.buildDocumentGraph(LayoutParsingType.REDACT_MANAGER, classificationDocument);
+        Document document = DocumentGraphFactory.buildDocumentGraph(LayoutParsingType.REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH, classificationDocument);

        TextBlock leafTextBlock = document.getFirstPage().getHeader().getLeafTextBlock();
        assertThat(leafTextBlock.getSearchText().contains(textToSearch)).isTrue();
@@ -190,7 +190,8 @@ public class PdfSegmentationServiceTest extends AbstractTest {
                                                                        imageMetadata.getGeometry().getHeight()),
                                                 ImageType.valueOf(imageMetadata.getClassification().getLabel().toUpperCase(Locale.ROOT)),
                                                 imageMetadata.isAlpha(),
-                                                 imageMetadata.getPosition().getPageNumber())));
+                                                 imageMetadata.getPosition().getPageNumber(),
+                                                 "")));

        System.out.println("object");
    }
@@ -103,15 +103,15 @@ public class RulingCleaningServiceTest extends BuildDocumentTest {
    @SneakyThrows
    private void writeJsons(Path filename) {

-        Document documentGraphBefore = DocumentGraphFactory.buildDocumentGraph(LayoutParsingType.REDACT_MANAGER,
-                layoutParsingPipeline.parseLayout(LayoutParsingType.REDACT_MANAGER,
+        Document documentGraphBefore = DocumentGraphFactory.buildDocumentGraph(LayoutParsingType.REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH,
+                layoutParsingPipeline.parseLayout(LayoutParsingType.REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH,
                        filename.toFile(),
                        new ImageServiceResponse(),
                        new TableServiceResponse(),
                        new VisualLayoutParsingResponse(),
                        Map.of("file",filename.toFile().toString())));
-        Document documentGraphAfter = DocumentGraphFactory.buildDocumentGraph(LayoutParsingType.REDACT_MANAGER,
-                layoutParsingPipeline.parseLayout(LayoutParsingType.REDACT_MANAGER,
+        Document documentGraphAfter = DocumentGraphFactory.buildDocumentGraph(LayoutParsingType.REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH,
+                layoutParsingPipeline.parseLayout(LayoutParsingType.REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH,
                        filename.toFile(),
                        new ImageServiceResponse(),
                        new TableServiceResponse(),
@@ -48,8 +48,9 @@ public abstract class BuildDocumentTest extends AbstractTest {
    @SneakyThrows
    protected Document buildGraph(String filename, LayoutParsingType layoutParsingType) {

+
        if (!filename.startsWith("files") && filename.startsWith("/")) {
-            LayoutParsingRequest layoutParsingRequest = buildDefaultLayoutParsingRequest(Path.of(filename).getFileName().toString(), LayoutParsingType.REDACT_MANAGER, true);
+            LayoutParsingRequest layoutParsingRequest = buildDefaultLayoutParsingRequest(Path.of(filename).getFileName().toString(), LayoutParsingType.REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH, true);
            prepareStorage(layoutParsingRequest, new File(filename));
            return DocumentGraphFactory.buildDocumentGraph(layoutParsingType,
                                                           layoutParsingPipeline.parseLayout(layoutParsingType,
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8afb731a307e1a3f827c59e902164b10bdabef96e14193b949fe081cd3aa859f
+size 168878
@@ -0,0 +1,135 @@
+{
+  "dossierId": "a91f19ff-11ba-4735-9f60-c650243f64a9",
+  "fileId": "6e8c5f114e2b71e103a32a20c5273188",
+  "targetFileExtension": "ORIGIN.pdf.gz",
+  "responseFileExtension": "IMAGE_INFO.json.gz",
+  "X-TENANT-ID": "redaction",
+  "data": [
+    {
+      "classification": {
+        "label": "other",
+        "probabilities": {
+          "other": 0.9126,
+          "formula": 0.0588,
+          "signature": 0.0261,
+          "logo": 0.0024
+        }
+      },
+      "representation": "70E1070C1030E081B7EF7FFFF",
+      "position": {
+        "x1": 61,
+        "x2": 394,
+        "y1": 155,
+        "y2": 470,
+        "pageNumber": 1
+      },
+      "geometry": {
+        "width": 333,
+        "height": 315
+      },
+      "alpha": false,
+      "filters": {
+        "geometry": {
+          "imageSize": {
+            "quotient": 0.5976,
+            "tooLarge": false,
+            "tooSmall": false
+          },
+          "imageFormat": {
+            "quotient": 1.0571,
+            "tooTall": false,
+            "tooWide": false
+          }
+        },
+        "probability": {
+          "unconfident": false
+        },
+        "allPassed": true
+      }
+    },
+    {
+      "classification": {
+        "label": "other",
+        "probabilities": {
+          "other": 0.9126,
+          "formula": 0.0588,
+          "signature": 0.0261,
+          "logo": 0.0024
+        }
+      },
+      "representation": "70E1070C1030E081B7EF7FFFF",
+      "position": {
+        "x1": 61,
+        "x2": 394,
+        "y1": 202,
+        "y2": 517,
+        "pageNumber": 11
+      },
+      "geometry": {
+        "width": 333,
+        "height": 315
+      },
+      "alpha": false,
+      "filters": {
+        "geometry": {
+          "imageSize": {
+            "quotient": 0.5976,
+            "tooLarge": false,
+            "tooSmall": false
+          },
+          "imageFormat": {
+            "quotient": 1.0571,
+            "tooTall": false,
+            "tooWide": false
+          }
+        },
+        "probability": {
+          "unconfident": false
+        },
+        "allPassed": true
+      }
+    },
+    {
+      "classification": {
+        "label": "other",
+        "probabilities": {
+          "other": 0.9126,
+          "formula": 0.0588,
+          "signature": 0.0261,
+          "logo": 0.0024
+        }
+      },
+      "representation": "70E1070C1030E081B7EF7FFFF",
+      "position": {
+        "x1": 47,
+        "x2": 379,
+        "y1": 289,
+        "y2": 604,
+        "pageNumber": 16
+      },
+      "geometry": {
+        "width": 332,
+        "height": 315
+      },
+      "alpha": false,
+      "filters": {
+        "geometry": {
+          "imageSize": {
+            "quotient": 0.5967,
+            "tooLarge": false,
+            "tooSmall": false
+          },
+          "imageFormat": {
+            "quotient": 1.054,
+            "tooTall": false,
+            "tooWide": false
+          }
+        },
+        "probability": {
+          "unconfident": false
+        },
+        "allPassed": true
+      }
+    }
+  ]
+}
@@ -12,6 +12,7 @@ dependencies {
    implementation("org.apache.pdfbox:pdfbox-tools:${pdfBoxVersion}")
    implementation("org.slf4j:slf4j-api:1.7.25")
    implementation("com.knecon.fforesight:tracing-commons:0.5.0")
+    implementation("com.pdftron:PDFNet:10.5.0")

