RED-9942 - Don't skip images on file

Merge branch 'RED-9760-NPE' into 'main'
RED-9760: Fixed nullpointer in TextPageBlock See merge request fforesight/layout-parser!194
2024-08-23 13:03:25 +03:00 · 2024-08-13 13:24:48 +02:00 · 2024-08-13 13:18:50 +02:00 · 2024-08-12 16:02:51 +02:00 · 2024-08-12 16:02:50 +02:00 · 2024-08-12 14:59:20 +02:00
167 changed files with 4791 additions and 2475 deletions
@@ -42,6 +42,15 @@ tasks.jacocoTestReport {
 }

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

+
 repositories {
    mavenLocal()
    mavenCentral()
@@ -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));
    }


@@ -19,6 +19,7 @@ public class SimplifiedText {
    @Schema(description = "Number of pages in the entire document.")
    private int numberOfPages;
    @Schema(description = "A List of simplified Sections, which contains almost exclusively text.")
+    @Builder.Default
    private List<SimplifiedSectionText> sectionTexts = new ArrayList<>();

 }
@@ -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.")
@@ -19,7 +19,6 @@ public record LayoutParsingRequest(
        @Schema(description = "Path to the original PDF file.")//
        @NonNull String originFileStorageId,//

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

 }
@@ -26,4 +26,8 @@ dependencies {
    implementation("org.springframework.boot:spring-boot-starter-web:3.1.3")
    implementation("org.jgrapht:jgrapht-core:1.5.2")
    implementation("org.tinspin:tinspin-indexes:2.1.3")
+    implementation("org.commonmark:commonmark:0.22.0")
+    implementation("org.commonmark:commonmark-ext-gfm-tables:0.22.0")
+    implementation("com.pdftron:PDFNet:10.11.0")
+
 }
@@ -9,6 +9,7 @@ import java.io.IOException;
 import java.nio.file.Files;
 import java.util.ArrayList;
 import java.util.HashMap;
+import java.util.HashSet;
 import java.util.List;
 import java.util.Map;
 import java.util.concurrent.atomic.AtomicReference;
@@ -24,6 +25,7 @@ import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.No
 import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingFinishedEvent;
 import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingRequest;
 import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingType;
+import com.knecon.fforesight.service.layoutparser.processor.markdown.MarkdownMapper;
 import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationDocument;
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
@@ -68,6 +70,7 @@ import com.knecon.fforesight.service.layoutparser.processor.services.parsing.PDF
 import com.knecon.fforesight.service.layoutparser.processor.services.visualization.LayoutGridService;
 import com.knecon.fforesight.service.layoutparser.processor.utils.MarkedContentUtils;
 import com.knecon.fforesight.service.layoutparser.processor.utils.PageInformation;
+import com.knecon.fforesight.service.layoutparser.processor.utils.TextPositionOperations;

 import io.micrometer.observation.Observation;
 import io.micrometer.observation.ObservationRegistry;
@@ -116,29 +119,14 @@ public class LayoutParsingPipeline {
        log.info("Starting layout parsing for {}", layoutParsingRequest.identifier());

        File originFile = layoutParsingStorageService.getOriginFile(layoutParsingRequest.originFileStorageId());
-//        File viewerDocumentFile = layoutParsingStorageService.getViewerDocFile(layoutParsingRequest.viewerDocumentStorageId()).orElse(originFile);
-        File viewerDocumentFile = originFile;
+        File viewerDocumentFile = layoutParsingStorageService.getViewerDocFile(layoutParsingRequest.viewerDocumentStorageId()).orElse(originFile);

-        VisualLayoutParsingResponse visualLayoutParsingResponse = new VisualLayoutParsingResponse();
-        if (layoutParsingRequest.visualLayoutParsingFileId()
-                .isPresent()) {
-            visualLayoutParsingResponse = layoutParsingStorageService.getVisualLayoutParsingFile(layoutParsingRequest.visualLayoutParsingFileId()
-                                                                                                         .get());
-        }
-
-        ImageServiceResponse imageServiceResponse = new ImageServiceResponse();
-        if (layoutParsingRequest.imagesFileStorageId()
-                .isPresent()) {
-            imageServiceResponse = layoutParsingStorageService.getImagesFile(layoutParsingRequest.imagesFileStorageId()
-                                                                                     .get());
-        }
-
-        TableServiceResponse tableServiceResponse = new TableServiceResponse();
-        if (layoutParsingRequest.tablesFileStorageId()
-                .isPresent()) {
-            tableServiceResponse = layoutParsingStorageService.getTablesFile(layoutParsingRequest.tablesFileStorageId()
-                                                                                     .get());
-        }
+        VisualLayoutParsingResponse visualLayoutParsingResponse = layoutParsingRequest.visualLayoutParsingFileId()
+                .map(layoutParsingStorageService::getVisualLayoutParsingFile).orElse(new VisualLayoutParsingResponse());
+        ImageServiceResponse imageServiceResponse = layoutParsingRequest.imagesFileStorageId()
+                .map(layoutParsingStorageService::getImagesFile).orElse(new ImageServiceResponse());
+        TableServiceResponse tableServiceResponse = layoutParsingRequest.tablesFileStorageId()
+                .map(layoutParsingStorageService::getTablesFile).orElse(new TableServiceResponse());

        ClassificationDocument classificationDocument = parseLayout(settings.getLayoutParsingTypeOverride() == null //
                                                                            ? layoutParsingRequest.layoutParsingType() : settings.getLayoutParsingTypeOverride(),
@@ -160,10 +148,13 @@ public class LayoutParsingPipeline {
        log.info("Storing resulting files for {}", layoutParsingRequest.identifier());

        layoutParsingStorageService.storeDocumentData(layoutParsingRequest, DocumentDataMapper.toDocumentData(documentGraph));
+        if (layoutParsingRequest.documentMarkdownFileStorageId().isPresent()) {
+            layoutParsingStorageService.storeMarkdownFile(layoutParsingRequest.documentMarkdownFileStorageId().get(), new MarkdownMapper().toMarkdownContent(documentGraph));
+        }
        layoutParsingStorageService.storeSimplifiedText(layoutParsingRequest, simplifiedSectionTextService.toSimplifiedText(documentGraph));
        layoutParsingStorageService.storeViewerDocument(layoutParsingRequest, viewerDocumentFile);

-        if (layoutParsingRequest.layoutParsingType().equals(LayoutParsingType.CLARIFYND)) {
+        if (layoutParsingRequest.researchDocumentStorageId() != null) {
            log.info("Building research document data for {}", layoutParsingRequest.identifier());
            var researchDocumentData = TaasDocumentDataMapper.fromDocument(documentGraph);
            layoutParsingStorageService.storeResearchDocumentData(layoutParsingRequest, researchDocumentData);
@@ -244,17 +235,18 @@ public class LayoutParsingPipeline {
        Map<Integer, List<TableCells>> pdfTableCells = cvTableParsingAdapter.buildCvParsedTablesPerPage(tableServiceResponse);
        Map<Integer, List<ClassifiedImage>> pdfImages = imageServiceResponseAdapter.buildClassifiedImagesPerPage(imageServiceResponse);
        Map<Integer, List<ClassifiedImage>> signatures = visualLayoutParsingAdapter.buildExtractedSignaturesPerPage(visualLayoutParsingResponse);
+
        ClassificationDocument classificationDocument = new ClassificationDocument();

        if (settings.isDebug() || identifier.containsKey("debug")) {
-            classificationDocument.getVisualizations().setActive(true);
+            classificationDocument.getLayoutDebugLayer().setActive(true);
        }

        List<ClassificationPage> classificationPages = new ArrayList<>();
        OutlineObject lastProcessedOutlineObject = null;

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

@@ -280,12 +272,14 @@ public class LayoutParsingPipeline {
            stripper.setStartPage(pageNumber);
            stripper.setEndPage(pageNumber);
            stripper.setPdpage(pdPage);
-            if (layoutParsingType.equals(LayoutParsingType.DOCUMINE_OLD)) {
-                stripper.setSortByPosition(true);
-            }
            stripper.getText(originDocument);
            List<TextPositionSequence> words = stripper.getTextPositionSequences();
-            classificationDocument.getVisualizations().addTextVisualizations(words, pageNumber);
+            if (layoutParsingType.equals(LayoutParsingType.DOCUMINE_OLD)) {
+                var lines = TextPositionOperations.groupByLine(new HashSet<>(words));
+                classificationDocument.getLayoutDebugLayer().addLineVisualizationsFromNestedTextPosition(lines, pageNumber);
+                words = TextPositionOperations.sortLines(lines);
+            }
+            classificationDocument.getLayoutDebugLayer().addTextVisualizations(words, pageNumber);

            PDRectangle pdr = pdPage.getMediaBox();

@@ -293,42 +287,34 @@ public class LayoutParsingPipeline {
            boolean isLandscape = pdr.getWidth() > pdr.getHeight() && (rotation == 0 || rotation == 180) || pdr.getHeight() > pdr.getWidth() && (rotation == 90 || rotation == 270);

            PDRectangle cropbox = pdPage.getCropBox();
-            classificationDocument.getVisualizations().addRulingVisualization(stripper.getRulings(), pageNumber);
+            classificationDocument.getLayoutDebugLayer().addRulingVisualization(stripper.getRulings(), pageNumber);
            CleanRulings cleanRulings = rulingCleaningService.deduplicateAndStraightenRulings(pdfTableCells.get(pageNumber), stripper.getRulings());

            PageInformation pageInformation = PageInformation.fromPDPage(pageNumber, pdPage);
            List<Cell> emptyTableCells = TableExtractionService.findCells(cleanRulings.getHorizontals(), cleanRulings.getVerticals(), pageInformation);
-            classificationDocument.getVisualizations().addCellVisualizations(emptyTableCells, pageNumber);
+            classificationDocument.getLayoutDebugLayer().addCellVisualizations(emptyTableCells, pageNumber);

            TextRulingsClassifier.classifyUnderlinedAndStrikethroughText(words, cleanRulings);

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

            pdfImages.computeIfAbsent(pageNumber, x -> new ArrayList<>())
                    .addAll(graphics.stream()
-                                    .map(g -> new ClassifiedImage(new Rectangle2D.Double(g.x1, g.y1, g.width(), g.height()), ImageType.GRAPHIC, false, stripper.getPageNumber()))
+                                    .map(g -> new ClassifiedImage(new Rectangle2D.Double(g.x1, g.y1, g.width(), g.height()),
+                                                                  ImageType.GRAPHIC,
+                                                                  false,
+                                                                  stripper.getPageNumber(),
+                                                                  ""))
                                    .toList());

            ClassificationPage classificationPage = switch (layoutParsingType) {
                case REDACT_MANAGER_OLD ->
-                        redactManagerBlockificationService.blockify(stripper.getTextPositionSequences(), cleanRulings, classificationDocument.getVisualizations());
+                        redactManagerBlockificationService.blockify(stripper.getTextPositionSequences(), cleanRulings, classificationDocument.getLayoutDebugLayer());
                case DOCUMINE_OLD -> docuMineBlockificationService.blockify(words, cleanRulings);
-                case DOCUMINE, REDACT_MANAGER, REDACT_MANAGER_PARAGRAPH_DEBUG, 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);
+                case DOCUMINE, REDACT_MANAGER, REDACT_MANAGER_PARAGRAPH_DEBUG, REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH ->
+                        docstrumBlockificationService.blockify(words, cleanRulings, true, classificationDocument.getLayoutDebugLayer(), layoutParsingType);
+                case CLARIFYND, CLARIFYND_PARAGRAPH_DEBUG ->
+                        docstrumBlockificationService.blockify(words, cleanRulings, false, classificationDocument.getLayoutDebugLayer(), layoutParsingType);
            };

            classificationPage.setCleanRulings(cleanRulings);
@@ -338,7 +324,7 @@ 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;
@@ -352,7 +338,7 @@ public class LayoutParsingPipeline {
                }
            }

-            classificationDocument.getVisualizations().addMarkedContentVisualizations(stripper.getMarkedContents(), pageNumber);
+            classificationDocument.getLayoutDebugLayer().addMarkedContentVisualizations(stripper.getMarkedContents(), pageNumber);
            // MarkedContent needs to be converted at this point, otherwise it leads to GC Problems in Pdfbox.
            classificationPage.setMarkedContentBboxPerType(convertMarkedContents(stripper.getMarkedContents()));

@@ -383,12 +369,12 @@ public class LayoutParsingPipeline {
        log.info("Calculating BodyTextFrame for {}", identifier);
        bodyTextFrameService.setBodyTextFrames(classificationDocument, layoutParsingType);
        for (ClassificationPage page : classificationDocument.getPages()) {
-            classificationDocument.getVisualizations().addCleanRulingVisualization(page.getCleanRulings(), page.getPageNumber());
+            classificationDocument.getLayoutDebugLayer().addCleanRulingVisualization(page.getCleanRulings(), page.getPageNumber());
        }
        log.info("Classify TextBlocks for {}", identifier);
        switch (layoutParsingType) {
-            case REDACT_MANAGER, REDACT_MANAGER_PARAGRAPH_DEBUG, REDACT_MANAGER_OLD, CLARIFYND_PARAGRAPH_DEBUG, REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH -> redactManagerClassificationService.classifyDocument(
-                    classificationDocument);
+            case REDACT_MANAGER, REDACT_MANAGER_PARAGRAPH_DEBUG, REDACT_MANAGER_OLD, CLARIFYND_PARAGRAPH_DEBUG, REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH ->
+                    redactManagerClassificationService.classifyDocument(classificationDocument);
            case DOCUMINE_OLD, DOCUMINE -> docuMineClassificationService.classifyDocument(classificationDocument);
            case CLARIFYND -> clarifyndClassificationService.classifyDocument(classificationDocument);
        }
@@ -5,7 +5,7 @@ import org.springframework.context.annotation.Bean;
 import org.springframework.context.annotation.ComponentScan;
 import org.springframework.context.annotation.Configuration;

-import com.knecon.fforesight.service.viewerdoc.service.ViewerDocumentService;
+import com.knecon.fforesight.service.viewerdoc.service.PDFTronViewerDocumentService;

 import io.micrometer.observation.ObservationRegistry;

@@ -13,12 +13,11 @@ import io.micrometer.observation.ObservationRegistry;
@ComponentScan
 public class LayoutParsingServiceProcessorConfiguration {

-
    @Bean
    @Autowired
-    public ViewerDocumentService viewerDocumentService(ObservationRegistry registry) {
+    public PDFTronViewerDocumentService viewerDocumentService(ObservationRegistry registry) {

-        return new ViewerDocumentService(registry);
+        return new PDFTronViewerDocumentService(registry);
    }

 }
@@ -1,9 +1,11 @@
 package com.knecon.fforesight.service.layoutparser.processor;

+import java.io.ByteArrayInputStream;
 import java.io.File;
 import java.io.FileInputStream;
 import java.io.IOException;
 import java.io.InputStream;
+import java.nio.charset.StandardCharsets;
 import java.nio.file.Files;
 import java.nio.file.Path;
 import java.nio.file.Paths;
@@ -21,6 +23,7 @@ import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsi
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.image.ImageServiceResponse;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.TableServiceResponse;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.VisualLayoutParsingResponse;
+import com.knecon.fforesight.service.viewerdoc.service.ViewerDocVersioningUtility;
 import com.knecon.fforesight.tenantcommons.TenantContext;

 import io.micrometer.observation.annotation.Observed;
@@ -36,6 +39,7 @@ public class LayoutParsingStorageService {
    private final StorageService storageService;
    private final ObjectMapper objectMapper;

+
    @Observed(name = "LayoutParsingStorageService", contextualName = "get-origin-file")
    public File getOriginFile(String storageId) throws IOException {

@@ -53,11 +57,18 @@ public class LayoutParsingStorageService {
        }
        File tempFile = createTempFile("viewerDocument", ".pdf");
        storageService.downloadTo(TenantContext.getTenantId(), storageId, tempFile);
+
+        if (!ViewerDocVersioningUtility.isCurrentVersion(tempFile)) {
+            assert tempFile.delete();
+            return Optional.empty();
+        }
+
        return Optional.of(tempFile);
    }


-    public ImageServiceResponse getImagesFile(String storageId) throws IOException {
+    @SneakyThrows
+    public ImageServiceResponse getImagesFile(String storageId) {

        try (InputStream inputStream = getObject(storageId)) {

@@ -68,7 +79,8 @@ public class LayoutParsingStorageService {
    }


-    public TableServiceResponse getTablesFile(String storageId) throws IOException {
+    @SneakyThrows
+    public TableServiceResponse getTablesFile(String storageId) {

        try (var tableClassificationStream = getObject(storageId)) {

@@ -78,11 +90,12 @@ public class LayoutParsingStorageService {
        }
    }

-    public VisualLayoutParsingResponse getVisualLayoutParsingFile(String storageId) throws IOException {
+
+@SneakyThrows
+    public VisualLayoutParsingResponse getVisualLayoutParsingFile(String storageId) {

        try (InputStream inputStream = getObject(storageId)) {
-            VisualLayoutParsingResponse visualLayoutParsingResponse = objectMapper.readValue(inputStream, VisualLayoutParsingResponse.class);
-            return visualLayoutParsingResponse;
+            return objectMapper.readValue(inputStream, VisualLayoutParsingResponse.class);
        }
    }

@@ -154,4 +167,16 @@ public class LayoutParsingStorageService {
        }
    }

+
+    @SneakyThrows
+    @Observed(name = "LayoutParsingStorageService", contextualName = "store-markdown-file")
+    public void storeMarkdownFile(String markdownFileStorageId, String markdownContent) {
+
+        try (InputStream inputStream = new ByteArrayInputStream(markdownContent.getBytes(StandardCharsets.UTF_8))) {
+
+            storageService.storeObject(TenantContext.getTenantId(), markdownFileStorageId, inputStream);
+        }
+
+    }
+
 }
@@ -17,4 +17,5 @@ public class LayoutparserSettings {

    boolean debug;
    LayoutParsingType layoutParsingTypeOverride;
+    String pdftronLicense;
 }
@@ -1,6 +1,7 @@
 package com.knecon.fforesight.service.layoutparser.processor.docstrum;

 import java.util.ArrayList;
+import java.util.EnumMap;
 import java.util.List;
 import java.util.stream.Collectors;

@@ -18,7 +19,7 @@ import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRul
 import com.knecon.fforesight.service.layoutparser.processor.model.text.RedTextPosition;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextDirection;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
-import com.knecon.fforesight.service.layoutparser.processor.visualization.LayoutparsingVisualizations;
+import com.knecon.fforesight.service.layoutparser.processor.visualization.LayoutDebugLayer;

 import lombok.RequiredArgsConstructor;

@@ -26,6 +27,7 @@ import lombok.RequiredArgsConstructor;
@RequiredArgsConstructor
 public class DocstrumSegmentationService {

+    public static final double SAME_DIRECTION_THRESHOLD = 0.9;
    private final NearestNeighbourService nearestNeighbourService;
    private final SpacingService spacingService;
    private final LineBuilderService lineBuilderService;
@@ -33,19 +35,50 @@ public class DocstrumSegmentationService {
    private final ReadingOrderService readingOrderService;


-    public List<Zone> segmentPage(List<TextPositionSequence> textPositions, boolean xyOrder, CleanRulings usedRulings, LayoutparsingVisualizations visualizations) {
+    public List<Zone> segmentPage(List<TextPositionSequence> textPositions, boolean xyOrder, CleanRulings usedRulings, LayoutDebugLayer visualizations) {

-        List<Zone> zones = new ArrayList<>();
-        zones.addAll(computeZones(textPositions, usedRulings, visualizations, TextDirection.ZERO));
-        zones.addAll(computeZones(textPositions, usedRulings, visualizations, TextDirection.QUARTER_CIRCLE));
-        zones.addAll(computeZones(textPositions, usedRulings, visualizations, TextDirection.HALF_CIRCLE));
-        zones.addAll(computeZones(textPositions, usedRulings, visualizations, TextDirection.THREE_QUARTER_CIRCLE));
+        EnumMap<TextDirection, Integer> directionCounts = new EnumMap<>(TextDirection.class);

-        return readingOrderService.resolve(zones, xyOrder);
+        List<Zone> newZones = computeZones(textPositions, usedRulings, visualizations, TextDirection.ZERO);
+        directionCounts.put(TextDirection.ZERO, newZones.size());
+        List<Zone> zones = new ArrayList<>(newZones);
+
+        newZones = computeZones(textPositions, usedRulings, visualizations, TextDirection.QUARTER_CIRCLE);
+        directionCounts.put(TextDirection.QUARTER_CIRCLE, newZones.size());
+        zones.addAll(newZones);
+
+        newZones = computeZones(textPositions, usedRulings, visualizations, TextDirection.HALF_CIRCLE);
+        directionCounts.put(TextDirection.HALF_CIRCLE, newZones.size());
+        zones.addAll(newZones);
+
+        newZones = computeZones(textPositions, usedRulings, visualizations, TextDirection.THREE_QUARTER_CIRCLE);
+        directionCounts.put(TextDirection.THREE_QUARTER_CIRCLE, newZones.size());
+        zones.addAll(newZones);
+
+        return readingOrderService.resolve(zones, xyOrder, mostSameDirection(directionCounts));
    }


-    private List<Zone> computeZones(List<TextPositionSequence> textPositions, CleanRulings rulings, LayoutparsingVisualizations visualizations, TextDirection direction) {
+    private boolean mostSameDirection(EnumMap<TextDirection, Integer> directionCounts) {
+
+        int total = directionCounts.values()
+                .stream()
+                .mapToInt(i -> i).sum();
+
+        if ((double) directionCounts.get(TextDirection.ZERO) / total > SAME_DIRECTION_THRESHOLD) {
+            return true;
+        } else if ((double) directionCounts.get(TextDirection.QUARTER_CIRCLE) / total > SAME_DIRECTION_THRESHOLD) {
+            return true;
+        } else if ((double) directionCounts.get(TextDirection.HALF_CIRCLE) / total > SAME_DIRECTION_THRESHOLD) {
+            return true;
+        } else if ((double) directionCounts.get(TextDirection.THREE_QUARTER_CIRCLE) / total > SAME_DIRECTION_THRESHOLD) {
+            return true;
+        }
+        return false;
+    }
+
+
+    private List<Zone> computeZones(List<TextPositionSequence> textPositions, CleanRulings rulings, LayoutDebugLayer visualizations, TextDirection direction) {

        List<RedTextPosition> positions = textPositions.stream()
                .filter(t -> t.getDir() == direction)
@@ -15,10 +15,16 @@ public class AngleFilter {

    public boolean matches(Neighbor neighbor) {

+        return matches(neighbor.getAngle());
+    }
+
+
+    public boolean matches(double angle) {
+
        if (lowerAngle <= upperAngle) {
-            return lowerAngle <= neighbor.getAngle() && neighbor.getAngle() < upperAngle;
+            return lowerAngle <= angle && angle < upperAngle;
        } else {
-            return lowerAngle <= neighbor.getAngle() || neighbor.getAngle() < upperAngle;
+            return lowerAngle <= angle || angle < upperAngle;
        }
    }

@@ -7,8 +7,12 @@ import java.util.List;
 import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;

 import lombok.Data;
+import lombok.NoArgsConstructor;
+import lombok.experimental.SuperBuilder;

@Data
+@SuperBuilder
+@NoArgsConstructor
 public abstract class BoundingBox {

    // Java coordinate system: (0, 0) is always upper left, x is increasing left to right and y is increasing from top to bottom.
@@ -19,7 +23,7 @@ public abstract class BoundingBox {
    // This rotates completely in 90 degree steps with page rotation.
    // Needs to be used when writing to a PDF.
    // Also, these are definitely correct and should be used whenever possible.
-    protected Rectangle2D bBoxInitialUserSpace;
+    protected Rectangle2D bBoxPdf;

    protected static final float VERTICAL_COMPARISON_THRESHOLD = 0.4f;

@@ -50,25 +54,25 @@ public abstract class BoundingBox {

    public double getPdfMinX() {

-        return bBoxInitialUserSpace.getMinX();
+        return bBoxPdf.getMinX();
    }


    public double getPdfMaxX() {

-        return bBoxInitialUserSpace.getMaxX();
+        return bBoxPdf.getMaxX();
    }


    public double getPdfMinY() {

-        return bBoxInitialUserSpace.getMinY();
+        return bBoxPdf.getMinY();
    }


    public double getPdfMaxY() {

-        return bBoxInitialUserSpace.getMaxY();
+        return bBoxPdf.getMaxY();
    }


@@ -129,13 +133,31 @@ public abstract class BoundingBox {
    }


-    public boolean intersectsY(BoundingBox other) {
+    private boolean intersectsX(BoundingBox other, float threshold) {
+
+        return this.getX() - threshold <= other.getMaxX() && this.getMaxX() + threshold >= other.getX();
+    }
+
+
+    public boolean intersectsPdf(BoundingBox other) {
+
+        return this.intersectsXPdf(other) && this.intersectsYPdf(other);
+    }
+
+
+    public boolean intersectsPdf(BoundingBox other, float yThreshold, float xThreshold) {
+
+        return this.intersectsXPdf(other, xThreshold) && this.intersectsYPdf(other, yThreshold);
+    }
+
+
+    public boolean intersectsYPdf(BoundingBox other) {

        return this.getPdfMinY() <= other.getPdfMaxY() && this.getPdfMaxY() >= other.getPdfMinY();
    }


-    public boolean intersectsYJava(BoundingBox other) {
+    public boolean intersectsY(BoundingBox other) {

        return this.getY() <= other.getMaxY() && this.getMaxY() >= other.getY();
    }
@@ -143,25 +165,31 @@ public abstract class BoundingBox {

    public boolean intersectsY(BoundingBox other, float threshold) {

+        return this.getY() - threshold <= other.getMaxY() && this.getMaxY() + threshold >= other.getY();
+    }
+
+
+    public boolean intersectsYPdf(BoundingBox other, float threshold) {
+
        return this.getPdfMinY() - threshold <= other.getPdfMaxY() && this.getPdfMaxY() + threshold >= other.getPdfMinY();
    }


-    public boolean intersectsX(BoundingBox other) {
+    public boolean intersectsXPdf(BoundingBox other) {

        return this.getPdfMinX() <= other.getPdfMaxX() && this.getPdfMaxX() >= other.getPdfMinX();
    }


-    public boolean intersectsXJava(BoundingBox other) {
+    public boolean intersectsX(BoundingBox other) {

        return this.getX() <= other.getMaxX() && this.getMaxX() >= other.getMinX();
    }


-    public boolean intersectsX(BoundingBox other, float threshold) {
+    public boolean intersectsXPdf(BoundingBox other, float threshold) {

-        return this.getPdfMinX() - threshold <= other.getMaxX() && this.getMaxX() + threshold >= other.getPdfMinX();
+        return this.getPdfMinX() - threshold <= other.getPdfMaxX() && this.getMaxX() + threshold >= other.getPdfMinX();
    }


@@ -170,8 +198,8 @@ public abstract class BoundingBox {
        this.bBox = components.stream()
                .map(BoundingBox::getBBox)
                .collect(RectangleTransformations.collectBBox());
-        this.bBoxInitialUserSpace = components.stream()
-                .map(BoundingBox::getBBoxInitialUserSpace)
+        this.bBoxPdf = components.stream()
+                .map(BoundingBox::getBBoxPdf)
                .collect(RectangleTransformations.collectBBox());
    }

@@ -229,25 +257,25 @@ public abstract class BoundingBox {

    public boolean rightOf(BoundingBox other) {

-        return this.intersectsYJava(other) && other.getMaxX() <= this.getMinX();
+        return this.intersectsY(other) && other.getMaxX() <= this.getMinX();
    }


    public boolean leftOf(BoundingBox other) {

-        return this.intersectsYJava(other) && other.getMinX() >= this.getMaxX();
+        return this.intersectsY(other) && other.getMinX() >= this.getMaxX();
    }


    public boolean isAbove(BoundingBox other) {

-        return this.intersectsXJava(other) && other.getMinY() >= this.getMaxY();
+        return this.intersectsX(other) && other.getMinY() >= this.getMaxY();
    }


    public boolean isBelow(BoundingBox other) {

-        return this.intersectsXJava(other) && this.getMinY() >= other.getMaxY();
+        return this.intersectsX(other) && this.getMinY() >= other.getMaxY();
    }

 }
@@ -35,7 +35,7 @@ public class Character {

    public double getHeight() {

-        return textPosition.getHeightDir();
+        return textPosition.getHeightDirAdj();
    }


@@ -65,9 +65,9 @@ public class Character {
        double s = Math.sin(-0);
        double c = Math.cos(-0);
        xs[0] = c * x - s * y;
-        xs[1] = c * (x + textPosition.getWidthDirAdj()) - s * (y + textPosition.getHeightDir());
+        xs[1] = c * (x + textPosition.getWidthDirAdj()) - s * (y + textPosition.getHeightDirAdj());
        xs[2] = c * other.x - s * other.y;
-        xs[3] = c * (other.x + other.textPosition.getWidthDirAdj()) - s * (other.y + other.textPosition.getHeightDir());
+        xs[3] = c * (other.x + other.textPosition.getWidthDirAdj()) - s * (other.y + other.textPosition.getHeightDirAdj());
        boolean overlapping = xs[1] >= xs[2] && xs[3] >= xs[0];
        Arrays.sort(xs);
        return Math.abs(xs[2] - xs[1]) * (overlapping ? 1 : -1);
@@ -1,18 +1,28 @@
 package com.knecon.fforesight.service.layoutparser.processor.docstrum.model;