    testImplementation("org.apache.logging.log4j:log4j-slf4j-impl:2.22.1")
    testImplementation("org.junit.jupiter:junit-jupiter")
@@ -3,8 +3,13 @@ package com.knecon.fforesight.service.viewerdoc.model;
 import org.apache.pdfbox.pdmodel.PDDocument;
 import org.apache.pdfbox.pdmodel.font.PDFont;

+import com.pdftron.pdf.Font;
+import com.pdftron.pdf.PDFDoc;
+
 public interface EmbeddableFont {

    PDFont embed(PDDocument document);

+    Font embed(PDFDoc doc);
+
 }
@@ -1,10 +1,15 @@
 package com.knecon.fforesight.service.viewerdoc.model;

+import java.util.Objects;
+
 import org.apache.pdfbox.pdmodel.PDDocument;
 import org.apache.pdfbox.pdmodel.font.PDFont;
 import org.apache.pdfbox.pdmodel.font.PDType1Font;
 import org.apache.pdfbox.pdmodel.font.Standard14Fonts;

+import com.pdftron.pdf.Font;
+import com.pdftron.pdf.PDFDoc;
+
 import lombok.RequiredArgsConstructor;
 import lombok.SneakyThrows;

@@ -13,10 +18,11 @@ public class Standard14EmbeddableFont implements EmbeddableFont {

    private final PDType1Font font;

+    private final int pdfTronIdentifier;

    public static Standard14EmbeddableFont helvetica() {

-        return new Standard14EmbeddableFont(new PDType1Font(Standard14Fonts.FontName.HELVETICA));
+        return new Standard14EmbeddableFont(new PDType1Font(Standard14Fonts.FontName.HELVETICA), Font.e_helvetica);
    }


@@ -34,4 +40,12 @@ public class Standard14EmbeddableFont implements EmbeddableFont {
        return font;
    }

+
+    @Override
+    @SneakyThrows
+    public Font embed(PDFDoc document) {
+
+        return Font.create(document, pdfTronIdentifier);
+    }
+
 }
@@ -23,4 +23,10 @@ public class VisualizationsOnPage {
    @Builder.Default
    List<FilledRectangle> filledRectangles = new LinkedList<>();

+
+    public boolean isEmpty() {
+
+        return placedTexts.isEmpty() && coloredLines.isEmpty() && coloredRectangles.isEmpty() && filledRectangles.isEmpty();
+    }
+
 }
@@ -0,0 +1,27 @@
+package com.knecon.fforesight.service.viewerdoc.service;
+
+import java.io.File;
+import java.util.List;
+
+import com.knecon.fforesight.service.viewerdoc.ContentStreams;
+import com.knecon.fforesight.service.viewerdoc.model.Visualizations;
+
+import io.micrometer.observation.ObservationRegistry;
+
+public interface IViewerDocumentService {
+
+    void addVisualizationsOnPage(File originFile, File destinationFile, List<Visualizations> visualizations);
+
+    default void enrichObservation(ObservationRegistry registry, int numberOfPages, List<ContentStreams.Identifier> layers) {
+
+        if (registry == null || registry.getCurrentObservation() == null || registry.isNoop()) {
+            return;
+        }
+        registry.getCurrentObservation().highCardinalityKeyValue("numberOfPages", String.valueOf(numberOfPages));
+        for (int i = 0; i < layers.size(); i++) {
+            ContentStreams.Identifier layer = layers.get(i);
+
+            registry.getCurrentObservation().highCardinalityKeyValue("layer_" + i, String.valueOf(layer.name()));
+        }
+    }
+}
@@ -48,7 +48,7 @@ import lombok.extern.slf4j.Slf4j;

@Slf4j
@RequiredArgsConstructor
-public class ViewerDocumentService {
+public class ViewerDocumentService implements IViewerDocumentService {

    private final ObservationRegistry registry;

@@ -64,7 +64,8 @@ public class ViewerDocumentService {

        PDDocument pdDocument = openPDDocument(tmpFile.toFile());

-        enrichObservation(pdDocument,
+        enrichObservation(registry,
+                          pdDocument.getNumberOfPages(),
                          visualizations.stream()
                                  .map(Visualizations::getLayer)
                                  .toList());
@@ -106,7 +107,11 @@ public class ViewerDocumentService {

                    contentStream.saveGraphicsState();