+import static com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence.BOLD;
+import static com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence.BOLD_ITALIC;
+import static com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence.ITALIC;
+import static com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence.STANDARD;
+
 import java.util.ArrayList;
 import java.util.Arrays;
+import java.util.Comparator;
+import java.util.EnumMap;
 import java.util.List;
+import java.util.Map;
+import java.util.concurrent.atomic.AtomicInteger;

 import com.knecon.fforesight.service.layoutparser.processor.docstrum.utils.FastAtan2;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.FontStyle;

 import lombok.Data;
 import lombok.EqualsAndHashCode;

@Data
-@EqualsAndHashCode(onlyExplicitlyIncluded = true)
-public class Line extends BoundingBox {
+@EqualsAndHashCode(onlyExplicitlyIncluded = true, callSuper = false)
+public class Line extends TextBoundingBox {

    private static final double WORD_DISTANCE_MULTIPLIER = 0.18;

@@ -28,6 +38,8 @@ public class Line extends BoundingBox {

    private final double height;

+    private FontStyle fontStyle;
+
    private final List<Character> characters;
    private final List<TextPositionSequence> words = new ArrayList<>();

@@ -67,6 +79,29 @@ public class Line extends BoundingBox {
        height = computeHeight();
        computeWords(wordSpacing * WORD_DISTANCE_MULTIPLIER);
        buildBBox();
+        computeFontStyle();
+    }
+
+
+    private void computeFontStyle() {
+
+        EnumMap<FontStyle, AtomicInteger> fontStyleCounter = new EnumMap<>(FontStyle.class);
+        for (FontStyle fontStyle : FontStyle.values()) {
+            fontStyleCounter.put(fontStyle, new AtomicInteger(0));
+        }
+        for (TextPositionSequence word : words) {
+            switch (word.getFontStyle()) {
+                case STANDARD -> fontStyleCounter.get(FontStyle.REGULAR).getAndIncrement();
+                case BOLD -> fontStyleCounter.get(FontStyle.BOLD).getAndIncrement();
+                case ITALIC -> fontStyleCounter.get(FontStyle.ITALIC).getAndIncrement();
+                case BOLD_ITALIC -> fontStyleCounter.get(FontStyle.BOLD_ITALIC).getAndIncrement();
+            }
+        }
+        fontStyle = fontStyleCounter.entrySet()
+                .stream()
+                .max(Comparator.comparing(entry -> entry.getValue().get()))
+                .map(Map.Entry::getKey)
+                .orElse(FontStyle.REGULAR);
    }


@@ -144,8 +179,8 @@ public class Line extends BoundingBox {
    private void buildBBox() {

        this.setToBBoxOfComponents(characters.stream()
-                                       .map(Character::getTextPosition)
-                                       .toList());
+                                           .map(Character::getTextPosition)
+                                           .toList());
    }


@@ -0,0 +1,102 @@
+package com.knecon.fforesight.service.layoutparser.processor.docstrum.model;
+
+import java.awt.geom.Rectangle2D;
+import java.util.List;
+import java.util.Set;
+import java.util.stream.Collectors;
+
+import com.knecon.fforesight.service.layoutparser.processor.model.text.TextDirection;
+import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
+
+import lombok.EqualsAndHashCode;
+import lombok.Getter;
+import lombok.NoArgsConstructor;
+import lombok.Setter;
+import lombok.experimental.SuperBuilder;
+
+@Getter
+@Setter
+@SuperBuilder
+@NoArgsConstructor
+@EqualsAndHashCode(callSuper = false)
+public abstract class TextBoundingBox extends BoundingBox {
+
+    protected Rectangle2D bBoxDirAdj;
+
+    protected TextDirection dir;
+
+
+    @Override
+    public void setToBBoxOfComponents(List<? extends BoundingBox> components) {
+
+        super.setToBBoxOfComponents(components);
+        this.bBoxDirAdj = components.stream()
+                .filter(c -> c instanceof TextBoundingBox)
+                .map(c -> (TextBoundingBox) c)
+                .map(TextBoundingBox::getBBoxDirAdj)
+                .collect(RectangleTransformations.collectBBox());
+
+        Set<TextDirection> textDirections = components.stream()
+                .filter(c -> c instanceof TextBoundingBox)
+                .map(c -> (TextBoundingBox) c)
+                .map(TextBoundingBox::getDir)
+                .collect(Collectors.toSet());
+
+        if (textDirections.isEmpty()) {
+            dir = TextDirection.ZERO;
+        } else if (textDirections.size() > 1) {
+            throw new IllegalArgumentException("More than one text direction found");
+        } else {
+            dir = textDirections.iterator().next();
+        }
+    }
+
+
+    public double getXDirAdj() {
+
+        return this.bBoxDirAdj.getX();
+    }
+
+
+    public double getYDirAdj() {
+
+        return this.bBoxDirAdj.getY();
+    }
+
+
+    public double getWidthDirAdj() {
+
+        return this.bBoxDirAdj.getWidth();
+    }
+
+
+    public double getHeightDirAdj() {
+
+        return this.bBoxDirAdj.getHeight();
+    }
+
+
+    public double getMaxXDirAdj() {
+
+        return this.bBoxDirAdj.getMaxX();
+    }
+
+
+    public double getMaxYDirAdj() {
+
+        return this.bBoxDirAdj.getMaxY();
+    }
+
+
+    public double getCenterYDirAdj() {
+
+        return this.bBoxDirAdj.getCenterY();
+    }
+
+
+    public double getCenterXDirAdj() {
+
+        return this.bBoxDirAdj.getCenterX();
+    }
+
+}
@@ -6,9 +6,11 @@ import java.util.List;
 import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;

 import lombok.Data;
+import lombok.EqualsAndHashCode;

@Data
-public class Zone extends BoundingBox {
+@EqualsAndHashCode(callSuper = false)
+public class Zone extends TextBoundingBox {

    private List<Line> lines;

@@ -1,15 +1,17 @@
 package com.knecon.fforesight.service.layoutparser.processor.docstrum.service;

+import java.awt.geom.Rectangle2D;
 import java.util.ArrayList;
 import java.util.Comparator;
 import java.util.HashMap;
 import java.util.List;
 import java.util.ListIterator;
 import java.util.Map;
+import java.util.stream.Collectors;

 import org.springframework.stereotype.Service;

-import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.BoundingBox;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.TextBoundingBox;
 import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Zone;
 import com.knecon.fforesight.service.layoutparser.processor.docstrum.utils.DoubleUtils;

@@ -19,21 +21,30 @@ public class ReadingOrderService {
    private static final double THRESHOLD = 5;
    public static final double MULTI_COLUMN_DETECTION_THRESHOLD = 1.5;

+    private static final Comparator<TextBoundingBox> COMPARATOR = //
+            Comparator.comparing(TextBoundingBox::getY, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD))
+                    .thenComparing(TextBoundingBox::getX, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD));

-    public List<Zone> resolve(List<Zone> zones, boolean xyReadingOrder) {
+    private static final Comparator<TextBoundingBox> COMPARATOR_DIR_ADJ = //
+            Comparator.comparing(TextBoundingBox::getYDirAdj, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD))
+                    .thenComparing(TextBoundingBox::getXDirAdj, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD));
+
+
+    public List<Zone> resolve(List<Zone> zones, boolean xyReadingOrder, boolean useDirAdjCoords) {

        if (zones.isEmpty() || zones.size() == 1) {
            return zones;
        }

        if (xyReadingOrder) {
-            return resolveSingleColumnReadingOrder(zones);
+            return resolveSingleColumnReadingOrder(zones, useDirAdjCoords);
        }

        Map<Long, Integer> histogram = new HashMap<>();
        for (Zone zone : zones) {
-            long minY = Math.round(zone.getBBox().getMinY());
-            long maxY = Math.round(zone.getBBox().getMaxY());
+            Rectangle2D bbox = useDirAdjCoords ? zone.getBBoxDirAdj() : zone.getBBox();
+            long minY = Math.round(bbox.getMinY());
+            long maxY = Math.round(bbox.getMaxY());
            for (long i = minY; i <= maxY; i++) {
                histogram.put(i, histogram.getOrDefault(i, 0) + 1);
            }
@@ -43,24 +54,32 @@ public class ReadingOrderService {
                    .stream()
                    .mapToInt(Integer::intValue).average()
                    .orElse(1) < MULTI_COLUMN_DETECTION_THRESHOLD) {
-            return resolveSingleColumnReadingOrder(zones);
+            return resolveSingleColumnReadingOrder(zones, useDirAdjCoords);
        } else {

-            return resolveMultiColumnReadingOder(zones);
+            return resolveMultiColumnReadingOder(zones, useDirAdjCoords);
        }

    }


-    private static List<Zone> resolveSingleColumnReadingOrder(List<Zone> zones) {
+    private static List<Zone> resolveSingleColumnReadingOrder(List<Zone> zones, boolean useDirAdjCoords) {

-        zones.sort(Comparator.comparing(BoundingBox::getY, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD))
-                           .thenComparing(BoundingBox::getX, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD)));
+        if (useDirAdjCoords) {
+            return zones.stream()
+                    .collect(Collectors.groupingBy(TextBoundingBox::getDir)).values()
+                    .stream()
+                    .flatMap(words -> words.stream()
+                            .sorted(COMPARATOR_DIR_ADJ))
+                    .toList();
+        }
+
+        zones.sort(COMPARATOR);
        return zones;
    }


-    private List<Zone> resolveMultiColumnReadingOder(List<Zone> zones) {
+    private List<Zone> resolveMultiColumnReadingOder(List<Zone> zones, boolean useDirAdjCoords) {

        // Simple reading order resolver for multi column page layout as described here : https://pub.towardsai.net/advanced-rag-02-unveiling-pdf-parsing-b84ae866344e
        // TODO implement a more fancy reading order resolver see https://github.com/BobLd/DocumentLayoutAnalysis/blob/master/README.md#reading-order
@@ -69,11 +88,12 @@ public class ReadingOrderService {
        double maxX = Double.NEGATIVE_INFINITY;

        for (Zone zone : zones) {
-            if (zone.getX() < minX) {
-                minX = zone.getX();
+            Rectangle2D bbox = useDirAdjCoords ? zone.getBBoxDirAdj() : zone.getBBox();
+            if (bbox.getX() < minX) {
+                minX = zone.getXDirAdj();
            }
-            if (zone.getX() + zone.getWidth() > maxX) {
-                maxX = zone.getX() + zone.getWidth();
+            if (bbox.getMaxX() > maxX) {
+                maxX = zone.getMaxXDirAdj();
            }
        }

@@ -82,24 +102,27 @@ public class ReadingOrderService {
        List<Zone> leftOf = new ArrayList<>();
        List<Zone> rightOf = new ArrayList<>();
        List<Zone> middle = new ArrayList<>();
+
        for (Zone zone : zones) {
-            if (zone.getX() < midLineXCoordinate && zone.getX() + zone.getWidth() < midLineXCoordinate) {
+            Rectangle2D bbox = useDirAdjCoords ? zone.getBBoxDirAdj() : zone.getBBox();
+            if (bbox.getX() < midLineXCoordinate && bbox.getX() + bbox.getWidth() < midLineXCoordinate) {
                leftOf.add(zone);
-            } else if (zone.getX() > midLineXCoordinate && zone.getX() + zone.getWidth() > midLineXCoordinate) {
+            } else if (bbox.getX() > midLineXCoordinate && bbox.getX() + bbox.getWidth() > midLineXCoordinate) {
                rightOf.add(zone);
            } else {
                middle.add(zone);
            }
        }

-        leftOf.sort(Comparator.comparing(BoundingBox::getY, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD))
-                            .thenComparing(BoundingBox::getX, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD)));
-
-        rightOf.sort(Comparator.comparing(BoundingBox::getY, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD))
-                             .thenComparing(BoundingBox::getX, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD)));
-
-        middle.sort(Comparator.comparing(BoundingBox::getY, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD))
-                            .thenComparing(BoundingBox::getX, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD)));
+        if (useDirAdjCoords) {
+            leftOf.sort(COMPARATOR_DIR_ADJ);
+            rightOf.sort(COMPARATOR_DIR_ADJ);
+            middle.sort(COMPARATOR_DIR_ADJ);
+        } else {
+            leftOf.sort(COMPARATOR);
+            rightOf.sort(COMPARATOR);
+            middle.sort(COMPARATOR);
+        }
 /*
        List<Zone> leftNotIntersecting = new ArrayList<>();
        for (Zone leftZone : leftOf) {
@@ -151,8 +174,9 @@ public class ReadingOrderService {

        while (itty.hasNext()) {
            Zone current = itty.next();
+            Rectangle2D bbox = useDirAdjCoords ? current.getBBoxDirAdj() : current.getBBox();
            for (int i = 0; i < sortedZones.size(); i++) {
-                if (current.getY() < sortedZones.get(i).getY()) {
+                if (bbox.getY() < sortedZones.get(i).getY()) {
                    sortedZones.add(i, current);
                    itty.remove();
                    break;
@@ -1,5 +1,7 @@
 package com.knecon.fforesight.service.layoutparser.processor.docstrum.service;

+import static com.knecon.fforesight.service.layoutparser.processor.model.SectionIdentifier.numericalIdentifierPattern;
+
 import java.util.ArrayList;
 import java.util.Comparator;
 import java.util.HashSet;
@@ -21,7 +23,7 @@ public class ZoneBuilderService {
    private static final double MIN_HORIZONTAL_DISTANCE_MULTIPLIER = -0.5;
    private static final double MAX_VERTICAL_DISTANCE_MULTIPLIER = 1.2;

-    private static final double MIN_HORIZONTAL_MERGE_DISTANCE_MULTIPLIER = -3.0;
+    private static final double MIN_HORIZONTAL_MERGE_DISTANCE_MULTIPLIER = -7;

    private static final double MAX_VERTICAL_MERGE_DISTANCE_MULTIPLIER = 0.5;

@@ -38,7 +40,7 @@ public class ZoneBuilderService {

        double minHorizontalDistance = characterSpacing * MIN_HORIZONTAL_DISTANCE_MULTIPLIER;
        double maxVerticalDistance = lineSpacing * MAX_VERTICAL_DISTANCE_MULTIPLIER;
-        double minHorizontalMergeDistance = characterSpacing * MIN_HORIZONTAL_MERGE_DISTANCE_MULTIPLIER;
+        double minHorizontalMergeDistance = lineSpacing * MIN_HORIZONTAL_MERGE_DISTANCE_MULTIPLIER;
        double maxVerticalMergeDistance = lineSpacing * MAX_VERTICAL_MERGE_DISTANCE_MULTIPLIER;

        UnionFind<Line> unionFind = new UnionFind<>(new HashSet<>(lines));
@@ -54,11 +56,26 @@ public class ZoneBuilderService {
                    return;
                }

-                double scale = Math.min(outerLine.getHeight(), innerLine.getHeight()) / meanHeight;
-                scale = Math.max(MIN_LINE_SIZE_SCALE, Math.min(scale, MAX_LINE_SIZE_SCALE));
+//                if (!innerLine.getFontStyle().equals(outerLine.getFontStyle()) //
+//                    && !outerLine.intersectsY(innerLine, -2f)) {
+//                    return;
+//                }

-                double horizontalDistance = outerLine.horizontalDistance(innerLine) / scale;
-                double verticalDistance = outerLine.verticalDistance(innerLine) / scale;
+                double horizontalScale = Math.min(outerLine.getHeightDirAdj(), innerLine.getHeightDirAdj()) / meanHeight;
+                horizontalScale = Math.max(MIN_LINE_SIZE_SCALE, Math.min(horizontalScale, MAX_LINE_SIZE_SCALE));
+                double verticalScale = horizontalScale;
+
+//                if (innerLine.toString().endsWith(":")
+//                    || outerLine.toString().endsWith(":")
+//                    || numericalIdentifierPattern.matcher(innerLine.toString()).matches()
+//                    || numericalIdentifierPattern.matcher(outerLine.toString()).matches()) {
+//
+//                    horizontalScale *= 5;
+//                    verticalScale /= 10;
+//                }
+
+                double horizontalDistance = outerLine.horizontalDistance(innerLine) / horizontalScale;
+                double verticalDistance = outerLine.verticalDistance(innerLine) / verticalScale;

                if ((!(minHorizontalDistance <= horizontalDistance) || !(verticalDistance <= maxVerticalDistance)) //
                    && (!(minHorizontalMergeDistance <= horizontalDistance) || !(verticalDistance <= maxVerticalMergeDistance))) {
@@ -87,7 +104,7 @@ public class ZoneBuilderService {
        double weights = 0.0;
        for (Line line : lines) {
            double weight = line.getLength();
-            meanHeight += line.getHeight() * weight;
+            meanHeight += line.getHeightDirAdj() * weight;
            weights += weight;
        }
        meanHeight /= weights;
@@ -7,9 +7,12 @@ public class DoubleUtils {
        if (Double.isNaN(d1) || Double.isNaN(d2)) {
            return Double.compare(d1, d2);
        }
-        long i1 = Math.round(d1 / (precision == 0 ? 1 : precision));
-        long i2 = Math.round(d2 / (precision == 0 ? 1 : precision));
-        return Long.compare(i1, i2);
+
+        if (Math.abs(d1 - d2) < precision) {
+            return 0;
+        }
+
+        return Double.compare(d1, d2);
    }

 }
@@ -0,0 +1,326 @@
+package com.knecon.fforesight.service.layoutparser.processor.markdown;
+
+import java.util.ArrayList;
+import java.util.HashSet;
+import java.util.LinkedList;
+import java.util.List;
+import java.util.Map;
+import java.util.Set;
+import java.util.TreeMap;
+
+import org.commonmark.Extension;
+import org.commonmark.ext.gfm.tables.TableBlock;
+import org.commonmark.ext.gfm.tables.TableBody;
+import org.commonmark.ext.gfm.tables.TableCell;
+import org.commonmark.ext.gfm.tables.TableHead;
+import org.commonmark.ext.gfm.tables.TableRow;
+import org.commonmark.ext.gfm.tables.TablesExtension;
+import org.commonmark.node.Block;
+import org.commonmark.node.Document;
+import org.commonmark.node.Emphasis;
+import org.commonmark.node.HardLineBreak;
+import org.commonmark.node.Heading;
+import org.commonmark.node.Node;
+import org.commonmark.node.Paragraph;
+import org.commonmark.node.StrongEmphasis;
+import org.commonmark.node.Text;
+import org.commonmark.renderer.markdown.MarkdownRenderer;
+
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.AbstractNodeVisitor;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.TextRange;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Headline;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.SemanticNode;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Table;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.FontStyle;
+
+public class MarkdownMapper extends AbstractNodeVisitor {
+
+    Document markdownDocument = new Document();
+
+
+    public String toMarkdownContent(com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document document) {
+
+        visit(document);
+
+        return buildRenderer().render(this.markdownDocument);
+    }
+
+
+    @Override
+    public void visit(Headline headline) {
+
+        markdownDocument.appendChild(parseHeadline(headline));
+    }
+
+
+    @Override
+    public void visit(com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Paragraph paragraph) {
+
+        markdownDocument.appendChild(parseParagraph(paragraph));
+    }
+
+
+    @Override
+    public void visit(Table table) {
+
+        markdownDocument.appendChild(parseTable(table));
+    }
+
+
+    private static MarkdownRenderer buildRenderer() {
+
+        List<Extension> extensions = List.of(TablesExtension.create());
+        return MarkdownRenderer.builder().extensions(extensions).build();
+    }
+
+
+    private Block parseTable(Table table) {
+
+//        if (table.getNumberOfRows() == 1 && table.getNumberOfCols() == 1) {
+//            org.commonmark.node.Paragraph markdownParagraph = new org.commonmark.node.Paragraph();
+//            parseTextBlock(table.getTextBlock(), true).forEach(markdownParagraph::appendChild);
+//            return markdownParagraph;
+//        }
+
+        TableBlock tableNode = new TableBlock();
+        TableHead head = new TableHead();
+        TableRow tableRow = createTableRow(table, 0);
+        head.appendChild(tableRow);
+        int row = 1;
+        tableNode.appendChild(head);
+        TableBody tableBody = new TableBody();
+        for (; row < table.getNumberOfRows(); row++) {
+            tableBody.appendChild(createTableRow(table, row));
+        }
+        tableNode.appendChild(tableBody);
+        return tableNode;
+    }
+
+
+    private TableRow createTableRow(Table table, int row) {
+
+        TableRow tableRow = new TableRow();
+        table.streamRow(row)
+                .map(this::createTableCell)
+                .forEach(tableRow::appendChild);
+        return tableRow;
+    }
+
+
+    private Node createTableCell(com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.TableCell tc) {
+
+        var cell = new TableCell();
+        List<SemanticNode> childNodes = tc.streamChildren()
+                .toList();
+        if (childNodes.isEmpty()) {
+            parseTextBlock(tc.getTextBlock(), false).forEach(cell::appendChild);
+        } else {
+            childNodes.forEach(semanticNode -> parseTextBlock(semanticNode.getTextBlock(), false).forEach(cell::appendChild));
+        }
+        return cell;
+    }
+
+
+    private Paragraph parseParagraph(com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Paragraph paragraph) {
+
+        org.commonmark.node.Paragraph markdownParagraph = new org.commonmark.node.Paragraph();
+        parseTextBlock(paragraph.getTextBlock(), true).forEach(markdownParagraph::appendChild);
+        return markdownParagraph;
+    }
+
+
+    private Heading parseHeadline(Headline headline) {
+
+        Heading heading = new Heading();
+        heading.setLevel(headline.getTreeId().size());
+        heading.appendChild(parseTextBlockAsText(headline.getTextBlock()));
+        return heading;
+
+    }
+
+
+    private Text parseTextBlockAsText(TextBlock textBlock) {
+
+        return new Text(textBlock.getSearchText());
+    }
+
+
+    private List<Node> parseTextBlock(TextBlock textBlock, boolean includeLineBreaks) {
+
+        LinkedList<Node> result = new LinkedList<>();
+        List<TextRangeWithTextType> textRanges = mergeTextStyles(textBlock);
+
+        String fullText = getText(textBlock, textBlock.getTextRange(), includeLineBreaks);
+        List<Integer> lineTextSizes = getLineTextSizes(fullText);
+        int idx = 0;
+        int charCount = 0;
+        for (TextRangeWithTextType textRange : textRanges) {
+            String text = getText(textBlock, textRange.textRange(), includeLineBreaks);
+            String[] lines = text.split("\n");
+            for (String line : lines) {
+                charCount += line.length();
+                switch (textRange.fontStyle()) {
+                    case REGULAR -> result.add(new Text(line));
+                    case BOLD -> {
+                        StrongEmphasis boldBlock = new StrongEmphasis();
+                        boldBlock.appendChild(new Text(line));
+                        result.add(boldBlock);
+                    }
+                    case ITALIC -> {
+                        Emphasis italicBlock = new Emphasis();
+                        italicBlock.appendChild(new Text(line));
+                        result.add(italicBlock);
+                    }
+                    case BOLD_ITALIC -> {
+                        Emphasis italicBlock = new Emphasis();
+
+                        StrongEmphasis boldBlock = new StrongEmphasis();
+                        boldBlock.appendChild(new Text(line));
+
+                        italicBlock.appendChild(boldBlock);
+                        result.add(italicBlock);
+                    }
+                }
+                if (includeLineBreaks && lineTextSizes.get(idx).equals(charCount)) {
+                    result.add(new HardLineBreak());
+                    idx++;
+                }
+
+            }
+        }
+        if (!result.isEmpty() && result.getLast() instanceof HardLineBreak) {
+            result.removeLast();
+        }
+        return result;
+    }
+
+
+    private static List<Integer> getLineTextSizes(String input) {
+
+        String[] parts = input.split("\n");
+        List<Integer> textSizes = new ArrayList<>();
+
+        int size = 0;
+        for (int i = 0; i < parts.length; i++) {
+            size += parts[i].length();
+            textSizes.add(size);
+        }
+
+        if (textSizes.isEmpty()) {
+            textSizes.add(0);
+        }
+
+        return textSizes;
+    }
+
+
+    private static String getText(TextBlock textBlock, TextRange textRange, boolean includeLineBreaks) {
+
+        return includeLineBreaks ? textBlock.subSequenceWithLineBreaks(textRange) : textBlock.subSequence(textRange).toString();
+    }
+
+
+    private List<TextRangeWithTextType> mergeTextStyles(TextBlock textBlock) {
+
+        List<TextRangeWithTextType> result = new ArrayList<>();
+
+        TreeMap<Integer, Set<FontStyleChange>> styleChanges = new TreeMap<>();
+
+        int start = textBlock.getTextRange().start();
+        int end = textBlock.getTextRange().end();
+
+        for (TextRange bold : textBlock.getBoldTextBoundaries()) {
+            styleChanges.computeIfAbsent(bold.start() + start, k -> new HashSet<>()).add(FontStyleChange.enter(FontStyle.BOLD));
+            styleChanges.computeIfAbsent(bold.end() + start, k -> new HashSet<>()).add(FontStyleChange.leave(FontStyle.BOLD));
+        }
+
+        for (TextRange italic : textBlock.getItalicTextBoundaries()) {
+            styleChanges.computeIfAbsent(italic.start() + start, k -> new HashSet<>()).add(FontStyleChange.enter(FontStyle.ITALIC));
+            styleChanges.computeIfAbsent(italic.end() + start, k -> new HashSet<>()).add(FontStyleChange.leave(FontStyle.ITALIC));
+        }
+
+        if (styleChanges.isEmpty()) {
+            result.add(new TextRangeWithTextType(new TextRange(start, end), FontStyle.REGULAR));
+            return result;
+        }
+
+        Set<FontStyle> currentStyles = new HashSet<>();
+        currentStyles.add(FontStyle.REGULAR);
+
+        for (Map.Entry<Integer, Set<FontStyleChange>> entry : styleChanges.entrySet()) {
+            int point = entry.getKey();
+            Set<FontStyleChange> changes = entry.getValue();
+
+            if (point > start) {
+                FontStyle style = determineFontStyle(currentStyles);
+                result.add(new TextRangeWithTextType(new TextRange(start, point), style));
+            }
+
+            changes.stream()
+                    .filter(FontStyleChange::leave)
+                    .map(FontStyleChange::style)
+                    .toList()
+                    .forEach(currentStyles::remove);
+
+            currentStyles.addAll(changes.stream()
+                                         .filter(FontStyleChange::enter)
+                                         .map(FontStyleChange::style)
+                                         .toList());
+
+            if (currentStyles.isEmpty()) {
+                currentStyles.add(FontStyle.REGULAR);
+            }
+
+            start = point;
+        }
+
+        if (start < end) {
+            FontStyle style = determineFontStyle(currentStyles);
+            result.add(new TextRangeWithTextType(new TextRange(start, textBlock.getTextRange().end()), style));
+        }
+
+        return result;
+    }
+
+
+    private FontStyle determineFontStyle(Set<FontStyle> styles) {
+
+        if (styles.contains(FontStyle.BOLD) && styles.contains(FontStyle.ITALIC)) {
+            return FontStyle.BOLD_ITALIC;
+        } else if (styles.contains(FontStyle.BOLD)) {
+            return FontStyle.BOLD;
+        } else if (styles.contains(FontStyle.ITALIC)) {
+            return FontStyle.ITALIC;
+        } else {
+            return FontStyle.REGULAR;
+        }
+    }
+
+
+
+    record FontStyleChange(boolean enter, FontStyle style) {
+
+        public static FontStyleChange enter(FontStyle style) {
+
+            return new FontStyleChange(true, style);
+        }
+
+
+        public static FontStyleChange leave(FontStyle style) {
+
+            return new FontStyleChange(false, style);
+        }
+
+
+        public boolean leave() {
+
+            return !enter;
+        }
+
+    }
+
+    record TextRangeWithTextType(TextRange textRange, FontStyle fontStyle) {
+
+    }
+}
@@ -7,7 +7,7 @@ import com.knecon.fforesight.service.layoutparser.processor.model.outline.Outlin
 import com.knecon.fforesight.service.layoutparser.processor.model.outline.TableOfContents;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.StringFrequencyCounter;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.UnclassifiedText;
-import com.knecon.fforesight.service.layoutparser.processor.visualization.LayoutparsingVisualizations;
+import com.knecon.fforesight.service.layoutparser.processor.visualization.LayoutDebugLayer;

 import lombok.Data;
 import lombok.NoArgsConstructor;
@@ -25,7 +25,7 @@ public class ClassificationDocument {
    private FloatFrequencyCounter fontSizeCounter = new FloatFrequencyCounter();
    private StringFrequencyCounter fontCounter = new StringFrequencyCounter();
    private StringFrequencyCounter fontStyleCounter = new StringFrequencyCounter();
-    private LayoutparsingVisualizations visualizations = new LayoutparsingVisualizations();
+    private LayoutDebugLayer layoutDebugLayer = new LayoutDebugLayer();
    private boolean headlines;

    private long rulesVersion;
@@ -18,6 +18,7 @@ import lombok.RequiredArgsConstructor;

@Data
@RequiredArgsConstructor
+
 public class ClassificationPage {

    @NonNull
@@ -25,7 +26,7 @@ public class ClassificationPage {

    private List<OutlineObject> outlineObjects = new ArrayList<>();

-    private List<AbstractPageBlock> headlines  = new ArrayList<>();
+    private List<AbstractPageBlock> headlines = new ArrayList<>();

    private List<ClassifiedImage> images = new ArrayList<>();

@@ -44,7 +45,7 @@ public class ClassificationPage {
    private float pageWidth;
    private float pageHeight;

-    CleanRulings cleanRulings;
+    private CleanRulings cleanRulings;

    private Map<String, List<Rectangle2D>> markedContentBboxPerType = new HashMap<>();

@@ -12,10 +12,10 @@ import lombok.Getter;
@Getter
 public class FloatFrequencyCounter {

-    Map<Float, Integer> countPerValue = new HashMap<>();
+    Map<Double, Integer> countPerValue = new HashMap<>();


-    public void add(float value) {
+    public void add(double value) {

        if (!countPerValue.containsKey(value)) {
            countPerValue.put(value, 1);
@@ -25,9 +25,9 @@ public class FloatFrequencyCounter {
    }


-    public void addAll(Map<Float, Integer> otherCounter) {
+    public void addAll(Map<Double, Integer> otherCounter) {

-        for (Map.Entry<Float, Integer> entry : otherCounter.entrySet()) {
+        for (Map.Entry<Double, Integer> entry : otherCounter.entrySet()) {
            if (countPerValue.containsKey(entry.getKey())) {
                countPerValue.put(entry.getKey(), countPerValue.get(entry.getKey()) + entry.getValue());
            } else {
@@ -37,10 +37,10 @@ public class FloatFrequencyCounter {
    }


-    public Float getMostPopular() {
+    public Double getMostPopular() {

-        Map.Entry<Float, Integer> mostPopular = null;
-        for (Map.Entry<Float, Integer> entry : countPerValue.entrySet()) {
+        Map.Entry<Double, Integer> mostPopular = null;
+        for (Map.Entry<Double, Integer> entry : countPerValue.entrySet()) {
            if (mostPopular == null || entry.getValue() >= mostPopular.getValue()) {
                mostPopular = entry;
            }
@@ -49,11 +49,11 @@ public class FloatFrequencyCounter {
    }


-    public List<Float> getHighterThanMostPopular() {
+    public List<Double> getHigherThanMostPopular() {

-        Float mostPopular = getMostPopular();
-        List<Float> higher = new ArrayList<>();
-        for (Float value : countPerValue.keySet()) {
+        Double mostPopular = getMostPopular();
+        List<Double> higher = new ArrayList<>();
+        for (Double value : countPerValue.keySet()) {
            if (value > mostPopular) {
                higher.add(value);
            }
@@ -63,10 +63,10 @@ public class FloatFrequencyCounter {
    }


-    public Float getHighest() {
+    public Double getHighest() {

-        Float highest = null;
-        for (Float value : countPerValue.keySet()) {
+        Double highest = null;
+        for (Double value : countPerValue.keySet()) {
            if (highest == null || value > highest) {
                highest = value;
            }
@@ -15,7 +15,7 @@ import lombok.experimental.FieldDefaults;
@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
 public class SectionIdentifier {

-    static Pattern numericalIdentifierPattern = Pattern.compile("^[\\s]?(\\d+)[\\s.,;]?(\\d+)?[\\s.,;]?(\\d+)?[\\s.,;]?(\\d+)?");
+    public static Pattern numericalIdentifierPattern = Pattern.compile("^[\\s]?(\\d+)[\\s.,;]?(\\d+)?[\\s.,;]?(\\d+)?[\\s.,;]?(\\d+)?");

    public enum Format {
        EMPTY,
@@ -0,0 +1,94 @@
+package com.knecon.fforesight.service.layoutparser.processor.model.graph;
+
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Footer;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Header;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Headline;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Image;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Paragraph;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Section;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.SemanticNode;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.SuperSection;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Table;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.TableCell;
+
+public abstract class AbstractNodeVisitor implements NodeVisitor {
+
+    @Override
+    public void visit(Document document) {
+
+        visitChildren(document);
+    }
+
+
+    @Override
+    public void visit(SuperSection superSection) {
+
+        visitChildren(superSection);
+    }
+
+
+    @Override
+    public void visit(Section section) {
+
+        visitChildren(section);
+    }
+
+
+    @Override
+    public void visit(Headline headline) {
+
+        visitChildren(headline);
+    }
+
+
+    @Override
+    public void visit(Paragraph paragraph) {
+
+        visitChildren(paragraph);
+    }
+
+
+    @Override
+    public void visit(Footer footer) {
+
+        visitChildren(footer);
+    }
+
+
+    @Override
+    public void visit(Header header) {
+
+        visitChildren(header);
+    }
+
+
+    @Override
+    public void visit(Image image) {
+
+        visitChildren(image);
+    }
+
+
+    @Override
+    public void visit(Table table) {
+
+        visitChildren(table);
+    }
+
+
+    @Override
+    public void visit(TableCell tableCell) {
+
+        visitChildren(tableCell);
+    }
+
+
+    private void visitChildren(SemanticNode semanticNode) {
+
+        semanticNode.streamChildren()
+                .forEach(node -> node.accept(this));
+
+    }
+
+}
@@ -2,6 +2,7 @@ package com.knecon.fforesight.service.layoutparser.processor.model.graph;

 import static java.lang.String.format;

+import java.util.ArrayList;
 import java.util.Collections;
 import java.util.LinkedList;
 import java.util.List;
@@ -39,7 +40,10 @@ public class DocumentTree {

    public TextBlock buildTextBlock() {

-        return allEntriesInOrder().map(Entry::getNode).filter(SemanticNode::isLeaf).map(SemanticNode::getLeafTextBlock).collect(new TextBlockCollector());
+        return allEntriesInOrder().map(Entry::getNode)
+                .filter(SemanticNode::isLeaf)
+                .map(SemanticNode::getLeafTextBlock)
+                .collect(new TextBlockCollector());
    }


@@ -113,13 +117,16 @@ public class DocumentTree {

    public Stream<SemanticNode> childNodes(List<Integer> treeId) {

-        return getEntryById(treeId).children.stream().map(Entry::getNode);
+        return getEntryById(treeId).children.stream()
+                .map(Entry::getNode);
    }


    public Stream<SemanticNode> childNodesOfType(List<Integer> treeId, NodeType nodeType) {

-        return getEntryById(treeId).children.stream().filter(entry -> entry.node.getType().equals(nodeType)).map(Entry::getNode);
+        return getEntryById(treeId).children.stream()
+                .filter(entry -> entry.node.getType().equals(nodeType))
+                .map(Entry::getNode);
    }


@@ -156,26 +163,32 @@ public class DocumentTree {

    public Stream<Entry> allEntriesInOrder() {

-        return Stream.of(root).flatMap(DocumentTree::flatten);
+        return Stream.of(root)
+                .flatMap(DocumentTree::flatten);
    }


    public Stream<Entry> allSubEntriesInOrder(List<Integer> parentId) {

-        return getEntryById(parentId).children.stream().flatMap(DocumentTree::flatten);
+        return getEntryById(parentId).children.stream()
+                .flatMap(DocumentTree::flatten);
    }


    @Override
    public String toString() {

-        return String.join("\n", allEntriesInOrder().map(Entry::toString).toList());
+        return String.join("\n",
+                           allEntriesInOrder().map(Entry::toString)
+                                   .toList());
    }


    private static Stream<Entry> flatten(Entry entry) {

-        return Stream.concat(Stream.of(entry), entry.children.stream().flatMap(DocumentTree::flatten));
+        return Stream.concat(Stream.of(entry),
+                             entry.children.stream()
+                                     .flatMap(DocumentTree::flatten));
    }


@@ -0,0 +1,45 @@
+package com.knecon.fforesight.service.layoutparser.processor.model.graph;
+
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Footer;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Header;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Headline;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Image;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Paragraph;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Section;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.SuperSection;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Table;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.TableCell;
+
+public interface NodeVisitor {
+
+    void visit(Document document);
+
+
+    void visit(SuperSection superSection);
+
+
+    void visit(Section section);
+
+
+    void visit(Headline headline);
+
+
+    void visit(Paragraph paragraph);
+
+
+    void visit(Footer footer);
+
+
+    void visit(Header header);
+
+
+    void visit(Image image);
+
+
+    void visit(Table table);
+
+
+    void visit(TableCell tableCell);
+
+}
@@ -13,13 +13,13 @@ import lombok.Setter;
@Setter
@EqualsAndHashCode
@SuppressWarnings("PMD.AvoidFieldNameMatchingMethodName")
-public class Boundary implements Comparable<Boundary> {
+public class TextRange implements Comparable<TextRange> {

    private int start;
    private int end;


-    public Boundary(int start, int end) {
+    public TextRange(int start, int end) {

        if (start > end) {
            throw new IllegalArgumentException(format("start: %d > end: %d", start, end));
@@ -47,15 +47,15 @@ public class Boundary implements Comparable<Boundary> {
    }


-    public boolean contains(Boundary boundary) {
+    public boolean contains(TextRange textRange) {

-        return start <= boundary.start() && boundary.end() <= end;
+        return start <= textRange.start() && textRange.end() <= end;
    }


-    public boolean containedBy(Boundary boundary) {
+    public boolean containedBy(TextRange textRange) {

-        return boundary.contains(this);
+        return textRange.contains(this);
    }


@@ -79,22 +79,33 @@ public class Boundary implements Comparable<Boundary> {

    public boolean contains(int index) {

+        return start <= index && index <= end;
+    }
+
+
+    public boolean containsExclusive(int index) {
+
        return start <= index && index < end;
    }


-    public boolean intersects(Boundary boundary) {
+    public boolean intersects(TextRange textRange) {

-        return boundary.start() < this.end && this.start < boundary.end();
+        return textRange.start() < this.end && this.start < textRange.end();
    }


-    public List<Boundary> split(List<Integer> splitIndices) {
+    public List<TextRange> split(List<Integer> splitIndices) {

-        if (splitIndices.stream().anyMatch(idx -> !this.contains(idx))) {
-            throw new IndexOutOfBoundsException(format("%s splitting indices are out of range for %s", splitIndices.stream().filter(idx -> !this.contains(idx)).toList(), this));
+        if (splitIndices.stream()
+                .anyMatch(idx -> !this.containsExclusive(idx))) {
+            throw new IndexOutOfBoundsException(format("%s splitting indices are out of range for %s",
+                                                       splitIndices.stream()
+                                                               .filter(idx -> !this.containsExclusive(idx))
+                                                               .toList(),
+                                                       this));
        }
-        List<Boundary> splitBoundaries = new LinkedList<>();
+        List<TextRange> splitBoundaries = new LinkedList<>();
        int previousIndex = start;
        for (int splitIndex : splitIndices) {

@@ -102,23 +113,31 @@ public class Boundary implements Comparable<Boundary> {
            if (splitIndex == previousIndex) {
                continue;
            }
-            splitBoundaries.add(new Boundary(previousIndex, splitIndex));
+            splitBoundaries.add(new TextRange(previousIndex, splitIndex));
            previousIndex = splitIndex;
        }
-        splitBoundaries.add(new Boundary(previousIndex, end));
+        splitBoundaries.add(new TextRange(previousIndex, end));
        return splitBoundaries;
    }

+
    public IntStream intStream() {

        return IntStream.range(start, end);
    }

-    public static Boundary merge(Collection<Boundary> boundaries) {

-        int minStart = boundaries.stream().mapToInt(Boundary::start).min().orElseThrow(IllegalArgumentException::new);
-        int maxEnd = boundaries.stream().mapToInt(Boundary::end).max().orElseThrow(IllegalArgumentException::new);
-        return new Boundary(minStart, maxEnd);
+    public static TextRange merge(Collection<TextRange> boundaries) {
+
+        int minStart = boundaries.stream()
+                .mapToInt(TextRange::start)
+                .min()
+                .orElseThrow(IllegalArgumentException::new);
+        int maxEnd = boundaries.stream()
+                .mapToInt(TextRange::end)
+                .max()
+                .orElseThrow(IllegalArgumentException::new);
+        return new TextRange(minStart, maxEnd);
    }


@@ -130,12 +149,12 @@ public class Boundary implements Comparable<Boundary> {


    @Override
-    public int compareTo(Boundary boundary) {
+    public int compareTo(TextRange textRange) {

-        if (end < boundary.end() && start < boundary.start()) {
+        if (end < textRange.end() && start < textRange.start()) {
            return -1;
        }
-        if (start > boundary.start() && end > boundary.end()) {
+        if (start > textRange.start() && end > textRange.end()) {
            return 1;
        }

@@ -11,7 +11,7 @@ import java.util.Map;
 import java.util.Set;

 import com.iqser.red.service.persistence.service.v1.api.shared.model.redactionlog.Engine;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.Boundary;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.TextRange;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Page;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.SemanticNode;
 import com.knecon.fforesight.service.layoutparser.processor.utils.IdBuilder;
@@ -28,11 +28,11 @@ import lombok.experimental.FieldDefaults;
@AllArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE)
@EqualsAndHashCode(onlyExplicitlyIncluded = true)
-public class RedactionEntity {
+public class TextEntity {

    // initial values
    @EqualsAndHashCode.Include
-    final Boundary boundary;
+    final TextRange textRange;
    @EqualsAndHashCode.Include
    final String type;
    @EqualsAndHashCode.Include
@@ -47,7 +47,7 @@ public class RedactionEntity {
    boolean dictionaryEntry;
    boolean dossierDictionaryEntry;
    Set<Engine> engines;
-    Set<RedactionEntity> references;
+    Set<TextEntity> references;
    @Builder.Default
    Deque<Integer> matchedRules = new LinkedList<>();
    String redactionReason;
@@ -66,9 +66,9 @@ public class RedactionEntity {
    SemanticNode deepestFullyContainingNode;


-    public static RedactionEntity initialEntityNode(Boundary boundary, String type, EntityType entityType) {
+    public static TextEntity initialEntityNode(TextRange textRange, String type, EntityType entityType) {

-        return RedactionEntity.builder().type(type).entityType(entityType).boundary(boundary).engines(new HashSet<>()).references(new HashSet<>()).build();
+        return TextEntity.builder().type(type).entityType(entityType).textRange(textRange).engines(new HashSet<>()).references(new HashSet<>()).build();
    }


@@ -132,7 +132,7 @@ public class RedactionEntity {
    public List<RedactionPosition> getRedactionPositionsPerPage() {

        if (redactionPositionsPerPage == null || redactionPositionsPerPage.isEmpty()) {
-            Map<Page, List<Rectangle2D>> rectanglesPerLinePerPage = deepestFullyContainingNode.getTextBlock().getPositionsPerPage(boundary);
+            Map<Page, List<Rectangle2D>> rectanglesPerLinePerPage = deepestFullyContainingNode.getTextBlock().getPositionsPerPage(textRange);

            Page firstPage = rectanglesPerLinePerPage.keySet()
                    .stream()
@@ -155,21 +155,21 @@ public class RedactionEntity {
    }


-    public boolean containedBy(RedactionEntity redactionEntity) {
+    public boolean containedBy(TextEntity textEntity) {

-        return this.boundary.containedBy(redactionEntity.getBoundary());
+        return this.textRange.containedBy(textEntity.getTextRange());
    }


-    public boolean contains(RedactionEntity redactionEntity) {
+    public boolean contains(TextEntity textEntity) {

-        return this.boundary.contains(redactionEntity.getBoundary());
+        return this.textRange.contains(textEntity.getTextRange());
    }


-    public boolean intersects(RedactionEntity redactionEntity) {
+    public boolean intersects(TextEntity textEntity) {

-        return this.boundary.intersects(redactionEntity.getBoundary());
+        return this.textRange.intersects(textEntity.getTextRange());
    }


@@ -185,13 +185,13 @@ public class RedactionEntity {
    }


-    public void addReference(RedactionEntity reference) {
+    public void addReference(TextEntity reference) {

        references.add(reference);
    }


-    public void addReferences(List<RedactionEntity> references) {
+    public void addReferences(List<TextEntity> references) {

        this.references.addAll(references);
    }
@@ -210,7 +210,7 @@ public class RedactionEntity {
        sb.append("Entity[\"");
        sb.append(value);
        sb.append("\", ");
-        sb.append(boundary);
+        sb.append(textRange);
        sb.append(", pages[");
        pages.forEach(page -> {
            sb.append(page.getNumber());
@@ -8,7 +8,7 @@ import java.util.Set;

 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.TextEntity;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;

 import lombok.AccessLevel;
@@ -39,7 +39,7 @@ public abstract class AbstractSemanticNode implements GenericSemanticNode {

    @Builder.Default
    @EqualsAndHashCode.Exclude
-    Set<RedactionEntity> entities = new HashSet<>();
+    Set<TextEntity> entities = new HashSet<>();

    @EqualsAndHashCode.Exclude
    Map<Page, Rectangle2D> bBoxCache;
@@ -11,8 +11,9 @@ import java.util.stream.Stream;

 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
-import com.knecon.fforesight.service.layoutparser.processor.visualization.LayoutparsingVisualizations;
+import com.knecon.fforesight.service.layoutparser.processor.visualization.LayoutDebugLayer;

 import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
@@ -33,7 +34,7 @@ public class Document extends AbstractSemanticNode {
    Set<Page> pages;
    Integer numberOfPages;

-    LayoutparsingVisualizations visualizations;
+    LayoutDebugLayer layoutDebugLayer;


    @Override
@@ -60,8 +61,8 @@ public class Document extends AbstractSemanticNode {
     *
     * @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.
+     * 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() {
@@ -101,8 +102,7 @@ public class Document extends AbstractSemanticNode {
    public Headline getHeadline() {

        return streamAllSubNodesOfType(NodeType.HEADLINE).map(node -> (Headline) node)
-                .findFirst()
-                .orElse(Headline.builder().build());
+                .findFirst().orElse(Headline.builder().build());
    }


@@ -163,4 +163,11 @@ public class Document extends AbstractSemanticNode {
        return bBox;
    }

+
+    @Override
+    public void accept(NodeVisitor visitor) {
+
+        visitor.visit(this);
+    }
+
 }
@@ -1,6 +1,7 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;

 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;

 import lombok.AccessLevel;
@@ -34,6 +35,12 @@ public class Footer extends AbstractSemanticNode {
    }


+    @Override
+    public void accept(NodeVisitor visitor) {
+        visitor.visit(this);
+    }
+
+
    @Override
    public TextBlock getTextBlock() {

@@ -1,6 +1,7 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;

 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;

 import lombok.AccessLevel;
@@ -27,6 +28,13 @@ public class Header extends AbstractSemanticNode {
    }


+    @Override
+    public void accept(NodeVisitor visitor) {
+
+        visitor.visit(this);
+    }
+
+
    @Override
    public NodeType getType() {

@@ -1,6 +1,7 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;

 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;

 import lombok.AccessLevel;
@@ -34,6 +35,13 @@ public class Headline extends AbstractSemanticNode {
    }


+    @Override
+    public void accept(NodeVisitor visitor) {
+
+        visitor.visit(this);
+    }
+
+
    @Override
    public TextBlock getTextBlock() {

@@ -7,6 +7,7 @@ import java.util.Map;
 import java.util.Set;

 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;

 import lombok.AccessLevel;
@@ -28,6 +29,8 @@ public class Image extends AbstractSemanticNode {

    String id;

+    String representationHash;
+
    ImageType imageType;
    boolean transparent;
    Rectangle2D position;
@@ -84,10 +87,29 @@ public class Image extends AbstractSemanticNode {
    }


+    @Override
+    public void accept(NodeVisitor visitor) {
+
+        visitor.visit(this);
+    }
+
+
    @Override
    public boolean isLeaf() {

        return true;
    }

+
+    public double getArea() {
+
+        return position.getWidth() * position.getHeight();
+    }
+
+
+    public boolean isFullPageImage() {
+
+        return imageType.equals(ImageType.OCR) || getArea() >= 0.5 * page.getArea();
+    }
+
 }
@@ -6,7 +6,6 @@ public enum ImageType {
    LOGO,
    FORMULA,
    SIGNATURE,
-
    SIGNATURE_VISUAL,
    OTHER,
    OCR,
@@ -6,7 +6,7 @@ import java.util.List;
 import java.util.Set;

 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.TextEntity;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlockCollector;

@@ -39,7 +39,7 @@ public class Page {

    @Builder.Default
    @EqualsAndHashCode.Exclude
-    Set<RedactionEntity> entities = new HashSet<>();
+    Set<TextEntity> entities = new HashSet<>();

    @Builder.Default
    @EqualsAndHashCode.Exclude
@@ -60,7 +60,10 @@ public class Page {

    public TextBlock getMainBodyTextBlock() {

-        return mainBody.stream().filter(SemanticNode::isLeaf).map(SemanticNode::getLeafTextBlock).collect(new TextBlockCollector());
+        return mainBody.stream()
+                .filter(SemanticNode::isLeaf)
+                .map(SemanticNode::getLeafTextBlock)
+                .collect(new TextBlockCollector());
    }


@@ -84,4 +87,10 @@ public class Page {
        return o instanceof Page && o.hashCode() == this.hashCode();
    }

+
+    public double getArea() {
+
+        return height * width;
+    }
+
 }
@@ -1,6 +1,7 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;

 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;

 import lombok.AccessLevel;
@@ -34,6 +35,13 @@ public class Paragraph extends AbstractSemanticNode {
    }


+    @Override
+    public void accept(NodeVisitor visitor) {
+
+        visitor.visit(this);
+    }
+
+
    @Override
    public TextBlock getTextBlock() {

@@ -1,6 +1,7 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;

 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;

 import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
@@ -26,15 +27,20 @@ public class Section extends AbstractSemanticNode {
    public Headline getHeadline() {

        return streamChildrenOfType(NodeType.HEADLINE).map(node -> (Headline) node)
-                .findFirst()
-                .orElseGet(() -> getParent().getHeadline());
+                .findFirst().orElseGet(() -> getParent().getHeadline());
+    }
+
+
+    @Override
+    public void accept(NodeVisitor visitor) {
+
+        visitor.visit(this);
    }


    public boolean hasTables() {

-        return streamAllSubNodesOfType(NodeType.TABLE).findAny()
-                .isPresent();
+        return streamAllSubNodesOfType(NodeType.TABLE).findAny().isPresent();
    }


@@ -14,14 +14,16 @@ import java.util.stream.Stream;

 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.Boundary;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.TextRange;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.EntityType;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.TextEntity;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.AtomicTextBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlockCollector;
+import com.knecon.fforesight.service.layoutparser.processor.utils.BBoxMergingUtility;
 import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;

 public interface SemanticNode {

@@ -42,7 +44,9 @@ public interface SemanticNode {
     */
    default TextBlock getTextBlock() {

-        return streamAllSubNodes().filter(SemanticNode::isLeaf).map(SemanticNode::getTextBlock).collect(new TextBlockCollector());
+        return streamAllSubNodes().filter(SemanticNode::isLeaf)
+                .map(SemanticNode::getTextBlock)
+                .collect(new TextBlockCollector());
    }


@@ -52,7 +56,7 @@ public interface SemanticNode {
     *
     * @return Set of all Entities associated with this Node
     */
-    Set<RedactionEntity> getEntities();
+    Set<TextEntity> getEntities();


    /**
@@ -68,7 +72,9 @@ public interface SemanticNode {

    default Page getFirstPage() {

-        return getTextBlock().getPages().stream().min(Comparator.comparingInt(Page::getNumber)).orElseThrow(() -> new IllegalStateException("SemanticNode has no Page!"));
+        return getTextBlock().getPages()
+                .stream()
+                .min(Comparator.comparingInt(Page::getNumber)).orElseThrow(() -> new IllegalStateException("SemanticNode has no Page!"));
    }


@@ -77,18 +83,19 @@ public interface SemanticNode {
     *
     * @return Set of PageNodes this node appears on.
     */
-    default Set<Page> getPages(Boundary boundary) {
+    default Set<Page> getPages(TextRange textRange) {

-        if (!getBoundary().contains(boundary)) {
-            throw new IllegalArgumentException(format("%s which was used to query for pages is not contained in the %s of this node!", boundary, getBoundary()));
+        if (!getBoundary().contains(textRange)) {
+            throw new IllegalArgumentException(format("%s which was used to query for pages is not contained in the %s of this node!", textRange, getBoundary()));
        }
-        return getTextBlock().getPages(boundary);
+        return getTextBlock().getPages(textRange);
    }


    default boolean isOnPage(int pageNumber) {

-        return getPages().stream().anyMatch(page -> page.getNumber() == pageNumber);
+        return getPages().stream()
+                .anyMatch(page -> page.getNumber() == pageNumber);
    }


@@ -203,7 +210,9 @@ public interface SemanticNode {
     */
    default boolean hasEntitiesOfType(String type) {

-        return getEntities().stream().filter(entity -> entity.getEntityType().equals(EntityType.ENTITY)).anyMatch(redactionEntity -> redactionEntity.getType().equals(type));
+        return getEntities().stream()
+                .filter(entity -> entity.getEntityType().equals(EntityType.ENTITY))
+                .anyMatch(redactionEntity -> redactionEntity.getType().equals(type));
    }


@@ -213,9 +222,11 @@ public interface SemanticNode {
     * @param type string representing the type of entities to return
     * @return List of RedactionEntities of any the type
     */
-    default List<RedactionEntity> getEntitiesOfType(String type) {
+    default List<TextEntity> getEntitiesOfType(String type) {

-        return getEntities().stream().filter(redactionEntity -> redactionEntity.getType().equals(type)).toList();
+        return getEntities().stream()
+                .filter(redactionEntity -> redactionEntity.getType().equals(type))
+                .toList();
    }


@@ -225,9 +236,11 @@ public interface SemanticNode {
     * @param types A list of strings representing the types of entities to return
     * @return List of RedactionEntities of any provided type
     */
-    default List<RedactionEntity> getEntitiesOfType(List<String> types) {
+    default List<TextEntity> getEntitiesOfType(List<String> types) {

-        return getEntities().stream().filter(redactionEntity -> redactionEntity.isAnyType(types)).toList();
+        return getEntities().stream()
+                .filter(redactionEntity -> redactionEntity.isAnyType(types))
+                .toList();
    }


@@ -279,7 +292,8 @@ public interface SemanticNode {
     */
    default boolean containsStrings(List<String> strings) {

-        return strings.stream().allMatch(this::containsString);
+        return strings.stream()
+                .allMatch(this::containsString);
    }


@@ -303,7 +317,8 @@ public interface SemanticNode {
     */
    default boolean containsAnyString(List<String> strings) {

-        return strings.stream().anyMatch(this::containsString);
+        return strings.stream()
+                .anyMatch(this::containsString);
    }


@@ -315,7 +330,8 @@ public interface SemanticNode {
     */
    default boolean containsAnyStringIgnoreCase(List<String> strings) {

-        return strings.stream().anyMatch(this::containsStringIgnoreCase);
+        return strings.stream()
+                .anyMatch(this::containsStringIgnoreCase);
    }


@@ -323,19 +339,19 @@ public interface SemanticNode {
     * This function is used during insertion of EntityNodes into the graph, it checks if the boundary of the RedactionEntity intersects or even contains the RedactionEntity.
     * It sets the fields accordingly and recursively calls this function on all its children.
     *
-     * @param redactionEntity RedactionEntity, which is being inserted into the graph
+     * @param textEntity RedactionEntity, which is being inserted into the graph
     */
-    default void addThisToEntityIfIntersects(RedactionEntity redactionEntity) {
+    default void addThisToEntityIfIntersects(TextEntity textEntity) {

        TextBlock textBlock = getTextBlock();
-        if (textBlock.getBoundary().intersects(redactionEntity.getBoundary())) {
-            if (textBlock.containsBoundary(redactionEntity.getBoundary())) {
-                redactionEntity.setDeepestFullyContainingNode(this);
+        if (textBlock.getTextRange().intersects(textEntity.getTextRange())) {
+            if (textBlock.containsBoundary(textEntity.getTextRange())) {
+                textEntity.setDeepestFullyContainingNode(this);
            }

-            redactionEntity.addIntersectingNode(this);
-            streamChildren().filter(semanticNode -> semanticNode.getBoundary().intersects(redactionEntity.getBoundary()))
-                    .forEach(node -> node.addThisToEntityIfIntersects(redactionEntity));
+            textEntity.addIntersectingNode(this);
+            streamChildren().filter(semanticNode -> semanticNode.getBoundary().intersects(textEntity.getTextRange()))
+                    .forEach(node -> node.addThisToEntityIfIntersects(textEntity));
        }
    }

@@ -386,7 +402,8 @@ public interface SemanticNode {
     */
    default Stream<SemanticNode> streamAllSubNodes() {

-        return getDocumentTree().allSubEntriesInOrder(getTreeId()).map(DocumentTree.Entry::getNode);
+        return getDocumentTree().allSubEntriesInOrder(getTreeId())
+                .map(DocumentTree.Entry::getNode);
    }


@@ -397,7 +414,9 @@ public interface SemanticNode {
     */
    default Stream<SemanticNode> streamAllSubNodesOfType(NodeType nodeType) {

-        return getDocumentTree().allSubEntriesInOrder(getTreeId()).filter(entry -> entry.getType().equals(nodeType)).map(DocumentTree.Entry::getNode);
+        return getDocumentTree().allSubEntriesInOrder(getTreeId())
+                .filter(entry -> entry.getType().equals(nodeType))
+                .map(DocumentTree.Entry::getNode);
    }


@@ -406,9 +425,9 @@ public interface SemanticNode {
     *
     * @return Boundary of this Node's TextBlock
     */
-    default Boundary getBoundary() {
+    default TextRange getBoundary() {

-        return getTextBlock().getBoundary();
+        return getTextBlock().getTextRange();
    }


@@ -453,17 +472,19 @@ public interface SemanticNode {
     */
    private Map<Page, Rectangle2D> getBBoxFromChildren() {

-        Map<Page, Rectangle2D> bBoxPerPage = new HashMap<>();
-        List<Map<Page, Rectangle2D>> childrenBBoxes = streamChildren().map(SemanticNode::getBBox).toList();
-        Set<Page> pages = childrenBBoxes.stream().flatMap(map -> map.keySet().stream()).collect(Collectors.toSet());
-        for (Page page : pages) {
-            Rectangle2D bBoxOnPage = childrenBBoxes.stream()
-                    .filter(childBboxPerPage -> childBboxPerPage.containsKey(page))
-                    .map(childBboxPerPage -> childBboxPerPage.get(page))
-                    .collect(RectangleTransformations.collectBBox());
-            bBoxPerPage.put(page, bBoxOnPage);
+        List<Map<Page, Rectangle2D>> childrenBBoxes = streamChildren().filter(child -> !isFullPageImage(child))
+                .map(SemanticNode::getBBox)
+                .toList();
+        return BBoxMergingUtility.mergeBBoxes(childrenBBoxes);
+    }
+
+
+    private static boolean isFullPageImage(SemanticNode child) {
+
+        if (!child.getType().equals(NodeType.IMAGE)) {
+            return false;
        }
-        return bBoxPerPage;
+        return ((Image) child).isFullPageImage();
    }