-                    drawVisualizationsToContentStream(pdDocument, visualization.getVisualizationsOnPages().get(pageNumber), contentStream, textDeRotationMatrix);
+                    drawVisualizationsToContentStream(pdDocument,
+                                                      visualization.getVisualizationsOnPages()
+                                                              .get(pageNumber),
+                                                      contentStream,
+                                                      textDeRotationMatrix);

                    contentStream.restoreGraphicsState();

@@ -185,8 +190,10 @@ public class ViewerDocumentService {
            contentStream.setFont(font, placedText.fontSize());
            contentStream.beginText();
            contentStream.setNonStrokingColor(placedText.color());
-            if (placedText.renderingMode().isPresent()) {
-                contentStream.setRenderingMode(placedText.renderingMode().get());
+            if (placedText.renderingMode()
+                    .isPresent()) {
+                contentStream.setRenderingMode(placedText.renderingMode()
+                                                       .get());
            } else {
                contentStream.setRenderingMode(RenderingMode.FILL);
            }
@@ -198,12 +205,12 @@ public class ViewerDocumentService {
    }


-    private void enrichObservation(PDDocument pdDocument, List<ContentStreams.Identifier> layers) {
+    private void enrichObservation(int numberOfPages, List<ContentStreams.Identifier> layers) {

        if (registry == null || registry.getCurrentObservation() == null || registry.isNoop()) {
            return;
        }
-        registry.getCurrentObservation().highCardinalityKeyValue("numberOfPages", String.valueOf(pdDocument.getNumberOfPages()));
+        registry.getCurrentObservation().highCardinalityKeyValue("numberOfPages", String.valueOf(numberOfPages));
        for (int i = 0; i < layers.size(); i++) {
            ContentStreams.Identifier layer = layers.get(i);