@@ -473,9 +494,14 @@ public interface SemanticNode {
    private Map<Page, Rectangle2D> getBBoxFromLeafTextBlock() {

        Map<Page, Rectangle2D> bBoxPerPage = new HashMap<>();
-        Map<Page, List<AtomicTextBlock>> atomicTextBlockPerPage = getTextBlock().getAtomicTextBlocks().stream().collect(Collectors.groupingBy(AtomicTextBlock::getPage));
+        Map<Page, List<AtomicTextBlock>> atomicTextBlockPerPage = getTextBlock().getAtomicTextBlocks()
+                .stream()
+                .collect(Collectors.groupingBy(AtomicTextBlock::getPage));
        atomicTextBlockPerPage.forEach((page, atbs) -> bBoxPerPage.put(page, RectangleTransformations.bBoxUnionAtomicTextBlock(atbs)));
        return bBoxPerPage;
    }

+
+    void accept(NodeVisitor visitor);
+
 }
@@ -1,6 +1,7 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;

 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;

 import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
@@ -26,8 +27,14 @@ public class SuperSection extends AbstractSemanticNode {
    public Headline getHeadline() {

        return streamChildrenOfType(NodeType.HEADLINE).map(node -> (Headline) node)
-                .findFirst()
-                .orElseGet(() -> getParent().getHeadline());
+                .findFirst().orElseGet(() -> getParent().getHeadline());
+    }
+
+
+    @Override
+    public void accept(NodeVisitor visitor) {
+
+        visitor.visit(this);
    }


@@ -15,7 +15,8 @@ import java.util.stream.Stream;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.TextEntity;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlockCollector;

@@ -43,7 +44,7 @@ public class Table implements SemanticNode {

    @Builder.Default
    @EqualsAndHashCode.Exclude
-    Set<RedactionEntity> entities = new HashSet<>();
+    Set<TextEntity> entities = new HashSet<>();

    @EqualsAndHashCode.Exclude
    Map<Page, Rectangle2D> bBoxCache;
@@ -54,7 +55,7 @@ public class Table implements SemanticNode {
     * @param strings Strings to check whether a row contains them
     * @return Stream of all entities in this table, that appear in a row, which contains any of the provided strings
     */
-    public Stream<RedactionEntity> streamEntitiesWhereRowContainsStringsIgnoreCase(List<String> strings) {
+    public Stream<TextEntity> streamEntitiesWhereRowContainsStringsIgnoreCase(List<String> strings) {

        return IntStream.range(0, numberOfRows).boxed()
                .filter(row -> rowContainsStringsIgnoreCase(row, strings))
@@ -88,7 +89,7 @@ public class Table implements SemanticNode {
     * @param value  the string which the table cell should contain
     * @return a stream of all entities, which appear in a row where at least one cell has the provided header and the provided value.
     */
-    public Stream<RedactionEntity> streamEntitiesWhereRowHasHeaderAndValue(String header, String value) {
+    public Stream<TextEntity> streamEntitiesWhereRowHasHeaderAndValue(String header, String value) {

        List<Integer> vertebrateStudyCols = streamHeaders().filter(headerNode -> headerNode.containsString(header))
                .map(TableCell::getCol)
@@ -107,7 +108,7 @@ public class Table implements SemanticNode {
     * @param values the strings which the table cell should contain
     * @return a stream of all entities, which appear in a row where at least one cell has the provided header and any provided value.
     */
-    public Stream<RedactionEntity> streamEntitiesWhereRowHasHeaderAndAnyValue(String header, List<String> values) {
+    public Stream<TextEntity> streamEntitiesWhereRowHasHeaderAndAnyValue(String header, List<String> values) {

        List<Integer> colsWithHeader = streamHeaders().filter(headerNode -> headerNode.containsString(header))
                .map(TableCell::getCol)
@@ -125,7 +126,7 @@ public class Table implements SemanticNode {
     * @param types type strings to check whether a row contains an entity like them
     * @return Stream of all entities in this table, that appear in a row, which contains at least one entity with any of the provided types.
     */
-    public Stream<RedactionEntity> streamEntitiesWhereRowContainsEntitiesOfType(List<String> types) {
+    public Stream<TextEntity> streamEntitiesWhereRowContainsEntitiesOfType(List<String> types) {

        List<Integer> rowsWithEntityOfType = IntStream.range(0, numberOfRows).boxed()
                .filter(rowNumber -> streamEntityTypesInRow(rowNumber).anyMatch(existingType -> types.stream()
@@ -145,7 +146,7 @@ public class Table implements SemanticNode {
     * @param types type strings to check whether a row doesn't contain an entity like it
     * @return Stream of all entities in this table, that appear in a row, which does not contain any entity with any of the provided types.
     */
-    public Stream<RedactionEntity> streamEntitiesWhereRowContainsNoEntitiesOfType(List<String> types) {
+    public Stream<TextEntity> streamEntitiesWhereRowContainsNoEntitiesOfType(List<String> types) {

        List<Integer> rowsWithNoEntityOfType = IntStream.range(0, numberOfRows).boxed()
                .filter(rowNumber -> streamEntityTypesInRow(rowNumber).noneMatch(existingType -> types.stream()
@@ -163,7 +164,7 @@ public class Table implements SemanticNode {

        return streamRow(rowNumber).map(TableCell::getEntities)
                .flatMap(Collection::stream)
-                .map(RedactionEntity::getType)
+                .map(TextEntity::getType)
                .distinct();
    }

@@ -304,12 +305,12 @@ public class Table implements SemanticNode {
     * Finds all entities of the provided type, which appear in the same row that the provided entity appears in.
     *
     * @param type            the type of entities to search for
-     * @param redactionEntity the entity, which appears in the row to search
+     * @param textEntity the entity, which appears in the row to search
     * @return List of all entities of the provided type, which appear in the same row that the provided entity appears in.
     */
-    public List<RedactionEntity> getEntitiesOfTypeInSameRow(String type, RedactionEntity redactionEntity) {
+    public List<TextEntity> getEntitiesOfTypeInSameRow(String type, TextEntity textEntity) {

-        return redactionEntity.getIntersectingNodes()
+        return textEntity.getIntersectingNodes()
                .stream()
                .filter(node -> node instanceof TableCell)
                .map(node -> (TableCell) node)
@@ -353,4 +354,10 @@ public class Table implements SemanticNode {
        return bBoxCache;
    }

+
+    @Override
+    public void accept(NodeVisitor visitor) {
+        visitor.visit(this);
+    }
+
 }
@@ -5,6 +5,7 @@ import java.util.HashMap;
 import java.util.Map;

 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlockCollector;

@@ -42,6 +43,13 @@ public class TableCell extends AbstractSemanticNode {
    }


+    @Override
+    public void accept(NodeVisitor visitor) {
+
+        visitor.visit(this);
+    }
+
+
    @Override
    public NodeType getType() {

@@ -10,10 +10,12 @@ import java.util.Collections;
 import java.util.HashMap;
 import java.util.List;
 import java.util.Map;
+import java.util.Set;
+import java.util.stream.Collectors;

 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentPositionData;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentTextData;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.Boundary;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.TextRange;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Page;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.SemanticNode;
 import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
@@ -36,14 +38,14 @@ public class AtomicTextBlock implements TextBlock {
    Page page;

    //string coordinates
-    Boundary boundary;
+    TextRange textRange;
    String searchText;
    @Builder.Default
    List<Integer> lineBreaks = new ArrayList<>();
    @Builder.Default
-    List<Boundary> boldTextBoundaries = new ArrayList<>();
+    List<TextRange> boldTextBoundaries = new ArrayList<>();
    @Builder.Default
-    List<Boundary> italicTextBoundaries = new ArrayList<>();
+    List<TextRange> italicTextBoundaries = new ArrayList<>();
    String orientation;
    int textDirection;

@@ -64,10 +66,44 @@ public class AtomicTextBlock implements TextBlock {
    }


+    @Override
+    public String subSequenceWithLineBreaks(TextRange stringTextRange) {
+
+        if (stringTextRange.length() == 0 || !getTextRange().contains(stringTextRange)) {
+            return "";
+        }
+
+        Set<Integer> lbInBoundary = lineBreaks.stream()
+                .map(i -> i + this.textRange.start())
+                .filter(stringTextRange::contains)
+                .collect(Collectors.toSet());
+        if (stringTextRange.end() == getTextRange().end()) {
+            lbInBoundary.add(getTextRange().end());
+        }
+        StringBuilder sb = new StringBuilder();
+        for (int i = stringTextRange.start(); i < stringTextRange.end(); i++) {
+            char character = this.charAt(i);
+            if (lbInBoundary.contains(i + 1)) {
+                // always plus one, due to the linebreaks being an exclusive end index
+                if (!Character.isWhitespace(character)) {
+                    lbInBoundary.remove(i + 1);
+                    lbInBoundary.add(i + 2);
+                    sb.append(character);
+                    continue;
+                }
+                sb.append("\n");
+            } else {
+                sb.append(character);
+            }
+        }
+        return sb.toString();
+    }
+
+
    public static AtomicTextBlock fromSearchTextWithTextPosition(String searchText,
                                                                 List<Integer> lineBreaks,
-                                                                 List<Boundary> boldTextBoundaries,
-                                                                 List<Boundary> italicTextBoundaries,
+                                                                 List<TextRange> boldTextBoundaries,
+                                                                 List<TextRange> italicTextBoundaries,
                                                                 List<Rectangle2D> positions,
                                                                 List<Integer> stringIdxToPositionIdx,
                                                                 long idx,
@@ -89,7 +125,7 @@ public class AtomicTextBlock implements TextBlock {
                .italicTextBoundaries(italicTextBoundaries)
                .positions(positions)
                .stringIdxToPositionIdx(stringIdxToPositionIdx)
-                .boundary(new Boundary(offset, offset + searchText.length()))
+                .textRange(new TextRange(offset, offset + searchText.length()))
                .textDirection(textDirection)
                .orientation(orientation)
                .build();
@@ -100,7 +136,7 @@ public class AtomicTextBlock implements TextBlock {

        return AtomicTextBlock.builder()
                .id(textBlockIdx)
-                .boundary(new Boundary(stringOffset, stringOffset))
+                .textRange(new TextRange(stringOffset, stringOffset))
                .searchText("")
                .page(page)
                .numberOnPage(numberOnPage)
@@ -109,19 +145,18 @@ public class AtomicTextBlock implements TextBlock {
    }


-    public static AtomicTextBlock fromAtomicTextBlockData(DocumentTextData documentTextData,
-                                                          DocumentPositionData documentPositionData,
-                                                          SemanticNode parent,
-                                                          Page page) {
+    public static AtomicTextBlock fromAtomicTextBlockData(DocumentTextData documentTextData, DocumentPositionData documentPositionData, SemanticNode parent, Page page) {

        return AtomicTextBlock.builder()
                .id(documentTextData.getId())
                .numberOnPage(documentTextData.getNumberOnPage())
                .page(page)
-                .boundary(new Boundary(documentTextData.getStart(), documentTextData.getEnd()))
+                .textRange(new TextRange(documentTextData.getStart(), documentTextData.getEnd()))
                .searchText(documentTextData.getSearchText())
-                .lineBreaks(Arrays.stream(documentTextData.getLineBreaks()).boxed().toList())
-                .stringIdxToPositionIdx(Arrays.stream(documentPositionData.getStringIdxToPositionIdx()).boxed().toList())
+                .lineBreaks(Arrays.stream(documentTextData.getLineBreaks()).boxed()
+                                    .toList())
+                .stringIdxToPositionIdx(Arrays.stream(documentPositionData.getStringIdxToPositionIdx()).boxed()
+                                                .toList())
                .positions(toRectangle2DList(documentPositionData.getPositions()))
                .parent(parent)
                .build();
@@ -130,7 +165,9 @@ public class AtomicTextBlock implements TextBlock {

    private static List<Rectangle2D> toRectangle2DList(float[][] positions) {

-        return Arrays.stream(positions).map(floatArr -> (Rectangle2D) new Rectangle2D.Float(floatArr[0], floatArr[1], floatArr[2], floatArr[3])).toList();
+        return Arrays.stream(positions)
+                .map(floatArr -> (Rectangle2D) new Rectangle2D.Float(floatArr[0], floatArr[1], floatArr[2], floatArr[3]))
+                .toList();
    }


@@ -140,11 +177,14 @@ public class AtomicTextBlock implements TextBlock {
            throw new IndexOutOfBoundsException(format("line %d out of range for AtomicTextBlock with %d lines", lineNumber, numberOfLines()));
        }
        if (lineNumber == 0) {
-            return subSequence(boundary.start(), lineBreaks.get(0) + boundary.start());
+            if (lineBreaks.isEmpty()) {
+                return searchText;
+            }
+            return subSequence(textRange.start(), lineBreaks.get(0) + textRange.start());
        } else if (lineNumber == numberOfLines() - 1) {
-            return subSequence(lineBreaks.get(lineBreaks.size() - 1) + boundary.start(), boundary.end());
+            return subSequence(lineBreaks.get(lineBreaks.size() - 1) + textRange.start(), textRange.end());
        }
-        return subSequence(lineBreaks.get(lineNumber - 1) + boundary.start(), lineBreaks.get(lineNumber) + boundary.start());
+        return subSequence(lineBreaks.get(lineNumber - 1) + textRange.start(), lineBreaks.get(lineNumber) + textRange.start());
    }


@@ -159,9 +199,9 @@ public class AtomicTextBlock implements TextBlock {
    public int getNextLinebreak(int fromIndex) {

        return lineBreaks.stream()//
-                .filter(linebreak -> linebreak > fromIndex - boundary.start()) //
-                .findFirst() //
-                .orElse(searchText.length()) + boundary.start();
+                       .filter(linebreak -> linebreak > fromIndex - textRange.start()) //
+                       .findFirst() //
+                       .orElse(searchText.length()) + textRange.start();
    }


@@ -169,43 +209,43 @@ public class AtomicTextBlock implements TextBlock {
    public int getPreviousLinebreak(int fromIndex) {

        return lineBreaks.stream()//
-                .filter(linebreak -> linebreak <= fromIndex - boundary.start())//
-                .reduce((a, b) -> b)//
-                .orElse(0) + boundary.start();
+                       .filter(linebreak -> linebreak <= fromIndex - textRange.start())//
+                       .reduce((a, b) -> b)//
+                       .orElse(0) + textRange.start();
    }


    @Override
    public Rectangle2D getPosition(int stringIdx) {

-        return positions.get(stringIdxToPositionIdx.get(stringIdx - boundary.start()));
+        return positions.get(stringIdxToPositionIdx.get(stringIdx - textRange.start()));
    }


    @Override
-    public List<Rectangle2D> getPositions(Boundary stringBoundary) {
+    public List<Rectangle2D> getPositions(TextRange stringTextRange) {

-        if (!containsBoundary(stringBoundary)) {
-            throw new IndexOutOfBoundsException(format("%s is out of bounds for %s", stringBoundary, this.boundary));
+        if (!containsBoundary(stringTextRange)) {
+            throw new IndexOutOfBoundsException(format("%s is out of bounds for %s", stringTextRange, this.textRange));
        }
-        if (stringBoundary.length() == 0) {
+        if (stringTextRange.length() == 0) {
            return Collections.emptyList();
        }

-        int startPositionIdx = stringIdxToPositionIdx.get(stringBoundary.start() - this.boundary.start());
+        int startPositionIdx = stringIdxToPositionIdx.get(stringTextRange.start() - this.textRange.start());

-        if (stringBoundary.end() == this.boundary.end()) {
+        if (stringTextRange.end() == this.textRange.end()) {
            return positions.subList(startPositionIdx, positions.size());
        }

-        return positions.subList(startPositionIdx, stringIdxToPositionIdx.get(stringBoundary.end() - this.boundary.start()));
+        return positions.subList(startPositionIdx, stringIdxToPositionIdx.get(stringTextRange.end() - this.textRange.start()));

    }


-    public Map<Page, List<Rectangle2D>> getPositionsPerPage(Boundary stringBoundary) {
+    public Map<Page, List<Rectangle2D>> getPositionsPerPage(TextRange stringTextRange) {

-        List<Rectangle2D> rectanglesPerLine = stringBoundary.split(getAllLineBreaksInBoundary(stringBoundary))
+        List<Rectangle2D> rectanglesPerLine = stringTextRange.split(getAllLineBreaksInBoundary(stringTextRange))
                .stream()
                .map(this::getPositions)
                .map(RectangleTransformations::rectangleBBoxWithGaps)
@@ -217,9 +257,12 @@ public class AtomicTextBlock implements TextBlock {
    }


-    private List<Integer> getAllLineBreaksInBoundary(Boundary boundary) {
+    protected List<Integer> getAllLineBreaksInBoundary(TextRange textRange) {

-        return getLineBreaks().stream().map(linebreak -> linebreak + this.boundary.start()).filter(boundary::contains).toList();
+        return getLineBreaks().stream()
+                .map(linebreak -> linebreak + this.textRange.start())
+                .filter(textRange::contains)
+                .toList();
    }


@@ -11,7 +11,7 @@ import java.util.List;
 import java.util.Map;
 import java.util.stream.Stream;

-import com.knecon.fforesight.service.layoutparser.processor.model.graph.Boundary;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.TextRange;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Page;

 import lombok.AccessLevel;
@@ -24,7 +24,7 @@ public class ConcatenatedTextBlock implements TextBlock {

    List<AtomicTextBlock> atomicTextBlocks;
    String searchText;
-    Boundary boundary;
+    TextRange textRange;


    public static ConcatenatedTextBlock empty() {
@@ -37,29 +37,30 @@ public class ConcatenatedTextBlock implements TextBlock {

        this.atomicTextBlocks = new LinkedList<>();
        if (atomicTextBlocks.isEmpty()) {
-            boundary = new Boundary(-1, -1);
+            textRange = new TextRange(-1, -1);
            return;
        }
        var firstTextBlock = atomicTextBlocks.get(0);
        this.atomicTextBlocks.add(firstTextBlock);
-        boundary = new Boundary(firstTextBlock.getBoundary().start(), firstTextBlock.getBoundary().end());
+        textRange = new TextRange(firstTextBlock.getTextRange().start(), firstTextBlock.getTextRange().end());

-        atomicTextBlocks.subList(1, atomicTextBlocks.size()).forEach(this::concat);
+        atomicTextBlocks.subList(1, atomicTextBlocks.size())
+                .forEach(this::concat);
    }


    public ConcatenatedTextBlock concat(TextBlock textBlock) {

-        int start = textBlock.getBoundary().start();
-        int end = textBlock.getBoundary().end();
+        int start = textBlock.getTextRange().start();
+        int end = textBlock.getTextRange().end();
        if (this.atomicTextBlocks.isEmpty()) {
-            boundary.setStart(start);
-            boundary.setEnd(end);
-        } else if (boundary.end() != start) {
-            throw new UnsupportedOperationException(format("Can only concat consecutive TextBlocks, trying to concat %s and %s", boundary, textBlock.getBoundary()));
+            textRange.setStart(start);
+            textRange.setEnd(end);
+        } else if (textRange.end() != start) {
+            throw new UnsupportedOperationException(format("Can only concat consecutive TextBlocks, trying to concat %s and %s", textRange, textBlock.getTextRange()));
        }
        this.atomicTextBlocks.addAll(textBlock.getAtomicTextBlocks());
-        boundary.setEnd(end);
+        textRange.setEnd(end);
        this.searchText = null;
        return this;
    }
@@ -67,13 +68,18 @@ public class ConcatenatedTextBlock implements TextBlock {

    private AtomicTextBlock getAtomicTextBlockByStringIndex(int stringIdx) {

-        return atomicTextBlocks.stream().filter(textBlock -> textBlock.getBoundary().contains(stringIdx)).findAny().orElseThrow(IndexOutOfBoundsException::new);
+        return atomicTextBlocks.stream()
+                .filter(textBlock -> textBlock.getTextRange().containsExclusive(stringIdx))
+                .findAny()
+                .orElseThrow(IndexOutOfBoundsException::new);
    }


-    private List<AtomicTextBlock> getAllAtomicTextBlocksPartiallyInStringBoundary(Boundary boundary) {
+    private List<AtomicTextBlock> getAllAtomicTextBlocksPartiallyInStringBoundary(TextRange textRange) {

-        return atomicTextBlocks.stream().filter(tb -> tb.getBoundary().intersects(boundary)).toList();
+        return atomicTextBlocks.stream()
+                .filter(tb -> tb.getTextRange().intersects(textRange))
+                .toList();
    }


@@ -92,7 +98,9 @@ public class ConcatenatedTextBlock implements TextBlock {
    @Override
    public int numberOfLines() {

-        return atomicTextBlocks.stream().map(AtomicTextBlock::getLineBreaks).mapToInt(List::size).sum();
+        return atomicTextBlocks.stream()
+                .map(AtomicTextBlock::getLineBreaks)
+                .mapToInt(List::size).sum();
    }


@@ -113,7 +121,10 @@ public class ConcatenatedTextBlock implements TextBlock {
    @Override
    public List<Integer> getLineBreaks() {

-        return getAtomicTextBlocks().stream().flatMap(atomicTextBlock -> atomicTextBlock.getLineBreaks().stream()).toList();
+        return getAtomicTextBlocks().stream()
+                .flatMap(atomicTextBlock -> atomicTextBlock.getLineBreaks()
+                        .stream())
+                .toList();
    }


@@ -125,47 +136,48 @@ public class ConcatenatedTextBlock implements TextBlock {


    @Override
-    public List<Rectangle2D> getPositions(Boundary stringBoundary) {
+    public List<Rectangle2D> getPositions(TextRange stringTextRange) {

-        List<AtomicTextBlock> textBlocks = getAllAtomicTextBlocksPartiallyInStringBoundary(stringBoundary);
+        List<AtomicTextBlock> textBlocks = getAllAtomicTextBlocksPartiallyInStringBoundary(stringTextRange);

        if (textBlocks.size() == 1) {
-            return textBlocks.get(0).getPositions(stringBoundary);
+            return textBlocks.get(0).getPositions(stringTextRange);
        }

        AtomicTextBlock firstTextBlock = textBlocks.get(0);
-        List<Rectangle2D> positions = new LinkedList<>(firstTextBlock.getPositions(new Boundary(stringBoundary.start(), firstTextBlock.getBoundary().end())));
+        List<Rectangle2D> positions = new LinkedList<>(firstTextBlock.getPositions(new TextRange(stringTextRange.start(), firstTextBlock.getTextRange().end())));

        for (AtomicTextBlock textBlock : textBlocks.subList(1, textBlocks.size() - 1)) {
            positions.addAll(textBlock.getPositions());
        }

        var lastTextBlock = textBlocks.get(textBlocks.size() - 1);
-        positions.addAll(lastTextBlock.getPositions(new Boundary(lastTextBlock.getBoundary().start(), stringBoundary.end())));
+        positions.addAll(lastTextBlock.getPositions(new TextRange(lastTextBlock.getTextRange().start(), stringTextRange.end())));

        return positions;
    }


    @Override
-    public Map<Page, List<Rectangle2D>> getPositionsPerPage(Boundary stringBoundary) {
+    public Map<Page, List<Rectangle2D>> getPositionsPerPage(TextRange stringTextRange) {

-        List<AtomicTextBlock> textBlocks = getAllAtomicTextBlocksPartiallyInStringBoundary(stringBoundary);
+        List<AtomicTextBlock> textBlocks = getAllAtomicTextBlocksPartiallyInStringBoundary(stringTextRange);

        if (textBlocks.size() == 1) {
-            return textBlocks.get(0).getPositionsPerPage(stringBoundary);
+            return textBlocks.get(0).getPositionsPerPage(stringTextRange);
        }

        AtomicTextBlock firstTextBlock = textBlocks.get(0);
-        Map<Page, List<Rectangle2D>> rectanglesPerLinePerPage = firstTextBlock.getPositionsPerPage(new Boundary(stringBoundary.start(), firstTextBlock.getBoundary().end()));
+        Map<Page, List<Rectangle2D>> rectanglesPerLinePerPage = firstTextBlock.getPositionsPerPage(new TextRange(stringTextRange.start(), firstTextBlock.getTextRange().end()));

        for (AtomicTextBlock textBlock : textBlocks.subList(1, textBlocks.size() - 1)) {
-            rectanglesPerLinePerPage = mergeEntityPositionsWithSamePageNode(rectanglesPerLinePerPage, textBlock.getPositionsPerPage(textBlock.getBoundary()));
+            rectanglesPerLinePerPage = mergeEntityPositionsWithSamePageNode(rectanglesPerLinePerPage, textBlock.getPositionsPerPage(textBlock.getTextRange()));
        }

        AtomicTextBlock lastTextBlock = textBlocks.get(textBlocks.size() - 1);
        rectanglesPerLinePerPage = mergeEntityPositionsWithSamePageNode(rectanglesPerLinePerPage,
-                lastTextBlock.getPositionsPerPage(new Boundary(lastTextBlock.getBoundary().start(), stringBoundary.end())));
+                                                                        lastTextBlock.getPositionsPerPage(new TextRange(lastTextBlock.getTextRange().start(),
+                                                                                                                        stringTextRange.end())));

        return rectanglesPerLinePerPage;
    }
@@ -174,11 +186,42 @@ public class ConcatenatedTextBlock implements TextBlock {
    private Map<Page, List<Rectangle2D>> mergeEntityPositionsWithSamePageNode(Map<Page, List<Rectangle2D>> map1, Map<Page, List<Rectangle2D>> map2) {

        Map<Page, List<Rectangle2D>> mergedMap = new HashMap<>(map1);
-        map2.forEach((pageNode, rectangles) -> mergedMap.merge(pageNode, rectangles, (l1, l2) -> Stream.concat(l1.stream(), l2.stream()).toList()));
+        map2.forEach((pageNode, rectangles) -> mergedMap.merge(pageNode,
+                                                               rectangles,
+                                                               (l1, l2) -> Stream.concat(l1.stream(), l2.stream())
+                                                                       .toList()));
        return mergedMap;
    }


+    @Override
+    public String subSequenceWithLineBreaks(TextRange stringTextRange) {
+
+        if (stringTextRange.length() == 0 || !getTextRange().contains(stringTextRange)) {
+            return "";
+        }
+
+        List<AtomicTextBlock> textBlocks = getAllAtomicTextBlocksPartiallyInStringBoundary(stringTextRange);
+
+        if (textBlocks.size() == 1) {
+            return textBlocks.get(0).subSequenceWithLineBreaks(stringTextRange);
+        }
+
+        StringBuilder sb = new StringBuilder();
+        AtomicTextBlock firstTextBlock = textBlocks.get(0);
+        sb.append(firstTextBlock.subSequenceWithLineBreaks(new TextRange(stringTextRange.start(), firstTextBlock.getTextRange().end())));
+
+        for (AtomicTextBlock textBlock : textBlocks.subList(1, textBlocks.size() - 1)) {
+            sb.append(textBlock.searchTextWithLineBreaks());
+        }
+
+        var lastTextBlock = textBlocks.get(textBlocks.size() - 1);
+        sb.append(lastTextBlock.subSequenceWithLineBreaks(new TextRange(lastTextBlock.getTextRange().start(), stringTextRange.end())));
+
+        return sb.toString();
+    }
+
+
    @Override
    public String toString() {

@@ -187,16 +230,22 @@ public class ConcatenatedTextBlock implements TextBlock {


    @Override
-    public List<Boundary> getBoldTextBoundaries() {
+    public List<TextRange> getBoldTextBoundaries() {

-        return getAtomicTextBlocks().stream().map(AtomicTextBlock::getBoldTextBoundaries).flatMap(Collection::stream).toList();
+        return getAtomicTextBlocks().stream()
+                .map(AtomicTextBlock::getBoldTextBoundaries)
+                .flatMap(Collection::stream)
+                .toList();
    }


    @Override
-    public List<Boundary> getItalicTextBoundaries() {
+    public List<TextRange> getItalicTextBoundaries() {

-        return getAtomicTextBlocks().stream().map(AtomicTextBlock::getItalicTextBoundaries).flatMap(Collection::stream).toList();
+        return getAtomicTextBlocks().stream()
+                .map(AtomicTextBlock::getItalicTextBoundaries)
+                .flatMap(Collection::stream)
+                .toList();
    }


@@ -10,7 +10,7 @@ import java.util.Map;
 import java.util.Set;
 import java.util.stream.Collectors;

-import com.knecon.fforesight.service.layoutparser.processor.model.graph.Boundary;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.TextRange;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Page;

 public interface TextBlock extends CharSequence {
@@ -21,10 +21,10 @@ public interface TextBlock extends CharSequence {
    List<AtomicTextBlock> getAtomicTextBlocks();


-    List<Boundary> getBoldTextBoundaries();
+    List<TextRange> getBoldTextBoundaries();


-    List<Boundary> getItalicTextBoundaries();
+    List<TextRange> getItalicTextBoundaries();


    String getOrientation();
@@ -33,7 +33,7 @@ public interface TextBlock extends CharSequence {
    int getTextDirection();


-    Boundary getBoundary();
+    TextRange getTextRange();


    int getNextLinebreak(int fromIndex);
@@ -48,31 +48,41 @@ public interface TextBlock extends CharSequence {
    Rectangle2D getPosition(int stringIdx);