@@ -234,7 +241,8 @@ public class ViewerDocumentService {
                                    (float) placedText.lineStart().getX(),
                                    (float) placedText.lineStart().getY());
        } else {
-            textMatrix = placedText.textMatrix().get();
+            textMatrix = placedText.textMatrix()
+                    .get();
        }
        return textMatrix;
    }
@@ -0,0 +1,73 @@
+package com.knecon.fforesight.service.viewerdoc.service.pdftron;
+
+import java.util.Deque;
+import java.util.Iterator;
+import java.util.LinkedList;
+import java.util.Set;
+
+public class MarkedContentStack {
+
+    private final Deque<MarkedContent> stack = new LinkedList<>();
+
+
+    public void enterMarkedContent(String name) {
+
+        stack.push(new MarkedContent(name));
+    }
+
+
+    public void leaveMarkedContent() {
+
+        stack.pop();
+    }
+
+
+    public String currentMarkedContent() {
+
+        if (stack.isEmpty()) {
+            return "";
+        }
+        return stack.peek().name();
+    }
+
+
+    public boolean currentMarkedContentContains(String name) {
+
+        Iterator<MarkedContent> markedContentIterator = stack.descendingIterator();
+        while (markedContentIterator.hasNext()) {
+            var markedContent = markedContentIterator.next();
+            if (markedContent.name().equals(name)) {
+                return true;
+            }
+        }
+        return false;
+    }
+
+
+    public boolean currentMarkedContentContainsAny(Set<String> names) {
+
+        if (stack.isEmpty()) {
+            return false;
+        }
+        Iterator<MarkedContent> markedContentIterator = stack.descendingIterator();
+        while (markedContentIterator.hasNext()) {
+            var markedContent = markedContentIterator.next();
+            if (names.contains(markedContent.name())) {
+                return true;
+            }
+        }
+        return false;
+    }
+
+
+    public void clear() {
+
+        stack.clear();
+    }
+
+
+    private record MarkedContent(String name) {
+
+    }
+
+}
@@ -0,0 +1,149 @@
+package com.knecon.fforesight.service.viewerdoc.service.pdftron;
+
+import java.io.File;
+import java.io.FileInputStream;
+import java.nio.file.Files;
+import java.nio.file.Path;
+import java.nio.file.StandardCopyOption;
+import java.util.Collection;
+import java.util.List;
+import java.util.Map;
+import java.util.Set;
+import java.util.function.Function;
+import java.util.stream.Collectors;
+import java.util.stream.Stream;
+
+import org.apache.pdfbox.cos.COSName;
+
+import com.knecon.fforesight.service.viewerdoc.ContentStreams;
+import com.knecon.fforesight.service.viewerdoc.model.EmbeddableFont;
+import com.knecon.fforesight.service.viewerdoc.model.PlacedText;
+import com.knecon.fforesight.service.viewerdoc.model.Visualizations;
+import com.knecon.fforesight.service.viewerdoc.model.VisualizationsOnPage;
+import com.knecon.fforesight.service.viewerdoc.service.IViewerDocumentService;
+import com.pdftron.pdf.ElementBuilder;
+import com.pdftron.pdf.ElementReader;
+import com.pdftron.pdf.ElementWriter;
+import com.pdftron.pdf.Font;
+import com.pdftron.pdf.PDFDoc;
+import com.pdftron.pdf.Page;
+import com.pdftron.pdf.PageIterator;
+import com.pdftron.pdf.ocg.Group;
+import com.pdftron.sdf.SDFDoc;
+
+import io.micrometer.observation.ObservationRegistry;
+import io.micrometer.observation.annotation.Observed;
+import lombok.RequiredArgsConstructor;
+import lombok.SneakyThrows;
+import lombok.extern.slf4j.Slf4j;
+
+@Slf4j
+@RequiredArgsConstructor
+public class PDFTronViewerDocumentService implements IViewerDocumentService {
+
+    private final ObservationRegistry registry;
+
+
+    @Override
+    @Observed(name = "PDFTronViewerDocumentService", contextualName = "add-visualizations")
+    @SneakyThrows
+    public synchronized void addVisualizationsOnPage(File originFile, File destinationFile, List<Visualizations> visualizations) {
+
+
+        // originFile and destinationFile might be the same, so we use a temp file.
+        // Otherwise, saving the document might corrupt the file
+        Path tmpFile = Files.createTempFile("tmpViewerDocument", ".pdf");
+        Files.copy(originFile.toPath(), tmpFile, StandardCopyOption.REPLACE_EXISTING);
+
+        try (PDFDoc pdfDoc = loadPdfDoc(tmpFile);//
+             ElementWriter pageWriter = new ElementWriter();//
+             ElementReader reader = new ElementReader();//
+             ElementBuilder builder = new ElementBuilder()//
+        ) {
+            enrichObservation(registry,
+                              pdfDoc.getPageCount(),
+                              visualizations.stream()
+                                      .map(Visualizations::getLayer)
+                                      .toList());
+
+            Map<ContentStreams.Identifier, Group> groupMap = PdftronLayerUtility.addLayersToDocument(visualizations, pdfDoc);
+
+            Map<EmbeddableFont, Font> fontMap = buildFontMap(visualizations, pdfDoc);
+
+            Set<String> markedContentToDraw = extractMarkedContentNames(visualizations.stream()
+                                                                                .map(Visualizations::getLayer));
+
+            Set<String> kneconMarkedContents = extractMarkedContentNames(ContentStreams.allContentStreams.stream());
+
+            PageContentCleaner pageContentCleaner = PageContentCleaner.builder()
+                    .writer(pageWriter)
+                    .reader(reader)
+                    .elementBuilder(builder)
+                    .markedContentToDraw(markedContentToDraw)
+                    .kneconMarkedContents(kneconMarkedContents)
+                    .build();
+
+            VisualizationWriter visualizationWriter = VisualizationWriter.builder()
+                    .writer(pageWriter)
+                    .builder(builder)