-    List<Rectangle2D> getPositions(Boundary stringBoundary);
+    List<Rectangle2D> getPositions(TextRange stringTextRange);


-    Map<Page, List<Rectangle2D>> getPositionsPerPage(Boundary stringBoundary);
+    Map<Page, List<Rectangle2D>> getPositionsPerPage(TextRange stringTextRange);


    int numberOfLines();


+    String subSequenceWithLineBreaks(TextRange stringTextRange);
+
+
+    default String searchTextWithLineBreaks() {
+
+        return subSequenceWithLineBreaks(getTextRange());
+    }
+
    default int indexOf(String searchTerm) {

-        return indexOf(searchTerm, getBoundary().start());
+        return indexOf(searchTerm, getTextRange().start());
    }


    default Set<Page> getPages() {

-        return getAtomicTextBlocks().stream().map(AtomicTextBlock::getPage).collect(Collectors.toUnmodifiableSet());
+        return getAtomicTextBlocks().stream()
+                .map(AtomicTextBlock::getPage)
+                .collect(Collectors.toUnmodifiableSet());
    }


-    default Set<Page> getPages(Boundary boundary) {
+    default Set<Page> getPages(TextRange textRange) {

        return getAtomicTextBlocks().stream()
-                .filter(atomicTextBlock -> atomicTextBlock.getBoundary().intersects(boundary))
+                .filter(atomicTextBlock -> atomicTextBlock.getTextRange().intersects(textRange))
                .map(AtomicTextBlock::getPage)
                .collect(Collectors.toUnmodifiableSet());
    }
@@ -80,38 +90,38 @@ public interface TextBlock extends CharSequence {

    default int indexOf(String searchTerm, int startOffset) {

-        int start = getSearchText().indexOf(searchTerm, startOffset - getBoundary().start());
+        int start = getSearchText().indexOf(searchTerm, startOffset - getTextRange().start());
        if (start == -1) {
            return -1;
        }
-        return start + getBoundary().start();
+        return start + getTextRange().start();
    }


    default CharSequence getFirstLine() {

-        return subSequence(getBoundary().start(), getNextLinebreak(getBoundary().start()));
+        return subSequence(getTextRange().start(), getNextLinebreak(getTextRange().start()));
    }


-    default boolean containsBoundary(Boundary boundary) {
+    default boolean containsBoundary(TextRange textRange) {

-        if (boundary.end() < boundary.start()) {
-            throw new IllegalArgumentException(format("Invalid %s, StartIndex must be smaller than EndIndex", boundary));
+        if (textRange.end() < textRange.start()) {
+            throw new IllegalArgumentException(format("Invalid %s, StartIndex must be smaller than EndIndex", textRange));
        }
-        return getBoundary().contains(boundary);
+        return getTextRange().contains(textRange);
    }


    default boolean containsIndex(int stringIndex) {

-        return getBoundary().contains(stringIndex);
+        return getTextRange().containsExclusive(stringIndex);
    }


-    default CharSequence subSequence(Boundary boundary) {
+    default CharSequence subSequence(TextRange textRange) {

-        return subSequence(boundary.start(), boundary.end());
+        return subSequence(textRange.start(), textRange.end());
    }


@@ -128,21 +138,21 @@ public interface TextBlock extends CharSequence {
    @Override
    default CharSequence subSequence(int start, int end) {

-        return getSearchText().substring(start - getBoundary().start(), end - getBoundary().start());
+        return getSearchText().substring(start - getTextRange().start(), end - getTextRange().start());
    }


    @Override
    default int length() {

-        return getBoundary().length();
+        return getTextRange().length();
    }


    @Override
    default char charAt(int index) {

-        return getSearchText().charAt(index - getBoundary().start());
+        return getSearchText().charAt(index - getTextRange().start());
    }

 }
@@ -10,8 +10,8 @@ import lombok.NonNull;
 import lombok.RequiredArgsConstructor;

@Data
-@RequiredArgsConstructor
@AllArgsConstructor
+@RequiredArgsConstructor
 public class ClassifiedImage {

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

 }
@@ -82,10 +82,18 @@ public class OutlineExtractorService {

        String title = item.getTitle();

-        PDPage page = item.findDestinationPage(document);
-        if (page == null) {
+        PDPage page;
+        try {
+            // Can throw: "Error: can't convert to Destination COSArray" for some OCR'd PDFs
+            page = item.findDestinationPage(document);
+            if (page == null) {
+                return Optional.empty();
+            }
+        }catch (IOException e){
+            log.info(String.format("Error occurred during position resolution for outline item with title %s: " + e, title));
            return Optional.empty();
        }
+
        int pageNumber = document.getPages().indexOf(page);

        Optional<Point2D> outlinePosition = Optional.empty();
@@ -12,12 +12,14 @@ import org.springframework.stereotype.Service;

 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;

+import io.micrometer.observation.annotation.Observed;
 import lombok.extern.slf4j.Slf4j;

@Service
@Slf4j
 public class OutlineValidationService {

+    @Observed(name = "OutlineValidationService", contextualName = "create-toc")
    public TableOfContents createToC(List<TextPageBlock> headlines) {

        List<TableOfContentItem> mainSections = new ArrayList<>();
@@ -167,7 +167,7 @@ public class TOCEnrichmentService {
            }
        }

-        if (!startBlocks.isEmpty()) {
+        if (!startBlocks.isEmpty() || !startImages.isEmpty()) {
            TableOfContentItem unassigned = new TableOfContentItem(null);
            unassigned.setSectionBlocks(startBlocks);
            unassigned.setImages(startImages);
@@ -35,14 +35,14 @@ public class Cell extends BoundingBox {

    public Cell(Point2D topLeft, Point2D bottomRight) {

-        this.bBoxInitialUserSpace = new Rectangle2D.Double(topLeft.getX(), topLeft.getY(), (bottomRight.getX() - topLeft.getX()), (bottomRight.getY() - topLeft.getY()));
-        this.bBox = bBoxInitialUserSpace;
+        this.bBoxPdf = new Rectangle2D.Double(topLeft.getX(), topLeft.getY(), (bottomRight.getX() - topLeft.getX()), (bottomRight.getY() - topLeft.getY()));
+        this.bBox = bBoxPdf;
    }


    public Cell(Rectangle2D bBoxInitialUserSpace, AffineTransform initialUserSpaceToJava) {

-        this.bBoxInitialUserSpace = bBoxInitialUserSpace;
+        this.bBoxPdf = bBoxInitialUserSpace;
        this.bBox = initialUserSpaceToJava.createTransformedShape(bBoxInitialUserSpace).getBounds2D();
    }

@@ -50,7 +50,7 @@ public class Cell extends BoundingBox {
    public static Cell copy(Cell cell) {

        Cell copy = new Cell();
-        copy.bBoxInitialUserSpace = cell.bBoxInitialUserSpace;
+        copy.bBoxPdf = cell.bBoxPdf;
        copy.bBox = cell.bBox;
        return copy;
    }
@@ -70,7 +70,7 @@ public class CleanRulings {

    public boolean lineBetween(BoundingBox a, BoundingBox b) {

-        return lineBetween(a.getBBoxInitialUserSpace(), b.getBBoxInitialUserSpace());
+        return lineBetween(a.getBBoxPdf(), b.getBBoxPdf());
    }


@@ -263,8 +263,8 @@ public class TablePageBlock extends AbstractPageBlock {

                        cells.stream()
                                .map(originalCell -> new CellWithIntersection(originalCell,
-                                                                              RectangleTransformations.calculateIntersectedArea(cellFromGridStructure.getBBoxInitialUserSpace(),
-                                                                                                                                originalCell.getBBoxInitialUserSpace())))
+                                                                              RectangleTransformations.calculateIntersectedArea(cellFromGridStructure.getBBoxPdf(),
+                                                                                                                                originalCell.getBBoxPdf())))
                                .filter(cellWithIntersection -> cellWithIntersection.intersectedArea > 0)
                                .filter(cellWithIntersection -> cellWithIntersection.originalCell.getArea() > cellWithIntersection.intersectedArea * CELL_AREA_CONTAINED_THRESHOLD)
                                .max(Comparator.comparing(CellWithIntersection::intersectedArea))
@@ -0,0 +1,9 @@
+package com.knecon.fforesight.service.layoutparser.processor.model.text;
+
+
+public enum FontStyle {
+    REGULAR,
+    BOLD,
+    ITALIC,
+    BOLD_ITALIC;
+}
@@ -5,64 +5,50 @@ import java.awt.geom.Rectangle2D;

 import org.apache.pdfbox.text.TextPosition;

-import com.fasterxml.jackson.annotation.JsonIgnore;
-import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.BoundingBox;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.TextBoundingBox;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.utils.FastAtan2;

+import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
-import lombok.Builder;
 import lombok.Data;
+import lombok.EqualsAndHashCode;
 import lombok.NoArgsConstructor;
 import lombok.SneakyThrows;
+import lombok.experimental.FieldDefaults;
+import lombok.experimental.SuperBuilder;

@Data
-@Builder
+@SuperBuilder
@NoArgsConstructor
@AllArgsConstructor
-public class RedTextPosition extends BoundingBox {
+@EqualsAndHashCode(callSuper = true)
+@FieldDefaults(level = AccessLevel.PRIVATE)
+public class RedTextPosition extends TextBoundingBox {

    public final static int HEIGHT_PADDING = 2;

-    private Rectangle2D.Float bBoxDirAdj; // adjusted to text rotation
+    String unicode;

-    @JsonIgnore
-    private int rotation;
+    // estimated using the TextMatrix in radians
+    float exactDir;

-    @JsonIgnore
-    private float pageHeight;
+    float widthOfSpace;

-    @JsonIgnore
-    private float pageWidth;
+    float fontSizeInPt;

-    private String unicode;
-
-    @JsonIgnore
-    private float dir;
-
-    // not used in reanalysis
-    @JsonIgnore
-    private float widthOfSpace;
-
-    // not used in reanalysis
-    @JsonIgnore
-    private float fontSizeInPt;
-
-    // not used in reanalysis
-    @JsonIgnore
-    private String fontName;
+    String fontName;


    @SneakyThrows
    public static RedTextPosition fromTextPosition(TextPosition textPosition) {

        var pos = new RedTextPosition();
-        pos.setRotation(textPosition.getRotation());
-        pos.setPageHeight(textPosition.getPageHeight());
-        pos.setPageWidth(textPosition.getPageWidth());
        pos.setUnicode(textPosition.getUnicode());
-        pos.setDir(textPosition.getDir());
        pos.setWidthOfSpace(textPosition.getWidthOfSpace());
        pos.setFontSizeInPt(textPosition.getFontSizeInPt());
        pos.setFontName(textPosition.getFont().getName());
+        pos.setExactDir((float) FastAtan2.fastAtan2(textPosition.getTextMatrix().getShearY(), textPosition.getTextMatrix().getScaleX()));
+        pos.setDir(TextDirection.fromDegrees(textPosition.getDir()));

        //TODO: There is a mismatch in the java coords of the text and the rulings,
        // I guess if we start with the initial user space positions and transform them the same way we do the rulings it would work.
@@ -73,18 +59,18 @@ public class RedTextPosition extends BoundingBox {
                                                                 textPosition.getYDirAdj() - textHeight,
                                                                 textPosition.getWidthDirAdj(),
                                                                 textHeight + HEIGHT_PADDING);
+
        pos.setBBoxDirAdj(dirAdjPosition);

        AffineTransform affineTransform = getRotationMatrix(TextDirection.fromDegrees(textPosition.getDir()), textPosition.getPageWidth(), textPosition.getPageHeight());
        Rectangle2D bBoxInitialUserSpace = affineTransform.createTransformedShape(dirAdjPosition).getBounds2D();

-        pos.setBBoxInitialUserSpace(bBoxInitialUserSpace); // These are definitely correct
+        pos.setBBoxPdf(bBoxInitialUserSpace); // These are definitely correct

        return pos;
    }


-
    private static AffineTransform getRotationMatrix(TextDirection textDirection, float pageWidth, float pageHeight) {

        AffineTransform transform = new AffineTransform();
@@ -103,32 +89,4 @@ public class RedTextPosition extends BoundingBox {
        return transform;
    }

-
-    @JsonIgnore
-    public float getXDirAdj() {
-
-        return this.bBoxDirAdj.x;
-    }
-
-
-    @JsonIgnore
-    public float getYDirAdj() {
-
-        return this.bBoxDirAdj.y;
-    }
-
-
-    @JsonIgnore
-    public float getWidthDirAdj() {
-
-        return this.bBoxDirAdj.width;
-    }
-
-
-    @JsonIgnore
-    public float getHeightDir() {
-
-        return this.bBoxDirAdj.height;
-    }
-
 }
@@ -44,4 +44,15 @@ public enum TextDirection {

        throw new IllegalArgumentException(String.format("A value of %f is not supported by TextDirection", degrees));
    }
+
+
+    public int getRotation() {
+
+        return switch (this) {
+            case ZERO -> 0;
+            case QUARTER_CIRCLE -> 1;
+            case HALF_CIRCLE -> 2;
+            case THREE_QUARTER_CIRCLE -> 3;
+        };
+    }
 }
@@ -8,6 +8,7 @@ import com.fasterxml.jackson.annotation.JsonIgnore;
 import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.FloatFrequencyCounter;
 import com.knecon.fforesight.service.layoutparser.processor.model.PageBlockType;
+import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
 import com.knecon.fforesight.service.layoutparser.processor.utils.TextNormalizationUtilities;

 import lombok.AllArgsConstructor;
@@ -26,17 +27,19 @@ public class TextPageBlock extends AbstractPageBlock {
    @Builder.Default
    private List<TextPositionSequence> sequences = new ArrayList<>();

+    private Rectangle2D bBoxDirAdj;
+
    private String mostPopularWordFont;

    private String mostPopularWordStyle;

-    private float mostPopularWordFontSize;
+    private double mostPopularWordFontSize;

-    private float mostPopularWordHeight;
+    private double mostPopularWordHeight;

-    private float mostPopularWordSpaceWidth;
+    private double mostPopularWordSpaceWidth;

-    private float highestFontSize;
+    private double highestFontSize;

    private PageBlockType classification;

@@ -46,39 +49,31 @@ public class TextPageBlock extends AbstractPageBlock {
    public TextPageBlock(List<TextPositionSequence> sequences) {

        this.sequences = sequences;
-        calculateFrequencyCounters();
+        if (!sequences.isEmpty()) {
+            calculateFrequencyCounters();
+        }
        calculateBBox();
    }


-    @JsonIgnore
    public TextDirection getDir() {

        return sequences.get(0).getDir();
    }


-    @JsonIgnore
-    public float getPageHeight() {
-
-        return sequences.get(0).getPageHeight();
-    }
-
-
-    @JsonIgnore
-    public float getPageWidth() {
-
-        return sequences.get(0).getPageWidth();
-    }
-
-
    private void calculateBBox() {

        if (sequences == null) {
            this.bBox = new Rectangle2D.Double();
-            this.bBoxInitialUserSpace = new Rectangle2D.Double();
+            this.bBoxPdf = new Rectangle2D.Double();
+            this.bBoxDirAdj = new Rectangle2D.Double();
            return;
        }
+        this.bBoxDirAdj = sequences.stream()
+                .map(TextPositionSequence::getBBoxDirAdj)
+                .collect(RectangleTransformations.collectBBox());
+
        setToBBoxOfComponents(sequences);
    }

@@ -1,6 +1,7 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.text;

-import java.awt.geom.Rectangle2D;
+import static com.knecon.fforesight.service.layoutparser.processor.model.text.RedTextPosition.HEIGHT_PADDING;
+
 import java.util.ArrayList;
 import java.util.List;
 import java.util.Locale;
@@ -8,8 +9,7 @@ import java.util.stream.Collectors;

 import org.apache.pdfbox.text.TextPosition;

-import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.BoundingBox;
-import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.TextBoundingBox;

 import lombok.AllArgsConstructor;
 import lombok.Builder;
@@ -23,23 +23,21 @@ import lombok.extern.slf4j.Slf4j;
@Builder
@NoArgsConstructor
@AllArgsConstructor
-@EqualsAndHashCode(onlyExplicitlyIncluded = true, callSuper = false)
-public class TextPositionSequence extends BoundingBox implements CharSequence {
+@EqualsAndHashCode(onlyExplicitlyIncluded = true, callSuper = true) // needs the bbox to be unique
+public class TextPositionSequence extends TextBoundingBox implements CharSequence {

-    public static final int HEIGHT_PADDING = 2;
+    public static final String STANDARD = "standard";
+    public static final String BOLD_ITALIC = "bold, italic";
+    public static final String BOLD = "bold";
+    public static final String ITALIC = "italic";

    @EqualsAndHashCode.Include
    private int page;

    @EqualsAndHashCode.Include
+    @Builder.Default
    private List<RedTextPosition> textPositions = new ArrayList<>();

-    private Rectangle2D bBoxDirAdj;
-    @EqualsAndHashCode.Include
-    private TextDirection dir;
-    private int rotation;
-    private float pageHeight;
-    private float pageWidth;
    private boolean isParagraphStart;
    private boolean strikethrough;
    private boolean underline;
@@ -51,10 +49,6 @@ public class TextPositionSequence extends BoundingBox implements CharSequence {
                .map(RedTextPosition::fromTextPosition)
                .collect(Collectors.toList());
        this.page = pageNumber;
-        this.dir = TextDirection.fromDegrees(textPositions.get(0).getDir());
-        this.rotation = textPositions.get(0).getRotation();
-        this.pageHeight = textPositions.get(0).getPageHeight();
-        this.pageWidth = textPositions.get(0).getPageWidth();
        this.isParagraphStart = isParagraphStart;
        calculateBBox();
    }
@@ -62,9 +56,6 @@ public class TextPositionSequence extends BoundingBox implements CharSequence {

    private void calculateBBox() {

-        this.bBoxDirAdj = textPositions.stream()
-                .map(RedTextPosition::getBBoxDirAdj)
-                .collect(RectangleTransformations.collectBBox());
        setToBBoxOfComponents(getTextPositions());
    }

@@ -73,10 +64,6 @@ public class TextPositionSequence extends BoundingBox implements CharSequence {

        this.textPositions = textPositions;
        this.page = page;
-        this.dir = TextDirection.fromDegrees(textPositions.get(0).getDir());
-        this.rotation = textPositions.get(0).getRotation();
-        this.pageHeight = textPositions.get(0).getPageHeight();
-        this.pageWidth = textPositions.get(0).getPageWidth();
        calculateBBox();
    }

@@ -112,9 +99,6 @@ public class TextPositionSequence extends BoundingBox implements CharSequence {
        textPositionSequence.textPositions = textPositions.subList(start, end);
        textPositionSequence.page = page;
        textPositionSequence.dir = dir;
-        textPositionSequence.rotation = rotation;
-        textPositionSequence.pageHeight = pageHeight;
-        textPositionSequence.pageWidth = pageWidth;
        textPositionSequence.setToBBoxOfComponents(getTextPositions());
        return textPositionSequence;
    }
@@ -141,10 +125,6 @@ public class TextPositionSequence extends BoundingBox implements CharSequence {

        this.textPositions.add(textPosition);
        this.page = textPositionSequence.getPage();
-        this.dir = textPositionSequence.getDir();
-        this.rotation = textPositionSequence.getRotation();
-        this.pageHeight = textPositionSequence.getPageHeight();
-        this.pageWidth = textPositionSequence.getPageWidth();
        calculateBBox();
    }

@@ -152,79 +132,18 @@ public class TextPositionSequence extends BoundingBox implements CharSequence {
    public void add(TextPosition textPosition) {

        this.textPositions.add(RedTextPosition.fromTextPosition(textPosition));
-        this.dir = TextDirection.fromDegrees(textPositions.get(0).getDir());
-        this.rotation = textPositions.get(0).getRotation();
-        this.pageHeight = textPositions.get(0).getPageHeight();
-        this.pageWidth = textPositions.get(0).getPageWidth();
        calculateBBox();
    }

+    public double getTextHeightNoPadding() {

-    /**
-     * This value is adjusted so that 0,0 is upper left and it is adjusted based on the text direction.
-     * This method ignores the page rotation but takes the text rotation and adjusts the coordinates to awt.
-     *
-     * @return the text direction adjusted minX value
-     */
-
-    public float getMinXDirAdj() {
-
-        return textPositions.get(0).getXDirAdj();
-
+        return textPositions.get(0).getHeightDirAdj();
    }


-    /**
-     * This value is adjusted so that 0,0 is upper left and it is adjusted based on the text direction.
-     * This method ignores the page rotation but takes the text rotation and adjusts the coordinates to awt.
-     *
-     * @return the text direction adjusted maxX value
-     */
+    public double getTextHeight() {

-    public float getMaxXDirAdj() {
-
-        return textPositions.get(textPositions.size() - 1).getXDirAdj() + textPositions.get(textPositions.size() - 1).getWidthDirAdj() + HEIGHT_PADDING;
-
-    }
-
-
-    /**
-     * This value is adjusted so that 0,0 is upper left and it is adjusted based on the text direction.
-     * This method ignores the page rotation but takes the text rotation and adjusts the coordinates to awt.
-     *
-     * @return the text direction adjusted minY value. The upper border of the bounding box of the word.
-     */
-
-    public float getMinYDirAdj() {
-
-        return textPositions.get(0).getYDirAdj() - getTextHeight();
-
-    }
-
-
-    /**
-     * This value is adjusted so that 0,0 is upper left and it is adjusted based on the text direction.
-     * This method ignores the page rotation but takes the text rotation and adjusts the coordinates to awt.
-     *
-     * @return the text direction adjusted maxY value. The lower border of the bounding box of the word.
-     */
-
-    public float getMaxYDirAdj() {
-
-        return textPositions.get(0).getYDirAdj();
-
-    }
-
-
-    public float getTextHeightNoPadding() {
-
-        return textPositions.get(0).getHeightDir();
-    }
-
-
-    public float getTextHeight() {
-
-        return textPositions.get(0).getHeightDir() + HEIGHT_PADDING;
+        return textPositions.get(0).getHeightDirAdj() + HEIGHT_PADDING;
    }


@@ -240,18 +159,18 @@ public class TextPositionSequence extends BoundingBox implements CharSequence {
    public String getFontStyle() {

        if (textPositions.get(0).getFontName() == null) {
-            return "standard";
+            return STANDARD;
        }
        String lowercaseFontName = textPositions.get(0).getFontName().toLowerCase(Locale.ROOT);

-        if (lowercaseFontName.contains("bold") && lowercaseFontName.contains("italic")) {
-            return "bold, italic";
-        } else if (lowercaseFontName.contains("bold")) {
-            return "bold";
-        } else if (lowercaseFontName.contains("italic")) {
-            return "italic";
+        if (lowercaseFontName.contains(BOLD) && lowercaseFontName.contains(ITALIC)) {
+            return BOLD_ITALIC;
+        } else if (lowercaseFontName.contains(BOLD)) {
+            return BOLD;
+        } else if (lowercaseFontName.contains(ITALIC)) {
+            return ITALIC;
        } else {
-            return "standard";
+            return STANDARD;
        }
    }

@@ -32,7 +32,7 @@ public class ImageServiceResponseAdapter {
                    .add(new ClassifiedImage(new Rectangle2D.Double(imageMetadata.getPosition().getX1(),
                            imageMetadata.getPosition().getY1(),
                            imageMetadata.getGeometry().getWidth(),
-                            imageMetadata.getGeometry().getHeight()), classification, imageMetadata.isAlpha(), imageMetadata.getPosition().getPageNumber()));
+                            imageMetadata.getGeometry().getHeight()), classification, imageMetadata.isAlpha(), imageMetadata.getPosition().getPageNumber(),imageMetadata.getRepresentation()));
        });

        // Currently This is a copy but, it will be changed later because i don' t think that we should unclassified images.
@@ -44,7 +44,7 @@ public class ImageServiceResponseAdapter {
                    .add(new ClassifiedImage(new Rectangle2D.Double(imageMetadata.getPosition().getX1(),
                            imageMetadata.getPosition().getY1(),
                            imageMetadata.getGeometry().getWidth(),
-                            imageMetadata.getGeometry().getHeight()), classification, imageMetadata.isAlpha(), imageMetadata.getPosition().getPageNumber()));
+                            imageMetadata.getGeometry().getHeight()), classification, imageMetadata.isAlpha(), imageMetadata.getPosition().getPageNumber(),imageMetadata.getRepresentation()));
        });

        return images;
@@ -56,7 +56,7 @@ public class ImageServiceResponseAdapter {
        classificationPage.getImages().forEach(image -> {
            if (image.getImageType().equals(ImageType.OTHER)) {
                for (AbstractPageBlock textblock : classificationPage.getTextBlocks()) {
-                    if (image.getPosition().contains(textblock.getBBoxInitialUserSpace())) {
+                    if (image.getPosition().contains(textblock.getBBoxPdf())) {
                        image.setImageType(ImageType.OCR);
                        return;
                    }
@@ -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);
            }
@@ -14,6 +14,7 @@ import lombok.NoArgsConstructor;
@AllArgsConstructor
 public class Classification {

+    @Builder.Default
    private Map<String, Float> probabilities = new HashMap<>();
    private String label;

@@ -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;
@@ -22,8 +22,10 @@ public class ImageServiceResponse {

    @JsonProperty(value = "imageMetadata")
    @JsonAlias("data")
+    @Builder.Default
    private List<ImageMetadata> data = new ArrayList<>();

+    @Builder.Default
    private List<ImageMetadata> dataCV = new ArrayList<>();


@@ -15,6 +15,7 @@ import lombok.NoArgsConstructor;
 public class TableData {

    private PageInfo pageInfo;
+    @Builder.Default
    private List<TableCells> tableCells = new ArrayList<>();

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

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

@@ -6,7 +6,6 @@ import java.util.List;

 import com.knecon.fforesight.service.layoutparser.processor.model.GapInformation;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
-import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;

 import lombok.AllArgsConstructor;
 import lombok.experimental.UtilityClass;
@@ -38,7 +37,7 @@ public class GapDetectionService {
        for (TextPositionSequence currentTextPosition : sortedTextPositionSequences.subList(1, sortedTextPositionSequences.size())) {

            double yDifference = Math.abs(currentTextPosition.getMaxYDirAdj() - previousTextPosition.getMaxYDirAdj());
-            double xGap = Math.abs(previousTextPosition.getMaxXDirAdj() - currentTextPosition.getMinXDirAdj());
+            double xGap = Math.abs(previousTextPosition.getMaxXDirAdj() - currentTextPosition.getXDirAdj());
            Rectangle2D previousTextPositionBBox = toRectangle2D(previousTextPosition);
            Rectangle2D currentTextPositionBBox = toRectangle2D(currentTextPosition);

@@ -71,7 +71,7 @@ public class LineDetectionService {

    private static boolean isXGap(TextPositionSequence currentTextPosition, TextPositionSequence previousTextPosition, double avgTextPositionHeight) {

-        return Math.abs(previousTextPosition.getMaxXDirAdj() - currentTextPosition.getMinXDirAdj()) > (avgTextPositionHeight * X_GAP_FACTOR);
+        return Math.abs(previousTextPosition.getMaxXDirAdj() - currentTextPosition.getXDirAdj()) > (avgTextPositionHeight * X_GAP_FACTOR);
    }


@@ -83,7 +83,7 @@ public class LineDetectionService {

    private static boolean isNewLine(TextPositionSequence currentTextPosition, TextPositionSequence previousTextPosition, double avgTextPositionHeight) {

-        return Math.abs(previousTextPosition.getMinYDirAdj() - currentTextPosition.getMinYDirAdj()) > avgTextPositionHeight;
+        return Math.abs(previousTextPosition.getYDirAdj() - currentTextPosition.getYDirAdj()) > avgTextPositionHeight;
    }


@@ -78,7 +78,7 @@ public class TableExtractionService {

            List<Cell> containedCells = new ArrayList<>();
            for (Cell c : cells) {
-                if (c.hasMinimumSize() && area.contains(c.getBBoxInitialUserSpace())) {
+                if (c.hasMinimumSize() && area.contains(c.getBBoxPdf())) {
                    containedCells.add(c);
                }
            }
@@ -31,13 +31,13 @@ public class TextRulingsClassifier {

    private static void handleVerticalText(CleanRulings cleanRulings, TextPositionSequence word) {

-        float lowerY = (float) (word.getBBoxInitialUserSpace().getMinY() + TEXT_BBOX_THRESHOLD_FACTOR * word.getWidth());
-        float upperY = (float) (word.getBBoxInitialUserSpace().getMaxY() - TEXT_BBOX_THRESHOLD_FACTOR * word.getWidth());
+        float lowerY = (float) (word.getBBoxPdf().getMinY() + TEXT_BBOX_THRESHOLD_FACTOR * word.getWidth());
+        float upperY = (float) (word.getBBoxPdf().getMaxY() - TEXT_BBOX_THRESHOLD_FACTOR * word.getWidth());

-        float strikethroughCenterX = (float) word.getBBoxInitialUserSpace().getCenterX();
+        float strikethroughCenterX = (float) word.getBBoxPdf().getCenterX();
        float strikethroughBoxHeight = (float) ((word.getBBoxDirAdj().getHeight() * STRIKETHROUGH_ZONE) / 2);

-        float underlineCenterX = (float) (word.getDir().equals(TextDirection.QUARTER_CIRCLE) ? word.getBBoxInitialUserSpace().getMaxX() : word.getBBoxInitialUserSpace().getMinX());
+        float underlineCenterX = (float) (word.getDir().equals(TextDirection.QUARTER_CIRCLE) ? word.getBBoxPdf().getMaxX() : word.getBBoxPdf().getMinX());
        float underlineBoxHeight = (float) ((word.getBBoxDirAdj().getHeight() * UNDERLINE_ZONE) / 2);

        float leftX = Math.min(underlineCenterX - underlineBoxHeight, strikethroughCenterX - strikethroughBoxHeight);
@@ -65,13 +65,13 @@ public class TextRulingsClassifier {

    private static void handleHorizontalText(CleanRulings cleanRulings, TextPositionSequence word) {