+                    .groupMap(groupMap)
+                    .visualizations(visualizations)
+                    .fontMap(fontMap)
+                    .build();
+
+            int pageNumber = 0;
+            for (PageIterator iterator = pdfDoc.getPageIterator(); iterator.hasNext(); pageNumber++) {
+
+                Page page = iterator.next();
+
+                pageContentCleaner.cleanPage(page);
+
+                visualizationWriter.drawVisualizationsOnPage(pageNumber, page);
+
+            }
+
+            saveDocument(pdfDoc, destinationFile);
+        }
+
+    }
+
+
+    private static Set<String> extractMarkedContentNames(Stream<ContentStreams.Identifier> visualizations) {
+
+        return visualizations.map(ContentStreams.Identifier::cosName)
+                .map(COSName::getName)
+                .collect(Collectors.toSet());
+    }
+
+
+    private static Map<EmbeddableFont, Font> buildFontMap(List<Visualizations> visualizations, PDFDoc pdfDoc) {
+
+        return visualizations.stream()
+                .map(Visualizations::getVisualizationsOnPages)
+                .map(Map::values)
+                .flatMap(Collection::stream)
+                .map(VisualizationsOnPage::getPlacedTexts)
+                .flatMap(Collection::stream)
+                .map(PlacedText::font)
+                .distinct()
+                .collect(Collectors.toMap(Function.identity(), font -> font.embed(pdfDoc)));
+
+    }
+
+
+    @SneakyThrows
+    private void saveDocument(PDFDoc doc, File target) {
+
+        doc.save(target.toString(), SDFDoc.SaveMode.REMOVE_UNUSED, null);
+    }
+
+
+    @SneakyThrows
+    private static PDFDoc loadPdfDoc(Path tmpFile) {
+
+        try (var in = new FileInputStream(tmpFile.toFile())) {
+            return new PDFDoc(in);
+        }
+    }
+
+}
@@ -0,0 +1,120 @@
+package com.knecon.fforesight.service.viewerdoc.service.pdftron;
+
+import java.util.Set;
+
+import com.pdftron.pdf.Element;
+import com.pdftron.pdf.ElementBuilder;
+import com.pdftron.pdf.ElementReader;
+import com.pdftron.pdf.ElementWriter;
+import com.pdftron.pdf.Page;
+
+import lombok.AccessLevel;
+import lombok.AllArgsConstructor;
+import lombok.Builder;
+import lombok.SneakyThrows;
+import lombok.experimental.FieldDefaults;
+
+@Builder
+@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
+public class PageContentCleaner {
+
+    ElementWriter writer;
+    ElementReader reader;
+    ElementBuilder elementBuilder;
+    Set<String> markedContentToDraw;
+    Set<String> kneconMarkedContents;
+
+    @Builder.Default
+    MarkedContentStack markedContentStack = new MarkedContentStack();
+
+
+    @SneakyThrows
+    public void cleanPage(Page page) {
+
+        begin(page);
+        boolean escaped = reader.next().getType() == Element.e_group_begin;
+
+        if (!escaped) {
+            writer.writeElement(elementBuilder.createGroupBegin());
+        }
+
+        copyElementsUntilFirstKneconMarkedContent();
+
+        if (!escaped) {
+            writer.writeElement(elementBuilder.createGroupEnd());
+        }
+
+        copyElementsExceptMarkedContentToDraw();
+        end();
+    }
+
+
+    @SneakyThrows
+    private void begin(Page page) {
+
+        writer.begin(page, ElementWriter.e_replacement, false, true, page.getResourceDict());
+        reader.begin(page);
+    }
+
+
+    @SneakyThrows
+    private void end() {
+
+        writer.end();
+        reader.end();
+    }
+
+
+    @SneakyThrows
+    private void copyElementsUntilFirstKneconMarkedContent() {
+
+        for (Element element = reader.current(); element != null; element = reader.next()) {
+
+            switch (element.getType()) {
+                case Element.e_marked_content_begin -> {
+                    markedContentStack.enterMarkedContent(element.getMCTag().getName());
+                    if (markedContentStack.currentMarkedContentContainsAny(kneconMarkedContents)) {
+                        break;
+                    }
+                    writer.writeElement(element);
+                }
+                case Element.e_marked_content_end -> {
+                    markedContentStack.leaveMarkedContent();
+                    writer.writeElement(element);
+                }
+                default -> writer.writeElement(element);
+            }
+        }
+    }
+
+
+    @SneakyThrows
+    private void copyElementsExceptMarkedContentToDraw() {
+
+        for (Element element = reader.current(); element != null; element = reader.next()) {
+
+            switch (element.getType()) {
+                case Element.e_marked_content_begin -> {
+                    markedContentStack.enterMarkedContent(element.getMCTag().getName());
+                    if (!markedContentStack.currentMarkedContentContainsAny(markedContentToDraw)) {
+                        writer.writeElement(element);
+                    }
+                }
+                case Element.e_marked_content_end -> {
+                    if (!markedContentStack.currentMarkedContentContainsAny(markedContentToDraw)) {
+                        writer.writeElement(element);
+                    }
+                    markedContentStack.leaveMarkedContent();
+                }
+                default -> {
+                    if (!markedContentStack.currentMarkedContentContainsAny(markedContentToDraw)) {
+                        writer.writeElement(element);
+                    }
+                }
+            }
+
+        }
+
+    }
+
+}
@@ -0,0 +1,96 @@
+package com.knecon.fforesight.service.viewerdoc.service.pdftron;
+
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+import java.util.Optional;
+
+import com.knecon.fforesight.service.viewerdoc.ContentStreams;
+import com.knecon.fforesight.service.viewerdoc.model.Visualizations;
+import com.pdftron.common.PDFNetException;
+import com.pdftron.pdf.PDFDoc;
+import com.pdftron.pdf.ocg.Config;
+import com.pdftron.pdf.ocg.Group;
+import com.pdftron.sdf.Obj;
+
+import lombok.SneakyThrows;
+import lombok.experimental.UtilityClass;
+
+@UtilityClass
+public class PdftronLayerUtility {
+
+    public Map<ContentStreams.Identifier, Group> addLayersToDocument(List<Visualizations> visualizations, PDFDoc pdfDoc) {
+
+        Map<ContentStreams.Identifier, Group> optionalContentGroupMap = new HashMap<>();
+        for (Visualizations visualization : visualizations) {
+            addLayerToDocument(visualization.getLayer(), pdfDoc, visualization.isLayerVisibilityDefaultValue())//
+                    .ifPresent(ocg -> optionalContentGroupMap.put(visualization.getLayer(), ocg));
+        }
+        return optionalContentGroupMap;
+    }
+
+
+    private Optional<Group> addLayerToDocument(ContentStreams.Identifier layer, PDFDoc pdfDoc, boolean layerVisibilityDefaultValue) {
+
+        if (layer.optionalContent()) {
+            return Optional.of(addLayerToDocument(pdfDoc, layer.name(), layerVisibilityDefaultValue));
+
+        }
+        return Optional.empty();
+
+    }
+
+
+    @SneakyThrows
+    private Group addLayerToDocument(PDFDoc doc, String layerName, boolean layerVisibilityDefaultValue) {
+
+        Optional<Group> existingGroup = findGroupInDoc(doc, layerName);
+
+        if (existingGroup.isPresent()) {
+            return existingGroup.get();
+        }
+
+        return addNewLayer(doc, layerName, layerVisibilityDefaultValue);
+    }
+
+
+    private Group addNewLayer(PDFDoc doc, String layerName, boolean layerVisibilityDefaultValue) throws PDFNetException {
+
+        Config cfg = doc.getOCGConfig();
+        if (cfg == null) {
+            cfg = Config.create(doc, true);
+            cfg.setName("Default");
+        }
+        Group grp = Group.create(doc, layerName);
+        grp.setInitialState(cfg, layerVisibilityDefaultValue);
+
+        // Add the new OCG to the list of layers that should appear in PDF viewer GUI.
+        Obj layerOrderArray = cfg.getOrder();
+        if (layerOrderArray == null) {
+            layerOrderArray = doc.createIndirectArray();
+            cfg.setOrder(layerOrderArray);
+        }
+        layerOrderArray.pushBack(grp.getSDFObj());
+
+        return grp;
+    }
+
+
+    @SneakyThrows
+    private Optional<Group> findGroupInDoc(PDFDoc doc, String layerName) {
+
+        Obj ocgs = doc.getOCGs();
+        if (ocgs != null) {
+            int i;
+            int sz = (int) ocgs.size();
+            for (i = 0; i < sz; ++i) {
+                Group ocg = new Group(ocgs.getAt(i));
+                if (ocg.getName().equals(layerName)) {
+                    return Optional.of(ocg);
+                }
+            }
+        }
+        return Optional.empty();
+    }
+
+}
@@ -0,0 +1,249 @@
+package com.knecon.fforesight.service.viewerdoc.service.pdftron;
+
+import java.awt.geom.AffineTransform;
+import java.awt.geom.Line2D;
+import java.awt.geom.Rectangle2D;
+import java.util.List;
+import java.util.Map;
+
+import com.knecon.fforesight.service.viewerdoc.ContentStreams;
+import com.knecon.fforesight.service.viewerdoc.model.ColoredLine;
+import com.knecon.fforesight.service.viewerdoc.model.ColoredRectangle;
+import com.knecon.fforesight.service.viewerdoc.model.EmbeddableFont;
+import com.knecon.fforesight.service.viewerdoc.model.FilledRectangle;
+import com.knecon.fforesight.service.viewerdoc.model.PlacedText;
+import com.knecon.fforesight.service.viewerdoc.model.Visualizations;
+import com.knecon.fforesight.service.viewerdoc.model.VisualizationsOnPage;
+import com.pdftron.common.Matrix2D;
+import com.pdftron.common.PDFNetException;
+import com.pdftron.pdf.ColorPt;
+import com.pdftron.pdf.ColorSpace;
+import com.pdftron.pdf.Element;
+import com.pdftron.pdf.ElementBuilder;
+import com.pdftron.pdf.ElementWriter;
+import com.pdftron.pdf.Font;
+import com.pdftron.pdf.GState;
+import com.pdftron.pdf.Page;
+import com.pdftron.pdf.ocg.Group;
+
+import lombok.AccessLevel;
+import lombok.Builder;
+import lombok.SneakyThrows;
+import lombok.experimental.FieldDefaults;
+
+@Builder
+@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
+public class VisualizationWriter {
+
+    ElementWriter writer;
+    ElementBuilder builder;
+    List<Visualizations> visualizations;
+    Map<ContentStreams.Identifier, Group> groupMap;
+    Map<EmbeddableFont, Font> fontMap;
+
+
+    @SneakyThrows
+    public void drawVisualizationsOnPage(int pageNumber, Page page) {
+
+        begin(page);
+
+        AffineTransform textDeRotationMatrix = getTextDeRotationTransform(page);
+
+        for (Visualizations visualization : visualizations) {
+
+            VisualizationsOnPage visualizationsOnPage = visualization.getVisualizationsOnPages()
+                    .get(pageNumber);
+
+            if (visualizationsOnPage == null || visualizationsOnPage.isEmpty()) {
+                continue;
+            }
+
+            Element markedContentStart = builder.createMarkedContentBeginInlineProperties(visualization.getLayer().cosName().getName());
+            writer.writeElement(markedContentStart);
+
+            if (visualization.getLayer().optionalContent()) {
+                Element ocgStart = builder.createMarkedContentBegin("OC", groupMap.get(visualization.getLayer()).getSDFObj());
+                writer.writeElement(ocgStart);
+            }
+
+            writeVisualization(visualizationsOnPage, textDeRotationMatrix);
+
+            if (visualization.getLayer().optionalContent()) {
+                Element ocgEnd = builder.createMarkedContentEnd();
+                writer.writeElement(ocgEnd);
+            }
+
+            Element markedContentEnd = builder.createMarkedContentEnd();
+            writer.writeElement(markedContentEnd);
+
+        }
+
+        end();
+
+    }
+
+
+    private void end() throws PDFNetException {
+
+        writer.end();
+    }
+
+