-        float leftX = (float) (word.getBBoxInitialUserSpace().getMinX() + TEXT_BBOX_THRESHOLD_FACTOR * word.getWidth());
-        float rightX = (float) (word.getBBoxInitialUserSpace().getMaxX() - TEXT_BBOX_THRESHOLD_FACTOR * word.getWidth());
+        float leftX = (float) (word.getBBoxPdf().getMinX() + TEXT_BBOX_THRESHOLD_FACTOR * word.getWidth());
+        float rightX = (float) (word.getBBoxPdf().getMaxX() - TEXT_BBOX_THRESHOLD_FACTOR * word.getWidth());

-        float strikethroughCenterY = (float) word.getBBoxInitialUserSpace().getCenterY();
+        float strikethroughCenterY = (float) word.getBBoxPdf().getCenterY();
        float strikethroughBoxHeight = (float) ((word.getBBoxDirAdj().getHeight() * STRIKETHROUGH_ZONE) / 2);

-        float underlineCenterY = (float) (word.getDir().equals(TextDirection.ZERO) ? word.getBBoxInitialUserSpace().getMinY() : word.getBBoxInitialUserSpace().getMaxY());
+        float underlineCenterY = (float) (word.getDir().equals(TextDirection.ZERO) ? word.getBBoxPdf().getMinY() : word.getBBoxPdf().getMaxY());
        float underlineBoxHeight = (float) ((word.getBBoxDirAdj().getHeight() * UNDERLINE_ZONE) / 2);

        float lowerY = Math.min(underlineCenterY - underlineBoxHeight, strikethroughCenterY - strikethroughBoxHeight);
@@ -2,12 +2,10 @@ package com.knecon.fforesight.service.layoutparser.processor.services.blockifica

 import static com.knecon.fforesight.service.layoutparser.processor.services.blockification.DocstrumBlockificationService.buildTextBlock;

-import java.awt.geom.Rectangle2D;
 import java.util.ArrayList;
 import java.util.List;
 import java.util.ListIterator;
 import java.util.Locale;
-import java.util.function.Function;

 import org.apache.commons.lang3.StringUtils;
 import org.springframework.stereotype.Service;
@@ -19,8 +17,6 @@ import com.knecon.fforesight.service.layoutparser.processor.model.SectionIdentif
 import com.knecon.fforesight.service.layoutparser.processor.model.outline.OutlineObject;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
-import com.knecon.fforesight.service.layoutparser.processor.services.factory.SearchTextWithTextPositionFactory;
-import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;

 import lombok.Data;

@@ -29,14 +25,6 @@ public class BlockificationPostprocessingService {

    private static final float BLOCK_TO_OUTLINE_DISTANCE_THRESHOLD = 5.0f;

-    private static final Function<TextPageBlock, Rectangle2D> blockToBoundingBox = (abstractPageBlock) -> abstractPageBlock.getSequences()
-            .stream()
-            .map(textPositionSequence -> textPositionSequence.getTextPositions()
-                    .stream()
-                    .map(tp -> SearchTextWithTextPositionFactory.mapRedTextPositionToInitialUserSpace(tp, textPositionSequence))
-                    .collect(RectangleTransformations.collectBBox()))
-            .collect(RectangleTransformations.collectBBox());
-

    public OutlineObject sanitizeOutlineBlocks(ClassificationPage classificationPage, OutlineObject notFoundOutlineObject) {

@@ -63,13 +51,13 @@ public class BlockificationPostprocessingService {
            }

            if (!contextsOverlap(notFoundOutlineObjectProcessionContext, firstOutlineObjectProcessionContext)) {
-                notFoundOutlineObject.setFound(selectMatch(classificationPage, notFoundOutlineObjectProcessionContext));
+                notFoundOutlineObject.setFound(selectMatch(classificationPage, notFoundOutlineObjectProcessionContext, pageHeight));
            }
            if (firstOutlineObject != null) {
                // re-create the context for the updated blocks
                firstOutlineObjectProcessionContext = new OutlineProcessionContext(firstOutlineObject);
                processTextBlocks(getTextPageBlocks(classificationPage), pageHeight, firstOutlineObjectProcessionContext);
-                firstOutlineObject.setFound(selectMatch(classificationPage, firstOutlineObjectProcessionContext));
+                firstOutlineObject.setFound(selectMatch(classificationPage, firstOutlineObjectProcessionContext, pageHeight));
            }

        }
@@ -77,7 +65,7 @@ public class BlockificationPostprocessingService {
        outlineObjectListIterator.forEachRemaining(outlineObject -> {
            OutlineProcessionContext outlineObjectProcessionContext = new OutlineProcessionContext(outlineObject);
            processTextBlocks(getTextPageBlocks(classificationPage), pageHeight, outlineObjectProcessionContext);
-            outlineObject.setFound(selectMatch(classificationPage, outlineObjectProcessionContext));
+            outlineObject.setFound(selectMatch(classificationPage, outlineObjectProcessionContext, pageHeight));
        });

        if (!outlineObjects.isEmpty()) {
@@ -160,7 +148,7 @@ public class BlockificationPostprocessingService {
    }


-    private boolean selectMatch(ClassificationPage classificationPage, OutlineProcessionContext context) {
+    private boolean selectMatch(ClassificationPage classificationPage, OutlineProcessionContext context, float pageHeight) {

        OutlineObject outlineObject = context.outlineObject;
        TextPageBlock directMatch = context.directMatch;
@@ -168,8 +156,8 @@ public class BlockificationPostprocessingService {
        TextPageBlock splitCandidate = context.splitCandidate;
        PageBlockType headlineType = PageBlockType.getHeadlineType(outlineObject.getTreeDepth());

-        double distanceToDirectMatch = directMatch != null ? calculateDistance(outlineObject, directMatch) : Double.MAX_VALUE;
-        double distanceToSplitCandidate = splitCandidate != null ? calculateDistance(outlineObject, splitCandidate) : Double.MAX_VALUE;
+        double distanceToDirectMatch = directMatch != null ? calculateDistance(outlineObject, directMatch, pageHeight) : Double.MAX_VALUE;
+        double distanceToSplitCandidate = splitCandidate != null ? calculateDistance(outlineObject, splitCandidate, pageHeight) : Double.MAX_VALUE;

        double distanceToBestMergeCandidates = Double.MAX_VALUE;
        List<TextPageBlock> bestMergeCandidateCombination = new ArrayList<>();
@@ -189,7 +177,7 @@ public class BlockificationPostprocessingService {

            for (List<TextPageBlock> combination : combinations) {
                double averageDistance = combination.stream()
-                        .map(block -> calculateDistance(outlineObject, block))
+                        .map(block -> calculateDistance(outlineObject, block, pageHeight))
                        .mapToDouble(Double::doubleValue).average()
                        .orElse(Double.MAX_VALUE);
                if (distanceToBestMergeCandidates > averageDistance) {
@@ -273,6 +261,10 @@ public class BlockificationPostprocessingService {
        List<TextPositionSequence> postSequence = new ArrayList<>();
        StringBuilder currentSequence = new StringBuilder();

+        if (target.isBlank()) {
+            return new WordSequenceResult();
+        }
+
        for (TextPositionSequence sequence : textPositionSequences) {

            currentSequence.append(sanitizeString(sequence.toString()));
@@ -332,7 +324,7 @@ public class BlockificationPostprocessingService {
        TextPositionSequence in = null;
        TextPositionSequence out;

-        String currentSequence = sequence.toString();
+        String currentSequence = sequence.toString().toLowerCase(Locale.ROOT);
        int index = currentSequence.indexOf(toRemove);
        int endIndex = index + toRemove.length();

@@ -414,10 +406,10 @@ public class BlockificationPostprocessingService {
    }


-    private double calculateDistance(OutlineObject outlineObject, TextPageBlock pageBlock) {
+    private double calculateDistance(OutlineObject outlineObject, TextPageBlock pageBlock, float pageHeight) {

        double deltaX = outlineObject.getPoint().getX() - pageBlock.getMinX();
-        double deltaY = pageBlock.getPageHeight() - outlineObject.getPoint().getY() - pageBlock.getMinY();
+        double deltaY = pageHeight - outlineObject.getPoint().getY() - pageBlock.getMinY();
        return Math.sqrt(deltaX * deltaX + deltaY * deltaY);
    }

@@ -1,7 +1,6 @@
 package com.knecon.fforesight.service.layoutparser.processor.services.blockification;

 import java.util.ArrayList;
-import java.util.Comparator;
 import java.util.List;
 import java.util.ListIterator;

@@ -10,14 +9,13 @@ import org.springframework.stereotype.Service;
 import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingType;
 import com.knecon.fforesight.service.layoutparser.processor.docstrum.DocstrumSegmentationService;
 import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Zone;
-import com.knecon.fforesight.service.layoutparser.processor.docstrum.utils.DoubleUtils;
 import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRulings;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.TablePageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
-import com.knecon.fforesight.service.layoutparser.processor.visualization.LayoutparsingVisualizations;
+import com.knecon.fforesight.service.layoutparser.processor.visualization.LayoutDebugLayer;

 import lombok.RequiredArgsConstructor;

@@ -35,12 +33,12 @@ public class DocstrumBlockificationService {
    public ClassificationPage blockify(List<TextPositionSequence> textPositions,
                                       CleanRulings rulings,
                                       boolean xyOrder,
-                                       LayoutparsingVisualizations visualizations,
+                                       LayoutDebugLayer visualizations,
                                       LayoutParsingType layoutParsingType) {

        CleanRulings usedRulings = rulings.withoutTextRulings();

-        var zones = docstrumSegmentationService.segmentPage(textPositions, xyOrder, usedRulings, visualizations);
+        List<Zone> zones = docstrumSegmentationService.segmentPage(textPositions, xyOrder, usedRulings, visualizations);

        if (!textPositions.isEmpty()) {
            visualizations.addZoneVisualizations(zones, textPositions.get(0).getPage());
@@ -48,11 +46,7 @@ public class DocstrumBlockificationService {
            visualizations.addCharactersWithNeighbours(zones, textPositions.get(0).getPage());
        }

-        var pageBlocks = toAbstractPageBlocks(zones, xyOrder, usedRulings);
-
-        if (xyOrder) {
-            sortPageBlocksXThenY(pageBlocks);
-        }
+        var pageBlocks = toAbstractPageBlocks(zones);

        var classificationPage = new ClassificationPage(pageBlocks);
        classificationPage.setCleanRulings(rulings);
@@ -73,21 +67,7 @@ public class DocstrumBlockificationService {
    }


-    private static void sortPageBlocksXThenY(List<AbstractPageBlock> pageBlocks) {
-
-        pageBlocks.sort(Comparator.comparing(AbstractPageBlock::getMinY, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD))
-                                .thenComparing(AbstractPageBlock::getMinX, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD)));
-        pageBlocks.sort(new Comparator<AbstractPageBlock>() {
-            @Override
-            public int compare(AbstractPageBlock o1, AbstractPageBlock o2) {
-
-                return Math.abs(o1.getMinY() - o2.getMinY()) < 5 && o1.getMinX() < o2.getMinX() == true ? -1 : 0;
-            }
-        });
-    }
-
-
-    private List<AbstractPageBlock> toAbstractPageBlocks(List<Zone> zones, boolean xyOrder, CleanRulings usedRulings) {
+    private List<AbstractPageBlock> toAbstractPageBlocks(List<Zone> zones) {

        List<AbstractPageBlock> abstractPageBlocks = new ArrayList<>();
        zones.forEach(zone -> {
@@ -296,6 +276,10 @@ public class DocstrumBlockificationService {
                    continue;
                }

+//                if (!current.getMostPopularWordStyle().equals(inner.getMostPopularWordStyle())) {
+//                    continue;
+//                }
+
                if (current.getDir() == inner.getDir() && current.intersects(inner, yThreshold, xThreshold)) {

                    boolean toDuplicate = current.isToDuplicate() || inner.isToDuplicate();
@@ -1,9 +1,6 @@
 package com.knecon.fforesight.service.layoutparser.processor.services.blockification;

-import static java.util.stream.Collectors.toSet;
-
 import java.util.ArrayList;
-import java.util.Comparator;
 import java.util.List;
 import java.util.regex.Matcher;
 import java.util.regex.Pattern;
@@ -13,10 +10,8 @@ import org.springframework.stereotype.Service;

 import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
-import com.knecon.fforesight.service.layoutparser.processor.model.FloatFrequencyCounter;
 import com.knecon.fforesight.service.layoutparser.processor.model.Orientation;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRulings;
-import com.knecon.fforesight.service.layoutparser.processor.model.text.StringFrequencyCounter;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;

@@ -44,31 +39,30 @@ public class DocuMineBlockificationService {

        CleanRulings usedRulings = cleanRulings.withoutTextRulings();

-        float minX = 1000;
-        float maxX = 0;
-        float minY = 1000;
-        float maxY = 0;
+        double minX = 1000;
+        double maxX = 0;
+        double minY = 1000;
+        double maxY = 0;
        TextPositionSequence prev = null;

        boolean wasSplitted = false;
-        Float splitX1 = null;
+        Double splitX1 = null;
        for (TextPositionSequence word : textPositions) {

-            boolean lineSeparation = prev != null && word.getMinYDirAdj() - maxY > Math.min(word.getHeight(), prev.getHeight()) * 1.25;
-            boolean startFromTop = prev != null && word.getMinYDirAdj() < prev.getMinYDirAdj() - prev.getTextHeight();
-            boolean splitByX = prev != null && maxX + 50 < word.getMinXDirAdj() && prev.getMinYDirAdj() == word.getMinYDirAdj();
-            boolean negativeXGap = prev != null && word.getMinXDirAdj() - minX < -5;
-            boolean newLineAfterSplit = prev != null && word.getMinYDirAdj() != prev.getMinYDirAdj() && wasSplitted && splitX1 != word.getMinXDirAdj();
+            boolean lineSeparation = prev != null && word.getYDirAdj() - prev.getMaxYDirAdj() > Math.min(word.getHeight(), prev.getHeight()) * 1.1;
+            boolean startFromTop = prev != null && word.getYDirAdj() < prev.getYDirAdj() - prev.getTextHeight();
+            boolean splitByX = prev != null && maxX + 50 < word.getXDirAdj() && prev.getYDirAdj() == word.getYDirAdj();
+            boolean negativeXGap = prev != null && word.getXDirAdj() - minX < -5;
+            boolean newLineAfterSplit = prev != null && word.getYDirAdj() != prev.getYDirAdj() && wasSplitted && splitX1 != word.getXDirAdj();
            boolean isSplitByRuling = prev != null && usedRulings.lineBetween(prev, word);
            boolean splitByDir = prev != null && !prev.getDir().equals(word.getDir());
-            boolean splitByOtherFontAndOtherY = prev != null && Math.abs(prev.getMaxYDirAdj() - word.getMaxYDirAdj()) > word.getTextHeight() * 0.2 && (word.getFontStyle()
-                                                                                                                                                               .contains("bold")
-                                                                                                                                                       && !prev.getFontStyle()
-                    .contains("bold") || prev.getFontStyle().contains("bold") && !word.getFontStyle().contains("bold"));
+            boolean splitByOtherFontAndOtherY = prev != null && Math.abs(prev.getMaxYDirAdj() - word.getMaxYDirAdj()) > word.getTextHeight() * 0.2 //
+                                                && (word.getFontStyle().contains("bold") && !prev.getFontStyle().contains("bold") //
+                                                    || prev.getFontStyle().contains("bold") && !word.getFontStyle().contains("bold"));

            Matcher matcher = pattern.matcher(chunkWords.stream()
                                                      .collect(Collectors.joining(" ")).toString());
-            boolean startsOnSameX = Math.abs(minX - word.getMinXDirAdj()) < 5 && matcher.matches();
+            boolean startsOnSameX = Math.abs(minX - word.getXDirAdj()) < 5 && matcher.matches();

            if (prev != null && (lineSeparation || startFromTop || splitByDir || isSplitByRuling || splitByOtherFontAndOtherY || negativeXGap || startsOnSameX)) {

@@ -84,7 +78,7 @@ public class DocuMineBlockificationService {
                if (splitByX && !isSplitByRuling) {
                    wasSplitted = true;
                    cb1.setOrientation(Orientation.LEFT);
-                    splitX1 = word.getMinXDirAdj();
+                    splitX1 = word.getXDirAdj();
                } else if (newLineAfterSplit && !isSplitByRuling) {
                    wasSplitted = false;
                    cb1.setOrientation(Orientation.RIGHT);
@@ -107,14 +101,14 @@ public class DocuMineBlockificationService {
            chunkWords.add(word);

            prev = word;
-            if (word.getMinXDirAdj() < minX) {
-                minX = word.getMinXDirAdj();
+            if (word.getXDirAdj() < minX) {
+                minX = word.getXDirAdj();
            }
            if (word.getMaxXDirAdj() > maxX) {
                maxX = word.getMaxXDirAdj();
            }
-            if (word.getMinYDirAdj() < minY) {
-                minY = word.getMinYDirAdj();
+            if (word.getYDirAdj() < minY) {
+                minY = word.getYDirAdj();
            }
            if (word.getMaxYDirAdj() > maxY) {
                maxY = word.getMaxYDirAdj();
@@ -126,7 +120,5 @@ public class DocuMineBlockificationService {
        return new ClassificationPage(textPageBlocks);
    }

-
-
 }

@@ -1,9 +1,6 @@
 package com.knecon.fforesight.service.layoutparser.processor.services.blockification;

-import static java.util.stream.Collectors.toSet;
-
 import java.util.ArrayList;
-import java.util.Comparator;
 import java.util.Iterator;
 import java.util.List;

@@ -11,13 +8,11 @@ import org.springframework.stereotype.Service;

 import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
-import com.knecon.fforesight.service.layoutparser.processor.model.FloatFrequencyCounter;
 import com.knecon.fforesight.service.layoutparser.processor.model.Orientation;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRulings;
-import com.knecon.fforesight.service.layoutparser.processor.model.text.StringFrequencyCounter;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
-import com.knecon.fforesight.service.layoutparser.processor.visualization.LayoutparsingVisualizations;
+import com.knecon.fforesight.service.layoutparser.processor.visualization.LayoutDebugLayer;

@SuppressWarnings("all")
@Service
@@ -35,7 +30,7 @@ public class RedactManagerBlockificationService {
     * @param visualizations
     * @return Page object that contains the Textblock and text statistics.
     */
-    public ClassificationPage blockify(List<TextPositionSequence> textPositions, CleanRulings cleanRulings, LayoutparsingVisualizations visualizations) {
+    public ClassificationPage blockify(List<TextPositionSequence> textPositions, CleanRulings cleanRulings, LayoutDebugLayer visualizations) {

        CleanRulings usedRulings = cleanRulings.withoutTextRulings();

@@ -43,18 +38,18 @@ public class RedactManagerBlockificationService {
        List<TextPositionSequence> chunkWords = new ArrayList<>();
        List<AbstractPageBlock> chunkBlockList = new ArrayList<>();

-        float minX = 1000, maxX = 0, minY = 1000, maxY = 0;
+        double minX = 1000, maxX = 0, minY = 1000, maxY = 0;
        TextPositionSequence prev = null;

        boolean wasSplitted = false;
-        Float splitX1 = null;
+        Double splitX1 = null;
        for (TextPositionSequence word : textPositions) {

-            boolean lineSeparation = word.getMinYDirAdj() - maxY > word.getHeight() * 1.25;
-            boolean startFromTop = prev != null && word.getMinYDirAdj() < prev.getMinYDirAdj() - prev.getTextHeight();
-            boolean splitByX = prev != null && maxX + 50 < word.getMinXDirAdj() && prev.getMinYDirAdj() == word.getMinYDirAdj();
-            boolean xIsBeforeFirstX = prev != null && word.getMinXDirAdj() < minX;
-            boolean newLineAfterSplit = prev != null && word.getMinYDirAdj() != prev.getMinYDirAdj() && wasSplitted && splitX1 != word.getMinXDirAdj();
+            boolean lineSeparation = word.getYDirAdj() - maxY > word.getHeight() * 1.25;
+            boolean startFromTop = prev != null && word.getYDirAdj() < prev.getYDirAdj() - prev.getTextHeight();
+            boolean splitByX = prev != null && maxX + 50 < word.getXDirAdj() && prev.getYDirAdj() == word.getYDirAdj();
+            boolean xIsBeforeFirstX = prev != null && word.getXDirAdj() < minX;
+            boolean newLineAfterSplit = prev != null && word.getYDirAdj() != prev.getYDirAdj() && wasSplitted && splitX1 != word.getXDirAdj();
            boolean isSplitByRuling = prev != null && usedRulings.lineBetween(prev, word);
            boolean splitByDir = prev != null && !prev.getDir().equals(word.getDir());

@@ -74,7 +69,7 @@ public class RedactManagerBlockificationService {
                if (splitByX && !isSplitByRuling) {
                    wasSplitted = true;
                    cb1.setOrientation(Orientation.LEFT);
-                    splitX1 = word.getMinXDirAdj();
+                    splitX1 = word.getXDirAdj();
                } else if (newLineAfterSplit && !isSplitByRuling) {
                    wasSplitted = false;
                    cb1.setOrientation(Orientation.RIGHT);
@@ -97,14 +92,14 @@ public class RedactManagerBlockificationService {
            chunkWords.add(word);

            prev = word;
-            if (word.getMinXDirAdj() < minX) {
-                minX = word.getMinXDirAdj();
+            if (word.getXDirAdj() < minX) {
+                minX = word.getXDirAdj();
            }
            if (word.getMaxXDirAdj() > maxX) {
                maxX = word.getMaxXDirAdj();
            }
-            if (word.getMinYDirAdj() < minY) {
-                minY = word.getMinYDirAdj();
+            if (word.getYDirAdj() < minY) {
+                minY = word.getYDirAdj();
            }
            if (word.getMaxYDirAdj() > maxY) {
                maxY = word.getMaxYDirAdj();
@@ -23,7 +23,7 @@ public class ClarifyndClassificationService {

    public void classifyDocument(ClassificationDocument document) {

-        List<Float> headlineFontSizes = document.getFontSizeCounter().getHighterThanMostPopular();
+        List<Double> headlineFontSizes = document.getFontSizeCounter().getHigherThanMostPopular();

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

@@ -35,7 +35,7 @@ public class ClarifyndClassificationService {
    }


-    private void classifyPage(HeadlineClassificationService headlineClassificationService, ClassificationPage page, ClassificationDocument document, List<Float> headlineFontSizes) {
+    private void classifyPage(HeadlineClassificationService headlineClassificationService, ClassificationPage page, ClassificationDocument document, List<Double> headlineFontSizes) {

        for (AbstractPageBlock textBlock : page.getTextBlocks()) {
            if (textBlock instanceof TextPageBlock) {
@@ -45,7 +45,7 @@ public class ClarifyndClassificationService {
    }


-    private void classifyBlock(HeadlineClassificationService headlineClassificationService, TextPageBlock textBlock, ClassificationPage page, ClassificationDocument document, List<Float> headlineFontSizes) {
+    private void classifyBlock(HeadlineClassificationService headlineClassificationService, TextPageBlock textBlock, ClassificationPage page, ClassificationDocument document, List<Double> headlineFontSizes) {

        var bodyTextFrame = page.getBodyTextFrame();

@@ -31,7 +31,7 @@ public class DocuMineClassificationService {

    public void classifyDocument(ClassificationDocument document) {

-        List<Float> headlineFontSizes = document.getFontSizeCounter().getHighterThanMostPopular();
+        List<Double> headlineFontSizes = document.getFontSizeCounter().getHigherThanMostPopular();

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

@@ -46,7 +46,7 @@ public class DocuMineClassificationService {
    private void classifyPage(HeadlineClassificationService headlineClassificationService,
                              ClassificationPage page,
                              ClassificationDocument document,
-                              List<Float> headlineFontSizes) {
+                              List<Double> headlineFontSizes) {

        for (AbstractPageBlock textBlock : page.getTextBlocks()) {
            if (textBlock instanceof TextPageBlock) {
@@ -60,7 +60,7 @@ public class DocuMineClassificationService {
                               TextPageBlock textBlock,
                               ClassificationPage page,
                               ClassificationDocument document,
-                               List<Float> headlineFontSizes) {
+                               List<Double> headlineFontSizes) {

        log.debug("headlineFontSizes: {}", headlineFontSizes);
        var bodyTextFrame = page.getBodyTextFrame();
@@ -25,7 +25,7 @@ public class RedactManagerClassificationService {

    public void classifyDocument(ClassificationDocument document) {

-        List<Float> headlineFontSizes = document.getFontSizeCounter().getHighterThanMostPopular();
+        List<Double> headlineFontSizes = document.getFontSizeCounter().getHigherThanMostPopular();

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

@@ -37,7 +37,7 @@ public class RedactManagerClassificationService {
    }


-    private void classifyPage(HeadlineClassificationService headlineClassificationService, ClassificationPage page, ClassificationDocument document, List<Float> headlineFontSizes) {
+    private void classifyPage(HeadlineClassificationService headlineClassificationService, ClassificationPage page, ClassificationDocument document, List<Double> headlineFontSizes) {

        for (AbstractPageBlock textBlock : page.getTextBlocks()) {
            if (textBlock instanceof TextPageBlock) {
@@ -47,7 +47,7 @@ public class RedactManagerClassificationService {
    }


-    private void classifyBlock(HeadlineClassificationService headlineClassificationService, TextPageBlock textBlock, ClassificationPage page, ClassificationDocument document, List<Float> headlineFontSizes) {
+    private void classifyBlock(HeadlineClassificationService headlineClassificationService, TextPageBlock textBlock, ClassificationPage page, ClassificationDocument document, List<Double> headlineFontSizes) {

        var bodyTextFrame = page.getBodyTextFrame();

@@ -56,7 +56,7 @@ public class RedactManagerClassificationService {
            return;
        }
        if (document.getFontSizeCounter().getMostPopular() == null) {
-            textBlock.setClassification(PageBlockType.OTHER);
+            textBlock.setClassification(PageBlockType.PARAGRAPH);
            return;
        }

@@ -129,7 +129,7 @@ public class RedactManagerClassificationService {
        } else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock)) {
            textBlock.setClassification(PageBlockType.PARAGRAPH_UNKNOWN);
        } else {
-            textBlock.setClassification(PageBlockType.OTHER);
+            textBlock.setClassification(PageBlockType.PARAGRAPH);
        }
    }

@@ -22,6 +22,7 @@ import com.knecon.fforesight.service.layoutparser.processor.model.Classification
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationHeader;
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.AbstractSemanticNode;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.DuplicatedParagraph;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Footer;
@@ -31,7 +32,6 @@ import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.He
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Image;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Page;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Paragraph;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.AbstractSemanticNode;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.AtomicTextBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.image.ClassifiedImage;
 import com.knecon.fforesight.service.layoutparser.processor.model.outline.TableOfContentItem;
@@ -45,15 +45,17 @@ import lombok.Builder;
 import lombok.Getter;
 import lombok.experimental.FieldDefaults;
 import lombok.experimental.UtilityClass;
+import lombok.extern.slf4j.Slf4j;

@UtilityClass
+@Slf4j
 public class DocumentGraphFactory {

    public Document buildDocumentGraph(LayoutParsingType layoutParsingType, ClassificationDocument document) {

        Document documentGraph = new Document();

-        documentGraph.setVisualizations(document.getVisualizations());
+        documentGraph.setLayoutDebugLayer(document.getLayoutDebugLayer());

        Context context = new Context(documentGraph);

@@ -109,10 +111,10 @@ public class DocumentGraphFactory {
        textBlocks.add(originalTextBlock);
        textBlocks.addAll(textBlocksToMerge);

-        AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock(TextPositionOperations.mergeAndSortTextPositionSequenceByYThenX(textBlocks), node, context, page);
+        AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock2(TextPositionOperations.mergeAndSort(textBlocks), node, context, page);

        if (node instanceof DuplicatedParagraph duplicatedParagraph) {
-            AtomicTextBlock unsortedTextBlock = context.textBlockFactory.buildAtomicTextBlock(textBlocks.stream()
+            AtomicTextBlock unsortedTextBlock = context.textBlockFactory.buildAtomicTextBlock2(textBlocks.stream()
                                                                                                      .flatMap(tb -> tb.getSequences()
                                                                                                              .stream())
                                                                                                      .collect(Collectors.toList()), node, context, page);
@@ -145,6 +147,7 @@ public class DocumentGraphFactory {
                .position(position)
                .transparent(image.isHasTransparency())
                .page(page)
+                .representationHash(image.getRepresentation())
                .documentTree(context.getDocumentTree())
                .build();
        page.getMainBody().add(imageNode);
@@ -188,7 +191,7 @@ public class DocumentGraphFactory {

        Page page = context.getPage(textBlocks.get(0).getPage());
        Footer footer = Footer.builder().documentTree(context.getDocumentTree()).build();
-        AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock(TextPositionOperations.mergeTextPositionSequence(textBlocks), footer, context, page);
+        AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock2(TextPositionOperations.merge(textBlocks), footer, context, page);
        List<Integer> tocId = context.getDocumentTree().createNewMainEntryAndReturnId(footer);
        footer.setTreeId(tocId);
        footer.setLeafTextBlock(textBlock);
@@ -200,7 +203,7 @@ public class DocumentGraphFactory {

        Page page = context.getPage(textBlocks.get(0).getPage());
        Header header = Header.builder().documentTree(context.getDocumentTree()).build();
-        AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock(TextPositionOperations.mergeTextPositionSequence(textBlocks), header, 0, page);
+        AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock(TextPositionOperations.merge(textBlocks), header, 0, page);
        List<Integer> tocId = context.getDocumentTree().createNewMainEntryAndReturnId(header);
        header.setTreeId(tocId);
        header.setLeafTextBlock(textBlock);
@@ -277,7 +280,8 @@ public class DocumentGraphFactory {
            return pages.keySet()
                    .stream()
                    .filter(page -> page.getNumber() == pageIndex)
-                    .findFirst().orElseThrow(() -> new NoSuchElementException(format("ClassificationPage with number %d not found", pageIndex)));
+                    .findFirst()
+                    .orElseThrow(() -> new NoSuchElementException(format("ClassificationPage with number %d not found", pageIndex)));
        }

    }
@@ -4,7 +4,7 @@ import java.awt.geom.Rectangle2D;
 import java.util.Collections;
 import java.util.List;

-import com.knecon.fforesight.service.layoutparser.processor.model.graph.Boundary;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.TextRange;

 import lombok.AccessLevel;
 import lombok.Builder;
@@ -19,8 +19,8 @@ public class SearchTextWithTextPositionDto {
    String searchText;
    List<Integer> lineBreaks;
    List<Integer> stringIdxToPositionIdx;
-    List<Boundary> boldTextBoundaries;
-    List<Boundary> italicTextBoundaries;
+    List<TextRange> boldTextBoundaries;
+    List<TextRange> italicTextBoundaries;
    List<Rectangle2D> positions;


@@ -1,6 +1,7 @@
 package com.knecon.fforesight.service.layoutparser.processor.services.factory;

-import java.awt.geom.AffineTransform;
+import static com.knecon.fforesight.service.layoutparser.processor.model.text.RedTextPosition.HEIGHT_PADDING;
+
 import java.awt.geom.Rectangle2D;
 import java.util.Collection;
 import java.util.Collections;
@@ -9,9 +10,8 @@ import java.util.List;
 import java.util.Locale;
 import java.util.Objects;

-import com.knecon.fforesight.service.layoutparser.processor.model.graph.Boundary;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.TextRange;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.RedTextPosition;
-import com.knecon.fforesight.service.layoutparser.processor.model.text.TextDirection;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;

 import lombok.experimental.UtilityClass;
@@ -19,13 +19,13 @@ import lombok.experimental.UtilityClass;
@UtilityClass
 public class SearchTextWithTextPositionFactory {

-    public final int HEIGHT_PADDING = 2;
    // when checking for a hyphen linebreak, we need to check after a linebreak if the last hyphen was less than three symbols away.
    // We detect a linebreak as either a "\n" character or if two adjacent symbol's position differ in y-coordinates by at least one character height.
    // If there is a hyphen linebreak, the hyphen will be 1 position in front of a "\n" or 2 positions in front of the character which has a lower y-coordinate
    // This is why, we need to initialize this to < -2, otherwise, if the very first symbol is a \n we would detect a hyphen linebreak that isn't there.
    // Also, Integer.MIN_VALUE is a bad idea due to potential overflow during arithmetic operations. This is why the default should be -3.
    public final int MAX_HYPHEN_LINEBREAK_DISTANCE = 3;
+    public static final double LINEBREAK_DELTA_TOLERANCE = 1.5;


    public SearchTextWithTextPositionDto buildSearchTextToTextPositionDto(List<TextPositionSequence> sequences) {
@@ -37,15 +37,13 @@ public class SearchTextWithTextPositionFactory {

        Context context = new Context();

-        RedTextPosition currentTextPosition = sequences.get(0).getTextPositions()
-                .get(0);
+        RedTextPosition currentTextPosition = sequences.get(0).getTextPositions().get(0);
        RedTextPosition previousTextPosition = RedTextPosition.builder().unicode(" ").bBoxDirAdj(currentTextPosition.getBBoxDirAdj()).build();

        for (TextPositionSequence word : sequences) {
            for (int i = 0; i < word.getTextPositions().size(); ++i) {

-                currentTextPosition = word.getTextPositions()
-                        .get(i);
+                currentTextPosition = word.getTextPositions().get(i);
                if (isLineBreak(currentTextPosition, previousTextPosition)) {
                    removeHyphenLinebreaks(context);
                    context.lineBreaksStringIdx.add(context.stringIdx);
@@ -67,11 +65,10 @@ public class SearchTextWithTextPositionFactory {
            ++context.stringIdx;
        }

-
        List<Rectangle2D> positions = sequences.stream()
                .map(TextPositionSequence::getTextPositions)
                .flatMap(Collection::stream)
-                .map(RedTextPosition::getBBoxInitialUserSpace)
+                .map(RedTextPosition::getBBoxPdf)
                .toList();

        assert context.stringBuilder.length() == context.stringIdxToPositionIdx.size();
@@ -118,23 +115,23 @@ public class SearchTextWithTextPositionFactory {
    }


-    private static List<Boundary> mergeToBoundaries(List<Integer> integers) {
+    private static List<TextRange> mergeToBoundaries(List<Integer> integers) {

        if (integers.isEmpty()) {
            return Collections.emptyList();
        }
-        List<Boundary> boundaries = new LinkedList<>();
+        List<TextRange> boundaries = new LinkedList<>();
        int start = integers.get(0);
        int end = integers.get(0) + 1;
        for (int current : integers) {
            if (current > end + 1) {
-                boundaries.add(new Boundary(start, end));
+                boundaries.add(new TextRange(start, end));
                start = current;
            }
            end = current + 1;
        }
        if (boundaries.isEmpty()) {
-            boundaries.add(new Boundary(start, end));
+            boundaries.add(new TextRange(start, end));
        }
        return boundaries;
    }
@@ -160,8 +157,8 @@ public class SearchTextWithTextPositionFactory {
            return false;
        }

-        double deltaY = Math.abs(currentPosition.getYDirAdj() - previousPosition.getYDirAdj());
-        return deltaY >= currentPosition.getHeightDir();
+        double deltaY = (Math.abs(currentPosition.getYDirAdj() - previousPosition.getYDirAdj()) * LINEBREAK_DELTA_TOLERANCE) + (2 * HEIGHT_PADDING);
+        return deltaY >= currentPosition.getHeightDirAdj() || deltaY >= previousPosition.getHeightDirAdj();
    }


@@ -187,32 +184,6 @@ public class SearchTextWithTextPositionFactory {
    }


-    public Rectangle2D mapRedTextPositionToInitialUserSpace(RedTextPosition textPosition, TextPositionSequence sequence) {
-
-        float textHeight = sequence.getTextHeight() + HEIGHT_PADDING;
-        Rectangle2D rectangle2D = new Rectangle2D.Double(textPosition.getXDirAdj(),
-                textPosition.getYDirAdj() - textHeight,
-                textPosition.getWidthDirAdj(),
-                textHeight + HEIGHT_PADDING);
-
-        AffineTransform transform = new AffineTransform();
-
-        if (sequence.getDir() == TextDirection.ZERO || sequence.getDir() == TextDirection.HALF_CIRCLE) {
-            transform.rotate(sequence.getDir().getRadians(), sequence.getPageWidth() / 2f, sequence.getPageHeight() / 2f);
-            transform.translate(0f, sequence.getPageHeight());
-        } else if (sequence.getDir() == TextDirection.QUARTER_CIRCLE) {
-            transform.rotate(sequence.getDir().getRadians(), sequence.getPageWidth() / 2f, sequence.getPageWidth() / 2f);
-            transform.translate(0f, sequence.getPageWidth());
-        } else {
-            transform.rotate(sequence.getDir().getRadians(), sequence.getPageHeight() / 2f, sequence.getPageHeight() / 2f);
-            transform.translate(0f, sequence.getPageWidth());
-        }
-        transform.scale(1., -1.);
-
-        return transform.createTransformedShape(rectangle2D).getBounds2D();
-    }
-
-
    private class Context {

        List<Integer> stringIdxToPositionIdx = new LinkedList<>();
@@ -234,7 +234,7 @@ public class SectionNodeFactory {
                .filter(abstractTextContainer -> !abstractTextContainer.equals(atc))
                .filter(abstractTextContainer -> abstractTextContainer.getPage() == atc.getPage())
                .filter(abstractTextContainer -> abstractTextContainer instanceof TextPageBlock)
-                .filter(abstractTextContainer -> abstractTextContainer.intersectsY(atc))
+                .filter(abstractTextContainer -> abstractTextContainer.intersectsYPdf(atc))
                .map(abstractTextContainer -> (TextPageBlock) abstractTextContainer)
                .filter(abstractTextContainer -> abstractTextContainer.getDir() == atc.getDir())
                .filter(abstractTextContainer -> !abstractTextContainer.isToDuplicate())
@@ -136,7 +136,7 @@ public class TableNodeFactory {
                .row(rowIndex)
                .col(colIndex)
                .header(cell.isHeaderCell())
-                .bBox(cell.getBBoxInitialUserSpace())
+                .bBox(cell.getBBoxPdf())
                .build();
        page.getMainBody().add(tableCell);

@@ -148,7 +148,7 @@ public class TableNodeFactory {
            tableCell.setLeafTextBlock(context.getTextBlockFactory().emptyTextBlock(tableNode, context, page));
        } else if (cell.getTextBlocks().size() == 1) {
            textBlock = context.getTextBlockFactory()
-                    .buildAtomicTextBlock(cell.getTextBlocks()
+                    .buildAtomicTextBlock2(cell.getTextBlocks()
                                                  .get(0).getSequences(), tableCell, context, page);
            tableCell.setLeafTextBlock(textBlock);
        } else if (firstTextBlockIsHeadline(cell)) {
@@ -163,8 +163,8 @@ public class TableNodeFactory {
                                          context,
                                          document);
        } else if (cellAreaIsSmallerThanPageAreaTimesThreshold(cell, page)) {
-            List<TextPositionSequence> sequences = TextPositionOperations.mergeAndSortTextPositionSequenceByYThenX(cell.getTextBlocks());
-            textBlock = context.getTextBlockFactory().buildAtomicTextBlock(sequences, tableCell, context, page);
+            List<TextPositionSequence> sequences = TextPositionOperations.mergeAndSort(cell.getTextBlocks());
+            textBlock = context.getTextBlockFactory().buildAtomicTextBlock2(sequences, tableCell, context, page);
            tableCell.setLeafTextBlock(textBlock);
        } else {
            cell.getTextBlocks()
@@ -17,7 +17,7 @@ public class TextBlockFactory {
    long textBlockIdx;


-    public AtomicTextBlock buildAtomicTextBlock(List<TextPositionSequence> sequences, SemanticNode parent, DocumentGraphFactory.Context context, Page page) {
+    public AtomicTextBlock buildAtomicTextBlock2(List<TextPositionSequence> sequences, SemanticNode parent, DocumentGraphFactory.Context context, Page page) {

        Integer numberOnPage = context.getAndIncrementTextBlockNumberOnPage(page);
        return buildAtomicTextBlock(sequences, parent, numberOnPage, page);
@@ -32,27 +32,27 @@ public class TextBlockFactory {
        long idx = textBlockIdx;
        textBlockIdx++;
        String orientation;
-        int textDirection;
+        int textRotation;
        if (sequences.isEmpty()) {
            orientation = null;
-            textDirection = 0;
+            textRotation = 0;
        } else {
            orientation = sequences.get(0).getDir().toString();
-            textDirection = sequences.get(0).getRotation();
+            textRotation = sequences.get(0).getDir().getRotation();
        }
        return AtomicTextBlock.fromSearchTextWithTextPosition(searchTextWithTextPositionDto.getSearchText(),
-                searchTextWithTextPositionDto.getLineBreaks(),
-                searchTextWithTextPositionDto.getBoldTextBoundaries(),
-                searchTextWithTextPositionDto.getItalicTextBoundaries(),
-                searchTextWithTextPositionDto.getPositions(),
-                searchTextWithTextPositionDto.getStringIdxToPositionIdx(),
-                idx,
-                parent,
-                numberOnPage,
-                page,
-                offset,
-                orientation,
-                textDirection);
+                                                              searchTextWithTextPositionDto.getLineBreaks(),
+                                                              searchTextWithTextPositionDto.getBoldTextBoundaries(),
+                                                              searchTextWithTextPositionDto.getItalicTextBoundaries(),
+                                                              searchTextWithTextPositionDto.getPositions(),
+                                                              searchTextWithTextPositionDto.getStringIdxToPositionIdx(),
+                                                              idx,
+                                                              parent,
+                                                              numberOnPage,
+                                                              page,
+                                                              offset,
+                                                              orientation,
+                                                              textRotation);
    }


@@ -8,12 +8,11 @@ import org.apache.pdfbox.pdmodel.PDDocument;
 import org.apache.pdfbox.pdmodel.PDPage;
 import org.springframework.stereotype.Service;

+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.BoundingBox;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRulings;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.Ruling;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
-import com.knecon.fforesight.service.layoutparser.processor.services.factory.SearchTextWithTextPositionFactory;
 import com.knecon.fforesight.service.layoutparser.processor.utils.PageInformation;
-import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;

 import lombok.RequiredArgsConstructor;
 import lombok.SneakyThrows;
@@ -67,10 +66,7 @@ public class GraphicExtractorService {
    private List<Box> getCharacterBBoxes(List<TextPositionSequence> textPositionSequences) {

        return textPositionSequences.stream()
-                .map(pos -> pos.getTextPositions()
-                        .stream()
-                        .map(tp -> SearchTextWithTextPositionFactory.mapRedTextPositionToInitialUserSpace(tp, pos))
-                        .collect(RectangleTransformations.collectBBox()))
+                .map(BoundingBox::getBBoxPdf)
                .map(Box::new)
                .collect(Collectors.toList());
    }
@@ -116,8 +116,8 @@ public class DocumentDataMapper {
                .page(atomicTextBlock.getPage().getNumber().longValue())
                .searchText(atomicTextBlock.getSearchText())
                .numberOnPage(atomicTextBlock.getNumberOnPage())
-                .start(atomicTextBlock.getBoundary().start())
-                .end(atomicTextBlock.getBoundary().end())
+                .start(atomicTextBlock.getTextRange().start())
+                .end(atomicTextBlock.getTextRange().end())
                .lineBreaks(toPrimitiveIntArray(atomicTextBlock.getLineBreaks()))
                .build();
    }
@@ -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;
    }

@@ -13,7 +13,7 @@ import com.knecon.fforesight.service.layoutparser.internal.api.data.taas.Researc
 import com.knecon.fforesight.service.layoutparser.internal.api.data.taas.RowData;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.taas.StructureObject;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.taas.TableData;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.Boundary;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.TextRange;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Page;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.SemanticNode;
@@ -82,15 +82,15 @@ public class TaasDocumentDataMapper {
    }


-    private static Range toRange(Boundary boundary) {
+    private static Range toRange(TextRange textRange) {

-        return new Range(boundary.start(), boundary.end());
+        return new Range(textRange.start(), textRange.end());
    }


-    private static List<Range> toRange(List<Boundary> boundary) {
+    private static List<Range> toRange(List<TextRange> textRange) {

-        return boundary.stream().map(TaasDocumentDataMapper::toRange).toList();
+        return textRange.stream().map(TaasDocumentDataMapper::toRange).toList();
    }


@@ -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())
@@ -82,6 +82,7 @@ public class LegacyPDFStreamEngine extends PDFStreamEngine {

    private int pageRotation;
    private PDRectangle pageSize;
+    private Matrix translateMatrix;
    private final GlyphList glyphList;
    private final Map<COSDictionary, Float> fontHeightMap = new WeakHashMap<COSDictionary, Float>();

@@ -133,6 +134,12 @@ public class LegacyPDFStreamEngine extends PDFStreamEngine {
        this.pageRotation = page.getRotation();
        this.pageSize = page.getCropBox();

+        if (pageSize.getLowerLeftX() == 0 && pageSize.getLowerLeftY() == 0) {
+            translateMatrix = null;
+        } else {
+            // translation matrix for cropbox
+            translateMatrix = Matrix.getTranslateInstance(-pageSize.getLowerLeftX(), -pageSize.getLowerLeftY());
+        }
        super.processPage(page);
    }

@@ -257,13 +264,22 @@ public class LegacyPDFStreamEngine extends PDFStreamEngine {
                return;
            }
        }
+        // adjust for cropbox if needed
+        Matrix translatedTextRenderingMatrix;
+        if (translateMatrix == null) {
+            translatedTextRenderingMatrix = textRenderingMatrix;
+        } else {
+            translatedTextRenderingMatrix = Matrix.concatenate(translateMatrix, textRenderingMatrix);
+            nextX -= pageSize.getLowerLeftX();
+            nextY -= pageSize.getLowerLeftY();
+        }

        // This is a hack for unicode letter with 2 chars e.g. RA see unicodeProblem.pdf
        if (unicodeMapping.length() == 2) {
            processTextPosition(new TextPosition(pageRotation,
                                                 pageSize.getWidth(),
                                                 pageSize.getHeight(),
-                                                 textRenderingMatrix,
+                                                 translatedTextRenderingMatrix,
                                                 nextX,
                                                 nextY,
                                                 Math.abs(dyDisplay),
@@ -277,7 +293,7 @@ public class LegacyPDFStreamEngine extends PDFStreamEngine {
            processTextPosition(new TextPosition(pageRotation,
                                                 pageSize.getWidth(),
                                                 pageSize.getHeight(),
-                                                 textRenderingMatrix,
+                                                 translatedTextRenderingMatrix,
                                                 nextX,
                                                 nextY,
                                                 Math.abs(dyDisplay),
@@ -293,7 +309,7 @@ public class LegacyPDFStreamEngine extends PDFStreamEngine {
            processTextPosition(new TextPosition(pageRotation,
                                                 pageSize.getWidth(),
                                                 pageSize.getHeight(),
-                                                 textRenderingMatrix,
+                                                 translatedTextRenderingMatrix,
                                                 nextX,
                                                 nextY,
                                                 Math.abs(dyDisplay),
@@ -1,41 +1,17 @@
 package com.knecon.fforesight.service.layoutparser.processor.services.visualization;

-import java.awt.Color;
-import java.awt.geom.Line2D;
-import java.awt.geom.Point2D;
-import java.awt.geom.Rectangle2D;
-import java.awt.geom.RectangularShape;
 import java.io.File;
-import java.util.ArrayList;
-import java.util.Collections;
-import java.util.Comparator;
-import java.util.HashMap;
 import java.util.List;
-import java.util.Map;
-import java.util.Optional;
-import java.util.stream.Collectors;
-import java.util.stream.Stream;

 import org.springframework.stereotype.Service;

-import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
-import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.DuplicatedParagraph;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Page;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.SemanticNode;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.SuperSection;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Headline;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Image;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Paragraph;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Table;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.TableCell;
-import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
-import com.knecon.fforesight.service.viewerdoc.ContentStreams;
-import com.knecon.fforesight.service.viewerdoc.model.ColoredLine;
-import com.knecon.fforesight.service.viewerdoc.model.ColoredRectangle;
-import com.knecon.fforesight.service.viewerdoc.model.LayoutGrid;
-import com.knecon.fforesight.service.viewerdoc.model.PlacedText;
-import com.knecon.fforesight.service.viewerdoc.model.Standard14EmbeddableFont;
-import com.knecon.fforesight.service.viewerdoc.model.Visualizations;
-import com.knecon.fforesight.service.viewerdoc.service.ViewerDocumentService;
+import com.knecon.fforesight.service.layoutparser.processor.visualization.LayoutGrid;
+import com.knecon.fforesight.service.viewerdoc.service.PDFTronViewerDocumentService;

 import io.micrometer.observation.annotation.Observed;
 import lombok.AccessLevel;
@@ -48,451 +24,41 @@ import lombok.experimental.FieldDefaults;
@FieldDefaults(level = AccessLevel.PRIVATE, makeFinal = true)
 public class LayoutGridService {

-    ViewerDocumentService viewerDocumentService;
-
-    static float FONT_SIZE = 10f;
-    static float LINE_WIDTH = 1f;
-    static Standard14EmbeddableFont FONT = Standard14EmbeddableFont.helvetica();
-
-    static Color INNER_LINES_COLOR = new Color(255, 175, 175);
-    static Color PARAGRAPH_COLOR = new Color(70, 130, 180);
-
-    static Color DUPLICATE_PARAGRAPH_COLOR = new Color(70, 180, 101);
-    static Color TABLE_COLOR = new Color(102, 205, 170);
-    static Color SECTION_COLOR = new Color(50, 50, 50);
-    static Color HEADLINE_COLOR = new Color(162, 56, 56);
-    static Color HEADER_COLOR = new Color(171, 131, 6);
-    static Color IMAGE_COLOR = new Color(253, 63, 146);
-
-    private record RectangleIdentifier(List<Integer> treeId, Integer pageNumber) {
-
-    }
-
-    HashMap<RectangleIdentifier, Rectangle2D> rectangleMap = new HashMap<>();
+    PDFTronViewerDocumentService viewerDocumentService;


    @SneakyThrows
    @Observed(name = "ViewerDocumentService", contextualName = "create-viewer-document")
    public void addLayoutGrid(File originFile, Document document, File destinationFile, boolean layerVisibilityDefaultValue, boolean writeVisualLayoutParsingGrid) {

-        List<Visualizations> allVisualizations;
-        Visualizations layoutGrid = this.addLayoutGrid(document, layerVisibilityDefaultValue, false);
-        if (writeVisualLayoutParsingGrid) {
-            Visualizations visualLayoutGrid = this.addLayoutGrid(document, layerVisibilityDefaultValue, true);
-            allVisualizations = Stream.concat(Stream.of(layoutGrid, visualLayoutGrid), document.getVisualizations().streamAll())
-                    .toList();
+        LayoutGrid layoutGrid = createLayoutGrid(document);
+        layoutGrid.setVisibleByDefault(layerVisibilityDefaultValue);
+//        Visualizations visualLayoutGrid = this.addLayoutGrid(document, layerVisibilityDefaultValue, true);
+        if (document.getLayoutDebugLayer().isActive()) {
+            viewerDocumentService.addLayerGroups(originFile, destinationFile, List.of(layoutGrid, document.getLayoutDebugLayer()));
        } else {
-            allVisualizations = Stream.concat(Stream.of(layoutGrid), document.getVisualizations().streamAll())
-                    .toList();
+            viewerDocumentService.addLayerGroups(originFile, destinationFile, List.of(layoutGrid));
+
        }
-
-        viewerDocumentService.addVisualizationsOnPage(originFile, destinationFile, allVisualizations);
    }


-    @SneakyThrows
-    @Observed(name = "ViewerDocumentService", contextualName = "create-viewer-document")
-    public Visualizations addLayoutGrid(Document document, boolean layerVisibilityDefaultValue, boolean visualParsingGrid) {
+    private LayoutGrid createLayoutGrid(Document document) {

-        LayoutGrid layoutGrid = createLayoutGrid(document, visualParsingGrid);
-
-        return Visualizations.builder()
-                .layer(visualParsingGrid ? ContentStreams.KNECON_VISUAL_PARSING : ContentStreams.KNECON_LAYOUT)
-                .visualizationsOnPages(layoutGrid.getVisualizationsPerPages())
-                .layerVisibilityDefaultValue(layerVisibilityDefaultValue)
-                .build();
-    }
-
-
-    private LayoutGrid createLayoutGrid(Document document, boolean visualParsingGrid) {
-
-        LayoutGrid layoutGrid = new LayoutGrid(document.getNumberOfPages());
+        LayoutGrid layoutGrid = new LayoutGrid();
        document.streamAllSubNodes()
-                .filter(node -> (node.getEngines().contains(LayoutEngine.AI) && visualParsingGrid) || (node.getEngines().contains(LayoutEngine.ALGORITHM) && !visualParsingGrid))
+                .peek(layoutGrid::addTreeId)
                .forEach(semanticNode -> {
-                    Color color = switch (semanticNode.getType()) {
-                        case PARAGRAPH -> PARAGRAPH_COLOR;
-                        case TABLE -> TABLE_COLOR;
-                        case SECTION, SUPER_SECTION -> SECTION_COLOR;
-                        case HEADLINE -> HEADLINE_COLOR;
-                        case HEADER, FOOTER -> HEADER_COLOR;
-                        case IMAGE -> IMAGE_COLOR;
-                        default -> null;
-                    };
-
-                    if (semanticNode instanceof DuplicatedParagraph) {
-                        color = DUPLICATE_PARAGRAPH_COLOR;
-                    }
-
-                    if (isNotSectionOrTableCellOrDocument(semanticNode)) {
-                        addAsRectangle(semanticNode, layoutGrid, color);
-                    }
-                    if (semanticNode.getType().equals(NodeType.SECTION) || semanticNode.getType().equals(NodeType.SUPER_SECTION)) {
-                        addSection(semanticNode, layoutGrid, color);
-                    }
-                    if (semanticNode.getType().equals(NodeType.TABLE)) {
-                        Table table = (Table) semanticNode;
-                        addInnerTableLines(table, layoutGrid);
+                    switch (semanticNode.getType()) {
+                        case SECTION, SUPER_SECTION -> layoutGrid.addSection(semanticNode);
+                        case HEADLINE -> layoutGrid.addHeadline((Headline) semanticNode);
+                        case PARAGRAPH -> layoutGrid.addParagraph((Paragraph) semanticNode);
+                        case TABLE -> layoutGrid.addTable((Table) semanticNode);
+                        case IMAGE -> layoutGrid.addImage((Image) semanticNode);
+                        case HEADER, FOOTER -> layoutGrid.addHeaderOrFooter(semanticNode);
                    }
                });
        return layoutGrid;
    }