+    private void begin(Page page) throws PDFNetException {
+
+        writer.begin(page, ElementWriter.e_overlay, false, true, page.getResourceDict());
+    }
+
+
+    @SneakyThrows
+    private void writeVisualization(VisualizationsOnPage visualizationsOnPage, AffineTransform textDeRotationMatrix) {
+
+        if (visualizationsOnPage.isMakePathsInvisible()) {
+            Element rect = builder.createRect(0, 0, 0, 0);
+            rect.setPathClip(true);
+            writer.writeElement(rect);
+        }
+
+        for (ColoredLine coloredLine : visualizationsOnPage.getColoredLines()) {
+
+            drawColoredLine(coloredLine);
+        }
+
+        for (ColoredRectangle coloredRectangle : visualizationsOnPage.getColoredRectangles()) {
+
+            drawColoredRectangle(coloredRectangle);
+        }
+
+        for (FilledRectangle filledRectangle : visualizationsOnPage.getFilledRectangles()) {
+
+            drawFilledRectangle(filledRectangle);
+        }
+
+        for (PlacedText placedText : visualizationsOnPage.getPlacedTexts()) {
+
+            writePlacedText(textDeRotationMatrix, placedText);
+        }
+    }
+
+
+    private void writePlacedText(AffineTransform textDeRotationMatrix, PlacedText placedText) throws PDFNetException {
+
+        float[] rgbComponents = placedText.color().getRGBColorComponents(null);
+        Font font = fontMap.get(placedText.font());
+
+        Element text = builder.createTextRun(placedText.text(), font, placedText.fontSize());
+
+        if (placedText.renderingMode()
+                .isPresent()) {
+            text.getGState()
+                    .setRenderingIntent(placedText.renderingMode()
+                                                .get().intValue());
+        } else {
+            try (ColorPt color = new ColorPt(rgbComponents[0], rgbComponents[1], rgbComponents[2])) {
+                text.getGState().setFillColor(color);
+            }
+            text.getGState().setRenderingIntent(GState.e_fill_text);
+        }
+
+        try (Matrix2D textMatrix = getTextMatrix(placedText, textDeRotationMatrix)) {
+            text.setTextMatrix(textMatrix);
+        }
+
+        writer.writeElement(text);
+    }
+
+
+    private void drawFilledRectangle(FilledRectangle filledRectangle) throws PDFNetException {
+
+        float[] rgbComponents = filledRectangle.color().getRGBColorComponents(null);
+        Rectangle2D r = filledRectangle.rectangle2D();
+
+        Element rect = builder.createRect(r.getX(), r.getY(), r.getWidth(), r.getHeight());
+
+        rect.setPathFill(true);
+        rect.setPathStroke(false);
+        rect.getGState().setFillColorSpace(ColorSpace.createDeviceRGB());
+        rect.getGState().setFillOpacity(filledRectangle.alpha());
+
+        try (ColorPt color = new ColorPt(rgbComponents[0], rgbComponents[1], rgbComponents[2])) {
+            rect.getGState().setFillColor(color);
+        }
+
+        writer.writeElement(rect);
+    }
+
+
+    private void drawColoredRectangle(ColoredRectangle coloredRectangle) throws PDFNetException {
+
+        float[] rgbComponents = coloredRectangle.color().getRGBColorComponents(null);
+        Rectangle2D r = coloredRectangle.rectangle2D();
+
+        Element rect = builder.createRect(r.getX(), r.getY(), r.getWidth(), r.getHeight());
+
+        rect.setPathStroke(true);
+        rect.setPathFill(false);
+        rect.getGState().setStrokeColorSpace(ColorSpace.createDeviceRGB());
+        rect.getGState().setLineWidth(coloredRectangle.lineWidth());
+
+        try (ColorPt color = new ColorPt(rgbComponents[0], rgbComponents[1], rgbComponents[2])) {
+            rect.getGState().setStrokeColor(color);
+        }
+
+        writer.writeElement(rect);
+    }
+
+
+    private void drawColoredLine(ColoredLine coloredLine) throws PDFNetException {
+
+        float[] rgbComponents = coloredLine.color().getRGBColorComponents(null);
+        Line2D l = coloredLine.line();
+
+        builder.pathBegin();
+        builder.moveTo(l.getX1(), l.getY1());
+        builder.lineTo(l.getX2(), l.getY2());
+        Element line = builder.pathEnd();
+
+        line.setPathStroke(true);
+        line.setPathFill(false);
+        line.getGState().setLineWidth(coloredLine.lineWidth());
+        line.getGState().setStrokeColorSpace(ColorSpace.createDeviceRGB());
+
+        try (ColorPt color = new ColorPt(rgbComponents[0], rgbComponents[1], rgbComponents[2])) {
+            line.getGState().setStrokeColor(color);
+        }
+        writer.writeElement(line);
+    }
+
+
+    @SneakyThrows
+    private static Matrix2D getTextMatrix(PlacedText placedText, AffineTransform textDeRotationMatrix) {
+
+        Matrix2D textMatrix;
+        if (placedText.textMatrix().isEmpty()) {
+            textMatrix = new Matrix2D(textDeRotationMatrix.getScaleX(),
+                                      textDeRotationMatrix.getShearX(),
+                                      textDeRotationMatrix.getShearY(),
+                                      textDeRotationMatrix.getScaleY(),
+                                      placedText.lineStart().getX(),
+                                      placedText.lineStart().getY());
+        } else {
+            var matrix = placedText.textMatrix()
+                    .get();
+            textMatrix = new Matrix2D(matrix.getScaleX(), matrix.getShearX(), matrix.getShearY(), matrix.getScaleY(), matrix.getTranslateX(), matrix.getTranslateY());
+        }
+        return textMatrix;
+    }
+
+
+    @SneakyThrows
+    private static AffineTransform getTextDeRotationTransform(Page page) {
+
+        return AffineTransform.getQuadrantRotateInstance(switch (page.getRotation()) {
+            case 90 -> 3;
+            case 180 -> 2;
+            case 270 -> 1;
+            default -> 0;
+        });
+    }
+
+}
@@ -1,5 +1,7 @@
 #!/bin/bash
-dir=${PWD##*/}
+
+set -e
+
 gradle assemble

 # Get the current Git branch
@@ -11,5 +13,31 @@ commit_hash=$(git rev-parse --short=5 HEAD)
 # Combine branch and commit hash
 buildName="${USER}-${branch}-${commit_hash}"

-gradle bootBuildImage --publishImage -PbuildbootDockerHostNetwork=true -Pversion=$buildName --no-build-cache
-echo "nexus.knecon.com:5001/ff/layoutparser-service-server:$buildName"
+gradle bootBuildImage --publishImage -PbuildbootDockerHostNetwork=true -Pversion=${buildName}
+
+newImageName="nexus.knecon.com:5001/ff/layoutparser-service-server:${buildName}"
+
+echo "full image name:"
+echo ${newImageName}
+echo ""
+
+if [ -z "$1" ]; then
+    exit 0
+fi
+
+namespace=${1}
+deployment_name="layoutparser-service"
+
+echo "deploying to ${namespace}"
+
+oldImageName=$(rancher kubectl -n ${namespace} get deployment ${deployment_name} -o=jsonpath='{.spec.template.spec.containers[*].image}')
+
+if [ "${newImageName}" = "${oldImageName}" ]; then
+    echo "Image tag did not change, redeploying..."
+    rancher kubectl rollout restart deployment ${deployment_name} -n ${namespace}
+else
+    echo "upgrading the image tag..."
+    rancher kubectl set image deployment/${deployment_name} ${deployment_name}=${newImageName} -n ${namespace}
+fi
+rancher kubectl rollout status deployment ${deployment_name} -n ${namespace}
+echo "Built ${deployment_name}:${buildName} and deployed to ${namespace}"
Author	SHA1	Message	Date
Kilian Schüttler	3173610be5	Merge branch 'CLARI-003' into 'main' CLARI-003: add treeId to StructureObject See merge request fforesight/layout-parser!176	2024-07-02 11:37:22 +02:00
Kilian Schuettler	e920eb5a78	CLARI-003: add treeId to StructureObject	2024-07-01 13:56:16 +02:00
Kilian Schüttler	7e4baea7e5	Merge branch 'RED-9353' into 'main' RED-9353: use azure ocr service See merge request fforesight/layout-parser!175	2024-07-01 11:13:27 +02:00
Kilian Schüttler	66d3433e04	RED-9353: use azure ocr service	2024-07-01 11:13:26 +02:00
Yannik Hampe	a2f559af51	Merge branch 'RED-3813' into 'main' RED-3813: Recategorize same image as experimental feature See merge request fforesight/layout-parser!155	2024-06-26 13:42:42 +02:00
Yannik Hampe	39f527a57c	Merge branch 'main' into 'RED-3813' # Conflicts: # layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/LayoutParsingPipeline.java	2024-06-26 09:10:59 +02:00
yhampe	5c2844fe31	RED-3813: Recategorize same image as experimental feature fixed failing test	2024-06-26 09:08:37 +02:00
Kilian Schüttler	b216f02e15	Merge branch 'RED-9194' into 'main' RED-9194: roll back single digit headline change See merge request fforesight/layout-parser!171	2024-06-21 15:13:40 +02:00
Kilian Schuettler	2e2f30ba35	RED-9194: roll back single digit headline change	2024-06-21 14:42:30 +02:00
Kilian Schuettler	9f7ed974ec	RED-9194: roll back single digit headline change	2024-06-21 14:41:30 +02:00
Kilian Schuettler	570a348a77	RED-9194: roll back single digit headline change	2024-06-21 14:39:47 +02:00
Maverick Studer	859dba2ecf	Merge branch 'RED-9374' into 'main' hotfix for table/paragraph section creation on document start before first headline See merge request fforesight/layout-parser!170	2024-06-18 17:36:04 +02:00
Maverick Studer	1c5d755111	hotfix for table/paragraph section creation on document start before first headline	2024-06-18 17:36:04 +02:00
Maverick Studer	133e06460f	Merge branch 'RED-9374' into 'main' RED-9374: Ner Entities are at wrong locations See merge request fforesight/layout-parser!169	2024-06-18 16:31:24 +02:00
Maverick Studer	da91fcff97	RED-9374: Ner Entities are at wrong locations	2024-06-18 16:31:24 +02:00
Kilian Schüttler	79795e408a	Merge branch 'RED-9194' into 'main' RED-9194: allow single digit headline identifiers See merge request fforesight/layout-parser!168	2024-06-07 09:09:25 +02:00
Kilian Schuettler	b719db86ab	RED-9194: allow single digit headline identifiers	2024-06-06 16:32:05 +02:00
Maverick Studer	797602e373	Merge branch 'thread-safe-hcs-fields' into 'main' fixed issue with thread-safety of local fields in the HeadlineClassificationService See merge request fforesight/layout-parser!167	2024-06-06 14:51:24 +02:00
maverickstuder	3d2f66cf10	fixed issue with thread-safety of local fields in the HeadlineClassificationService: * HeadlineClassificationService is no singleton anymore * instead initialize it in the ClassificationService and pass it to the classifyMethods as required	2024-06-06 14:39:23 +02:00
Kilian Schüttler	e304a9f2d7	Merge branch 'RED-7074-le' into 'main' RED-7074: Design Subsection section tree structure algorithm See merge request fforesight/layout-parser!166	2024-06-06 13:22:14 +02:00
Maverick Studer	c05f67cf44	RED-7074: Design Subsection section tree structure algorithm	2024-06-06 13:22:14 +02:00
yhampe	9ecf9ca19f	RED-3813: Recategorize same image as experimental feature now writing hash into structure	2024-06-05 14:20:33 +02:00
Corina Olariu	3a2ee903af	Merge branch 'RED-9206-2' into 'main' RED-9206 - Sections are no longer correctly separated from each other in the test file See merge request fforesight/layout-parser!165	2024-06-05 13:39:15 +02:00
Corina Olariu	072a8aa3da	RED-9206 - Sections are no longer correctly separated from each other in the test file - add REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH case	2024-06-05 14:26:54 +03:00
Corina Olariu	b5cfa7b63d	Merge branch 'RED-9206' into 'main' RED-9206 - Sections are no longer correctly separated from each other in the test file See merge request fforesight/layout-parser!163	2024-06-05 13:13:45 +02:00
Corina Olariu	5f5a6258c5	Merge branch 'main' into RED-9206	2024-06-05 13:34:14 +03:00
Maverick Studer	ac0e83725a	Merge branch 'RED-7074-lgs' into 'main' RED-7074: Design Subsection section tree structure algorithm See merge request fforesight/layout-parser!164	2024-06-05 12:28:00 +02:00
Maverick Studer	5d33ad570e	RED-7074: Design Subsection section tree structure algorithm	2024-06-05 12:28:00 +02:00
Corina Olariu	fd698a78fc	RED-9206 - Sections are no longer correctly separated from each other in the test file - introduce new layout parsing type: REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH to include changes from REDACT_MANAGER apart from duplicate paragraph. - updated junit tests -	2024-06-04 20:55:37 +03:00
Maverick Studer	c3edeb3c7d	Merge branch 'RED-7074-test' into 'main' RED-7074: Design Subsection section tree structure algorithm See merge request fforesight/layout-parser!162	2024-06-04 15:07:40 +02:00
Maverick Studer	fc06dba2ce	RED-7074: Design Subsection section tree structure algorithm	2024-06-04 15:07:40 +02:00
yhampe	9be672c728	RED-3813: Recategorize same image as experimental feature working on pushing properties to persistence service	2024-05-28 13:51:45 +02:00
yhampe	a5fcebce30	RED-3813: Recategorize same image as experimental feature added representation to image and DocumentStructure	2024-05-17 07:34:05 +02:00