-
-    private void addInnerTableLines(Table table, LayoutGrid layoutGrid) {
-
-        if (table.getNumberOfCols() < 1 || table.getNumberOfRows() < 1) {
-            return;
-        }
-        for (Page page : table.getPages()) {
-
-            Optional<Integer> optionalFirstRowOnPage = table.streamCol(0)
-                    .filter(tableCell -> tableCell.isOnPage(page.getNumber()))
-                    .map(TableCell::getRow)
-                    .findFirst();
-            if (optionalFirstRowOnPage.isEmpty()) {
-                continue;
-            }
-            int firstRowOnPage = optionalFirstRowOnPage.get();
-
-            Stream<Double> xStream = switch (page.getRotation()) {
-                case 90 -> streamBBoxOfCellsOnPage(table.streamCol(0), page).map(RectangularShape::getMinX);
-                case 180 -> streamBBoxOfCellsOnPage(table.streamRow(firstRowOnPage), page).map(RectangularShape::getMaxX);
-                case 270 -> streamBBoxOfCellsOnPage(table.streamCol(0), page).map(RectangularShape::getMaxX);
-                default -> streamBBoxOfCellsOnPage(table.streamRow(firstRowOnPage), page).map(RectangularShape::getMinX);
-            };
-            List<Double> xs = xStream.collect(Collectors.toList());
-            xs.remove(0);
-
-            Stream<Double> yStream = switch (page.getRotation()) {
-                case 90 -> streamBBoxOfCellsOnPage(table.streamRow(firstRowOnPage), page).map(RectangularShape::getMinY);
-                case 180 -> streamBBoxOfCellsOnPage(table.streamCol(0), page).map(RectangularShape::getMinY);
-                case 270 -> streamBBoxOfCellsOnPage(table.streamRow(firstRowOnPage), page).map(RectangularShape::getMaxY);
-                default -> streamBBoxOfCellsOnPage(table.streamCol(0), page).map(RectangularShape::getMaxY);
-            };
-            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();
-            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));
-            });
-            ys.forEach(y -> {
-                Line2D line = new Line2D.Double(new Point2D.Double(tableBBox.getMinX(), y), new Point2D.Double(tableBBox.getMaxX(), y));
-                coloredLines.add(new ColoredLine(line, INNER_LINES_COLOR, LINE_WIDTH));
-            });
-        }
-    }
-
-
-    private static Stream<Rectangle2D> streamBBoxOfCellsOnPage(Stream<TableCell> table, Page page) {
-
-        return table.filter(tableCell -> tableCell.isOnPage(page.getNumber()))
-                .map(TableCell::getBBox)
-                .map(bBoxMap -> bBoxMap.get(page));
-    }
-
-
-    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 (bBoxMap.values().size() == 1) {
-            handleSinglePage(semanticNode, layoutGrid, color, firstPage, bBoxMap.get(firstPage), treeIdString, maxChildDepth, ownDepth);
-            return;
-        }
-        List<Page> pagesInOrder = bBoxMap.keySet()
-                .stream()
-                .sorted(Comparator.comparingInt(Page::getNumber))
-                .collect(Collectors.toList());
-        pagesInOrder.remove(0);
-        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, 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 + LINE_WIDTH + 2 * maxChildDepth);
-
-        Point2D upperLeftCorner;
-        Point2D translationVector;
-        switch (page.getRotation()) {
-            case 90 -> {
-                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 -> {
-                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 -> {
-
-                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 -> {
-
-                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);
-
-        List<PlacedText> placedTexts = layoutGrid.getVisualizationsPerPages()
-                .get(page.getNumber() - 1).getPlacedTexts();
-
-        PlacedText newText = PlacedText.textFacingUp(s, upperLeftCorner, FONT_SIZE, Color.BLACK, FONT);
-
-        Optional<PlacedText> conflictingText = placedTexts.stream()
-                .filter(pt -> Math.abs(pt.lineStart().getY() - newText.lineStart().getY()) <= FONT_SIZE)
-                .findFirst();
-
-        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 handleSinglePage(SemanticNode semanticNode,
-                                  LayoutGrid layoutGrid,
-                                  Color color,
-                                  Page page,
-                                  Rectangle2D rectangle2D,
-                                  String treeIdString,
-                                  Integer maxChildDepth,
-                                  Integer ownDepth) {
-
-        RectangleAndLinesResult result = createLinesAndPlaceText(semanticNode, page, rectangle2D, treeIdString, layoutGrid, maxChildDepth, ownDepth);
-        // 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 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()
-                .stream()
-                .map(Object::toString)
-                .collect(Collectors.joining("."));
-    }
-
-
-    /*
-    A __________________ B
-     |                  |
-     |                  |
-     |                  |
-     |                  |
-    D|__________________| C
-    The returned List are the lines [AB, BC, DC, AD]
-    The List is reordered, such that the order of the returned lines are always as viewed on the page.
-     */
-    private List<Line2D> createLinesFromRectangle(Rectangle2D r, int pageRotation) {
-        // +0.5 to join the lines
-        List<Line2D> lines = new ArrayList<>(4);
-        float lineWidthCorrection = LINE_WIDTH * 0.5f;
-        Point2D.Float a = new Point2D.Float((float) r.getMinX(), (float) r.getMaxY());
-        Point2D.Float a1 = new Point2D.Float((float) r.getMinX() - lineWidthCorrection, (float) r.getMaxY());
-        Point2D.Float b = new Point2D.Float((float) r.getMaxX(), (float) r.getMaxY());
-        Point2D.Float b1 = new Point2D.Float((float) r.getMaxX() + lineWidthCorrection, (float) r.getMaxY());
-        Point2D.Float c = new Point2D.Float((float) r.getMaxX(), (float) r.getMinY());
-        Point2D.Float c1 = new Point2D.Float((float) r.getMaxX() + lineWidthCorrection, (float) r.getMinY());
-        Point2D.Float d = new Point2D.Float((float) r.getMinX(), (float) r.getMinY());
-        Point2D.Float d1 = new Point2D.Float((float) r.getMinX() - lineWidthCorrection, (float) r.getMinY());
-        lines.add(new Line2D.Float(a1, b1));
-        lines.add(new Line2D.Float(b, c));
-        lines.add(new Line2D.Float(d1, c1));
-        lines.add(new Line2D.Float(a, d));
-
-        return switch (pageRotation) {
-            case 90 -> {
-                Collections.rotate(lines, 1);
-                yield lines;
-            }
-            case 180 -> {
-                Collections.rotate(lines, 2);
-                yield lines;
-            }
-            case 270 -> {
-                Collections.rotate(lines, 3);
-                yield lines;
-            }
-
-            default -> lines;
-        };
-    }
-
-
-    private static boolean isNotSectionOrTableCellOrDocument(SemanticNode semanticNode) {
-
-        return !(semanticNode.getType().equals(NodeType.DOCUMENT)
-                 || semanticNode.getType().equals(NodeType.SECTION)
-                 || semanticNode.getType().equals(NodeType.SUPER_SECTION)
-                 || semanticNode.getType().equals(NodeType.TABLE_CELL));
-    }
-
-
-    private void addAsRectangle(SemanticNode semanticNode, LayoutGrid layoutGrid, Color color) {
-
-        semanticNode.getBBox()
-                .forEach((page, textBBox) -> layoutGrid.getVisualizationsPerPages()
-                        .get(page.getNumber() - 1).getColoredRectangles().add(new ColoredRectangle(textBBox, color, LINE_WIDTH)));
-    }
-
-
-    private Point2D add(Point2D a, Point2D b) {
-
-        return new Point2D.Double(a.getX() + b.getX(), a.getY() + b.getY());
-    }
-
 }
@@ -0,0 +1,34 @@
+package com.knecon.fforesight.service.layoutparser.processor.utils;
+
+import java.awt.geom.Rectangle2D;
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+import java.util.Set;
+import java.util.stream.Collectors;
+
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Page;
+
+import lombok.experimental.UtilityClass;
+
+@UtilityClass
+public class BBoxMergingUtility {
+
+    public Map<Page, Rectangle2D> mergeBBoxes(List<Map<Page, Rectangle2D>> bboxesToMerge) {
+
+        Map<Page, Rectangle2D> bBoxPerPage = new HashMap<>();
+        Set<Page> pages = bboxesToMerge.stream()
+                .flatMap(map -> map.keySet()
+                        .stream())
+                .collect(Collectors.toSet());
+        for (Page page : pages) {
+            Rectangle2D bBoxOnPage = bboxesToMerge.stream()
+                    .filter(childBboxPerPage -> childBboxPerPage.containsKey(page))
+                    .map(childBboxPerPage -> childBboxPerPage.get(page))
+                    .collect(RectangleTransformations.collectBBox());
+            bBoxPerPage.put(page, bBoxOnPage);
+        }
+        return bBoxPerPage;
+    }
+
+}
@@ -96,7 +96,7 @@ public class HeaderFooterDetection {
                        continue;
                    }

-                    int distance = calculateHammingDistanceWithPreprocessing(testString, paddedString);
+                    int distance = StringDistances.hamming(testString, paddedString);
                    double normalizedScore = 1 - (double) distance / Math.max(testString.length(), paddedString.length());
                    score += normalizedScore * (j < weights.length ? weights[j] : 1);
                }
@@ -180,44 +180,4 @@ public class HeaderFooterDetection {
        return headerCandidates;
    }

-
-    /**
-     * Calculate the Hamming distance between two strings after preprocessing to make them the same length
-     * and replacing all digits with a special character '@' since they are a common occurrence in headers/footers.
-     *
-     * @param firstCandidate  First string
-     * @param secondCandidate Second string
-     * @return The Hamming distance between the two preprocessed strings.
-     */
-    private int calculateHammingDistanceWithPreprocessing(String firstCandidate, String secondCandidate) {
-
-        int maxLength = Math.max(firstCandidate.length(), secondCandidate.length());
-
-        String cleanFirstCandidate = padString(firstCandidate, maxLength, '\0').replaceAll("\\d", "@");
-        String cleanSecondCandidate = padString(secondCandidate, maxLength, '\0').replaceAll("\\d", "@");
-
-        int distance = 0;
-        for (int i = 0; i < maxLength; i++) {
-            if (cleanFirstCandidate.charAt(i) != cleanSecondCandidate.charAt(i)) {
-                distance++;
-            }
-        }
-        return distance;
-    }
-
-
-    private String padString(String input, int length, char padChar) {
-
-        if (input.length() >= length) {
-            return input;
-        }
-
-        StringBuilder sb = new StringBuilder(input);
-
-        while (sb.length() < length) {
-            sb.append(padChar);
-        }
-        return sb.toString();
-    }
-
 }
@@ -48,7 +48,7 @@ public class MarkedContentUtils {

        return markedContentByYPosition.values()
                .stream()
-                .map(textPositions -> new TextPositionSequence(textPositions, 0, true).getBBoxInitialUserSpace())
+                .map(textPositions -> new TextPositionSequence(textPositions, 0, true).getBBoxPdf())
                .map(t -> new Rectangle2D.Double(t.getX(), t.getY() - Math.abs(t.getHeight()), t.getWidth(), Math.abs(t.getHeight())))
                .collect(Collectors.toList());
    }
@@ -90,7 +90,7 @@ public class MarkedContentUtils {
                    .map(content -> (TextPosition) content)
                    .filter(content -> !content.getUnicode().equals(" "))
                    .map(textPositions -> new TextPositionSequence(List.of(textPositions), 0, true))
-                    .map(BoundingBox::getBBoxInitialUserSpace)
+                    .map(BoundingBox::getBBoxPdf)
                    .collect(Collectors.toList());
        }

@@ -108,7 +108,7 @@ public final class PositionUtils {
    }


-    public float getHeightDifferenceBetweenChunkWordAndDocumentWord(TextPageBlock textBlock, Float documentMostPopularWordHeight) {
+    public double getHeightDifferenceBetweenChunkWordAndDocumentWord(TextPageBlock textBlock, Double documentMostPopularWordHeight) {

        return textBlock.getMostPopularWordHeight() - documentMostPopularWordHeight;
    }
@@ -116,7 +116,7 @@ public final class PositionUtils {

    public double getApproxLineCount(TextPageBlock textBlock) {

-        return textBlock.getHeight() / textBlock.getMostPopularWordHeight();
+        return textBlock.getBBoxDirAdj().getHeight() / textBlock.getMostPopularWordHeight();
    }

 }
@@ -28,7 +28,7 @@ public class SpreadsheetFinder {
        Map<Point2D, Point2D> edgesV = new HashMap<>();

        for (Cell cell : cells) {
-            for (Point2D pt : getPoints(cell.getBBoxInitialUserSpace())) {
+            for (Point2D pt : getPoints(cell.getBBoxPdf())) {
                if (pointSet.contains(pt)) { // shared vertex, remove it
                    pointSet.remove(pt);
                } else {
@@ -0,0 +1,49 @@
+package com.knecon.fforesight.service.layoutparser.processor.utils;
+
+import lombok.experimental.UtilityClass;
+
+@UtilityClass
+public class StringDistances {
+
+
+    /**
+     * Calculate the Hamming distance between two strings after preprocessing to make them the same length
+     * and replacing all digits with a special character '@' since they are a common occurrence in headers/footers.
+     *
+     * @param s1  First string
+     * @param s2 Second string
+     * @return The Hamming distance between the two preprocessed strings.
+     */
+    public int hamming(String s1, String s2) {
+
+        int maxLength = Math.max(s1.length(), s2.length());
+
+        String cleanFirstCandidate = padString(s1, maxLength, '\0').replaceAll("\\d", "@");
+        String cleanSecondCandidate = padString(s2, maxLength, '\0').replaceAll("\\d", "@");
+
+        int distance = 0;
+        for (int i = 0; i < maxLength; i++) {
+            if (cleanFirstCandidate.charAt(i) != cleanSecondCandidate.charAt(i)) {
+                distance++;
+            }
+        }
+        return distance;
+    }
+
+
+    private String padString(String input, int length, char padChar) {
+
+        if (input.length() >= length) {
+            return input;
+        }
+
+        StringBuilder sb = new StringBuilder(input);
+
+        while (sb.length() < length) {
+            sb.append(padChar);
+        }
+        return sb.toString();
+    }
+
+
+}
@@ -1,30 +1,143 @@
 package com.knecon.fforesight.service.layoutparser.processor.utils;

+import java.awt.geom.Rectangle2D;
+import java.awt.geom.RectangularShape;
+import java.util.Collection;
+import java.util.Comparator;
+import java.util.HashSet;
 import java.util.List;
+import java.util.Set;
 import java.util.stream.Collectors;

+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.AngleFilter;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.TextBoundingBox;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.UnionFind;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.utils.DoubleUtils;
+import com.knecon.fforesight.service.layoutparser.processor.docstrum.utils.FastAtan2;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;

+import lombok.experimental.UtilityClass;
+
+@UtilityClass
 public class TextPositionOperations {

-    private static final TextPositionSequenceComparator comparator = new TextPositionSequenceComparator();
+    public static final double ANGLE_TOLERANCE = Math.PI / 35;
+    public static final AngleFilter ANGLE_FILTER = new AngleFilter(-ANGLE_TOLERANCE, ANGLE_TOLERANCE);
+    public static final double MAX_LINE_HEIGHT_FACTOR = 0.66; // multiplied with max word height
+    public static final double MAX_WORD_DISTANCE_FACTOR = 3.5; // multiplied with max word width
+
+    private static final double Y_THRESHOLD = 5;
+    private static final double X_THRESHOLD = 3;
+    private static final Comparator<TextBoundingBox> COMPARATOR_DIR_ADJ = //
+            Comparator.comparing(TextBoundingBox::getDir)
+                    .thenComparing(TextBoundingBox::getMaxYDirAdj, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, Y_THRESHOLD))
+                    .thenComparing(TextBoundingBox::getXDirAdj, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, X_THRESHOLD));


-    public static List<TextPositionSequence> mergeAndSortTextPositionSequenceByYThenX(List<TextPageBlock> textBlocks) {
+    public List<TextPositionSequence> mergeAndSort(List<TextPageBlock> textBlocks) {

-        var sequence = textBlocks.stream().flatMap(tb -> tb.getSequences().stream()).collect(Collectors.toList());
-
-        // because the TextPositionSequenceComparator is not transitive, but
-        // JDK7+ enforces transitivity on comparators, we need to use
-        // a custom quicksort implementation (which is slower, unfortunately).
-        QuickSort.sort(sequence, comparator);
-        return sequence;
+        var sequences = textBlocks.stream()
+                .flatMap(tb -> tb.getSequences()
+                        .stream())
+                .collect(Collectors.toSet());
+        return sortUsingLineDetection(sequences);
    }

-    public static List<TextPositionSequence> mergeTextPositionSequence(List<TextPageBlock> textBlocks) {

-       return textBlocks.stream().flatMap(tb -> tb.getSequences().stream()).collect(Collectors.toList());
+    public List<TextPositionSequence> sort(List<TextPositionSequence> sequences) {
+
+        return sortUsingLineDetection(new HashSet<>(sequences));
+    }
+
+
+    private List<TextPositionSequence> sortUsingLineDetection(Set<TextPositionSequence> sequences) {
+
+        return sortLines(groupByLine(sequences));
+
+    }
+
+
+    public List<TextPositionSequence> sortLines(Collection<Set<TextPositionSequence>> lines) {
+
+        return lines.stream()
+                .map(TextPositionOperations::sortByXDirAdj)
+                .filter(line -> !line.isEmpty())
+                .sorted(Comparator.comparing(line -> line.get(0), COMPARATOR_DIR_ADJ))
+                .flatMap(Collection::stream)
+                .toList();
+    }
+
+
+    private List<TextPositionSequence> sortByXDirAdj(Set<TextPositionSequence> line) {
+
+        return line.stream()
+                .sorted(Comparator.comparing(TextPositionSequence::getXDirAdj))
+                .toList();
+    }
+
+
+    public Collection<Set<TextPositionSequence>> groupByLine(Set<TextPositionSequence> sequences) {
+
+        double maxLineDistance = sequences.stream()
+                                         .map(TextPositionSequence::getBBoxDirAdj)
+                                         .mapToDouble(RectangularShape::getHeight).average()
+                                         .orElse(10) * MAX_LINE_HEIGHT_FACTOR;
+        double maxXGap = sequences.stream()
+                                 .map(TextPositionSequence::getBBoxDirAdj)
+                                 .mapToDouble(RectangularShape::getWidth).average()
+                                 .orElse(75) * MAX_WORD_DISTANCE_FACTOR;
+
+        UnionFind<TextPositionSequence> unionFind = new UnionFind<>(sequences);
+
+        for (TextPositionSequence sequence : sequences) {
+            for (TextPositionSequence sequence2 : sequences) {
+
+                if (sequence.equals(sequence2) || unionFind.inSameSet(sequence, sequence2)) {
+                    continue;
+                }
+
+                double angle = computeAngle(sequence.getBBoxDirAdj(), sequence2.getBBoxDirAdj());
+
+                double normalizedVerticalDistance = Math.abs(sequence.getBBoxDirAdj().getCenterY() - sequence2.getBBoxDirAdj().getCenterY()) / maxLineDistance;
+                double normalizedHorizontalDistance = Math.abs(sequence.getBBoxDirAdj().getCenterX() - sequence2.getBBoxDirAdj().getCenterX()) / maxXGap;
+
+                if (sequence.getDir() != sequence2.getDir()
+                    || Math.abs(sequence.getFontSize() - sequence2.getFontSize()) > 0.5 * Math.min(sequence.getFontSize(),
+                                                                                                   sequence2.getFontSize())
+                    || Math.pow(normalizedVerticalDistance, 2) + Math.pow(normalizedHorizontalDistance, 2) > 1
+                    || !ANGLE_FILTER.matches(angle)) {
+                    continue;
+                }
+
+                unionFind.union(sequence, sequence2);
+            }
+        }
+
+        return unionFind.getGroups();
+    }
+
+
+    public double computeAngle(Rectangle2D rect1, Rectangle2D rect2) {
+
+        double rect1CentroidX = rect1.getCenterX();
+        double rect1CentroidY = rect1.getCenterY();
+        double rect2CentroidX = rect2.getCenterX();
+        double rect2CentroidY = rect2.getCenterY();
+
+        double deltaX = rect2CentroidX - rect1CentroidX;
+        double deltaY = rect2CentroidY - rect1CentroidY;
+
+        return FastAtan2.fastAtan2(deltaY, deltaX);
+    }
+
+
+    public List<TextPositionSequence> merge(List<TextPageBlock> textBlocks) {
+
+        return textBlocks.stream()
+                .map(TextPageBlock::getSequences)
+                .flatMap(Collection::stream)
+                .collect(Collectors.toList());
    }

 }
@@ -1,64 +0,0 @@
-/*
- * Licensed to the Apache Software Foundation (ASF) under one or more
- * contributor license agreements.  See the NOTICE file distributed with
- * this work for additional information regarding copyright ownership.
- * The ASF licenses this file to You under the Apache License, Version 2.0
- * (the "License"); you may not use this file except in compliance with
- * the License.  You may obtain a copy of the License at
- *
- *      http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-package com.knecon.fforesight.service.layoutparser.processor.utils;
-
-import java.util.Comparator;
-
-import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
-
-/**
- * This class is a comparator for TextPosition operators.  It handles
- * pages with text in different directions by grouping the text based
- * on direction and sorting in that direction. This allows continuous text
- * in a given direction to be more easily grouped together.
- *
- * @author Ben Litchfield
- */
-public class TextPositionSequenceComparator implements Comparator<TextPositionSequence> {
-
-    @Override
-    public int compare(TextPositionSequence pos1, TextPositionSequence pos2) {
-        // only compare text that is in the same direction
-        int cmp1 = Float.compare(pos1.getDir().getDegrees(), pos2.getDir().getDegrees());
-        if (cmp1 != 0) {
-            return cmp1;
-        }
-
-        // get the text direction adjusted coordinates
-        double x1 = pos1.getBBox().getX();
-        double x2 = pos2.getBBox().getX();
-
-        double pos1YBottom = pos1.getBBox().getMaxY();
-        double pos2YBottom = pos2.getBBox().getMaxY();
-
-        // note that the coordinates have been adjusted so 0,0 is in upper left
-        double pos1YTop = pos1YBottom - pos1.getBBox().getHeight();
-        double pos2YTop = pos2YBottom - pos2.getBBox().getHeight();
-
-        double yDifference = Math.abs(pos1YBottom - pos2YBottom);
-
-        // we will do a simple tolerance comparison
-        if (yDifference < .1 || pos2YBottom >= pos1YTop && pos2YBottom <= pos1YBottom || pos1YBottom >= pos2YTop && pos1YBottom <= pos2YBottom) {
-            return Double.compare(x1, x2);
-        } else if (pos1YBottom < pos2YBottom) {
-            return -1;
-        } else {
-            return 1;
-        }
-    }
-
-}
@@ -0,0 +1,111 @@
+package com.knecon.fforesight.service.layoutparser.processor.visualization;
+
+import java.awt.geom.AffineTransform;
+import java.awt.geom.Line2D;
+import java.awt.geom.Point2D;
+import java.awt.geom.Rectangle2D;
+
+import lombok.experimental.UtilityClass;
+
+@UtilityClass
+public class ConnectionLineUtil {
+
+    public static Line2D[] splitRectangleIntoLines(Rectangle2D rect) {
+
+        double x = rect.getX();
+        double y = rect.getY();
+        double width = rect.getWidth();
+        double height = rect.getHeight();
+
+        Line2D[] lines = new Line2D[4];
+        lines[0] = new Line2D.Double(x, y, x + width, y);           // Top
+        lines[1] = new Line2D.Double(x + width, y, x + width, y + height); // Right
+        lines[2] = new Line2D.Double(x + width, y + height, x, y + height); // Bottom
+        lines[3] = new Line2D.Double(x, y + height, x, y);          // Left
+
+        return lines;
+    }
+
+
+    public static Line2D transform(Line2D line2D, AffineTransform affineTransform) {
+
+        var p1 = affineTransform.transform(line2D.getP1(), null);
+        var p2 = affineTransform.transform(line2D.getP2(), null);
+        return new Line2D.Double(p1, p2);
+    }
+
+
+    public static double length(Line2D line2D) {
+
+        return line2D.getP1().distance(line2D.getP2());
+    }
+
+
+    public static Line2D findClosestMidpointLine(Rectangle2D rect1, Rectangle2D rect2) {
+
+        Line2D[] lines1 = splitRectangleIntoLines(rect1);
+        Line2D[] lines2 = splitRectangleIntoLines(rect2);
+
+        Line2D closestLine1 = null;
+        Line2D closestLine2 = null;
+        double minDistance = Double.MAX_VALUE;
+
+        for (Line2D line1 : lines1) {
+            for (Line2D line2 : lines2) {
+                double distance = lineDistance(line1, line2);
+                if (distance < minDistance) {
+                    minDistance = distance;
+                    closestLine1 = line1;
+                    closestLine2 = line2;
+                }
+            }
+        }
+
+        if (closestLine1 == null || closestLine2 == null) {
+            throw new IllegalStateException("Could not find closest lines");
+        }
+
+        Point2D midpoint1 = getMidpoint(closestLine1);
+        Point2D midpoint2 = getMidpoint(closestLine2);
+
+        return new Line2D.Double(midpoint1, midpoint2);
+    }
+
+
+    private static double lineDistance(Line2D line1, Line2D line2) {
+
+        return Math.abs(getMidpoint(line1).distance(getMidpoint(line2)));
+    }
+
+
+    private static Point2D getMidpoint(Line2D line) {
+
+        double x = (line.getX1() + line.getX2()) / 2;
+        double y = (line.getY1() + line.getY2()) / 2;
+        return new Point2D.Double(x, y);
+    }
+
+
+    public static Line2D[] createArrowHead(Line2D line, double arrowLength) {
+
+        Point2D start = line.getP1();
+        Point2D end = line.getP2();
+
+        // Calculate the angle of the line
+        double angle = Math.atan2(end.getY() - start.getY(), end.getX() - start.getX());
+
+        // Calculate the points for the two arrow lines
+        double arrowHeadAngle = Math.PI / 6;
+        double x1 = end.getX() - arrowLength * Math.cos(angle - arrowHeadAngle);
+        double y1 = end.getY() - arrowLength * Math.sin(angle - arrowHeadAngle);
+        double x2 = end.getX() - arrowLength * Math.cos(angle + arrowHeadAngle);
+        double y2 = end.getY() - arrowLength * Math.sin(angle + arrowHeadAngle);
+
+        // Create and return the two arrow lines
+        Line2D arrow1 = new Line2D.Double(end, new Point2D.Double(x1, y1));
+        Line2D arrow2 = new Line2D.Double(end, new Point2D.Double(x2, y2));
+
+        return new Line2D[]{arrow1, arrow2};
+    }
+
+}
@@ -6,8 +6,8 @@ import java.awt.geom.Point2D;
 import java.awt.geom.Rectangle2D;
 import java.util.Collection;
 import java.util.List;
+import java.util.Set;
 import java.util.concurrent.atomic.AtomicInteger;
-import java.util.stream.Stream;

 import org.apache.pdfbox.pdmodel.documentinterchange.markedcontent.PDMarkedContent;

@@ -21,12 +21,10 @@ import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageB
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
 import com.knecon.fforesight.service.layoutparser.processor.utils.MarkedContentUtils;
 import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
-import com.knecon.fforesight.service.viewerdoc.ContentStreams;
+import com.knecon.fforesight.service.viewerdoc.layers.LayoutDebugLayerConfig;
 import com.knecon.fforesight.service.viewerdoc.model.ColoredLine;
 import com.knecon.fforesight.service.viewerdoc.model.ColoredRectangle;
 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.model.VisualizationsOnPage;

 import lombok.AccessLevel;
@@ -36,72 +34,15 @@ import lombok.NoArgsConstructor;
 import lombok.Setter;
 import lombok.experimental.FieldDefaults;

+@Setter
@Getter
@NoArgsConstructor
@AllArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE)
-public class LayoutparsingVisualizations {
+public class LayoutDebugLayer extends LayoutDebugLayerConfig {

-    static final Standard14EmbeddableFont FONT = Standard14EmbeddableFont.helvetica();
-
-    static final Color WORDS_COLOR = new Color(68, 84, 147);
-    static final Color LINES_COLOR = new Color(152, 45, 179);
-    static final Color ZONES_COLOR = new Color(131, 38, 38);
-
-    static final Color RULINGS_COLOR = new Color(21, 221, 174);
-    static final Color TABLE_RULINGS_COLOR = new Color(255, 175, 175);
-    static final Color HEADER_RULING_COLOR = new Color(171, 131, 6);
-    static final Color FOOTER_RULING_COLOR = new Color(106, 82, 2);
-    static final Color UNDERLINE_RULING_COLOR = new Color(6, 39, 171);
-    static final Color STRIKETROUGH_RULING_COLOR = new Color(171, 6, 6);
-
-    static final Color CELLS_COLOR = new Color(31, 214, 27);
-
-    static final Color MAIN_BODY_COLOR = new Color(171, 131, 6);
-    static final Color MARKED_CONTENT_COLOR = new Color(171, 131, 6);
-
-    static final List<Color> ROTATING_CHARACTER_COLOR = List.of(new Color(255, 87, 51),
-                                                                new Color(255, 195, 0),
-                                                                new Color(76, 175, 80),
-                                                                new Color(33, 150, 243),
-                                                                new Color(155, 89, 182),
-                                                                new Color(233, 30, 99),
-                                                                new Color(0, 188, 212),
-                                                                new Color(121, 85, 72));
-
-    @Setter
    boolean active;

-    final Visualizations words = Visualizations.builder().layer(ContentStreams.WORDS).build();
-    final Visualizations lines = Visualizations.builder().layer(ContentStreams.LINES).build();
-    final Visualizations zones = Visualizations.builder().layer(ContentStreams.ZONES).build();
-    final Visualizations mainBody = Visualizations.builder().layer(ContentStreams.MAIN_BODY).build();
-    final Visualizations clean_rulings = Visualizations.builder().layer(ContentStreams.CLEAN_RULINGS).build();
-    final Visualizations rulings = Visualizations.builder().layer(ContentStreams.RULINGS).build();
-    final Visualizations cells = Visualizations.builder().layer(ContentStreams.CELLS).build();
-    final Visualizations markedContent = Visualizations.builder().layer(ContentStreams.MARKED_CONTENT).build();
-    final Visualizations neighbours = Visualizations.builder().layer(ContentStreams.NEIGHBOURS).build();
-    final Visualizations characters = Visualizations.builder().layer(ContentStreams.CHARACTERS).build();
-
-
-    public Stream<Visualizations> streamAll() {
-
-        if (!active) {
-            return Stream.empty();
-        }
-        return Stream.of(characters, //
-                         neighbours,//
-                         words, //
-                         lines, //
-                         zones, //
-                         rulings, //
-                         clean_rulings, //
-                         cells, //
-                         mainBody, //
-                         markedContent //
-        );
-    }
-

    public void addTextVisualizations(List<TextPositionSequence> textPositionSequences, int pageNumber) {

@@ -111,7 +52,7 @@ public class LayoutparsingVisualizations {
        VisualizationsOnPage visualizationsOnPage = getOrCreateVisualizationsOnPage(pageNumber, this.words);
        visualizationsOnPage.getColoredRectangles()
                .addAll(textPositionSequences.stream()
-                                .map(BoundingBox::getBBoxInitialUserSpace)
+                                .map(BoundingBox::getBBoxPdf)
                                .map(rect -> new ColoredRectangle(rect, WORDS_COLOR, 1))
                                .toList());
    }
@@ -130,6 +71,7 @@ public class LayoutparsingVisualizations {
                                .toList());
    }

+
    public void addRulingVisualization(List<Ruling> rulings, int pageNumber) {

        if (!active) {
@@ -137,8 +79,7 @@ public class LayoutparsingVisualizations {
        }
        VisualizationsOnPage visualizationsOnPage = getOrCreateVisualizationsOnPage(pageNumber, this.rulings);
        visualizationsOnPage.getColoredLines()
-                .addAll(rulings
-                                .stream()
+                .addAll(rulings.stream()
                                .map(ruling -> new ColoredLine(ruling, decideOnRulingColor(ruling), 0.5f))
                                .toList());
    }
@@ -165,7 +106,7 @@ public class LayoutparsingVisualizations {
        VisualizationsOnPage visualizationsOnPage = getOrCreateVisualizationsOnPage(pageNumber, this.cells);
        visualizationsOnPage.getColoredRectangles()
                .addAll(cells.stream()
-                                .map(cell -> new ColoredRectangle(cell.getBBoxInitialUserSpace(), CELLS_COLOR, 1))
+                                .map(cell -> new ColoredRectangle(cell.getBBoxPdf(), CELLS_COLOR, 1))
                                .toList());
    }

@@ -179,7 +120,7 @@ public class LayoutparsingVisualizations {
        VisualizationsOnPage visualizationsOnPage = getOrCreateVisualizationsOnPage(page, this.zones);
        visualizationsOnPage.getColoredRectangles()
                .addAll(zones.stream()
-                                .map(BoundingBox::getBBoxInitialUserSpace)
+                                .map(BoundingBox::getBBoxPdf)
                                .map(zone -> new ColoredRectangle(zone, ZONES_COLOR, 1))
                                .toList());

@@ -204,12 +145,29 @@ public class LayoutparsingVisualizations {
        VisualizationsOnPage visualizationsOnPage = getOrCreateVisualizationsOnPage(page, this.lines);
        visualizationsOnPage.getColoredRectangles()
                .addAll(lines.stream()
-                                .map(BoundingBox::getBBoxInitialUserSpace)
+                                .map(BoundingBox::getBBoxPdf)
                                .map(line -> new ColoredRectangle(line, LINES_COLOR, 0.5f))
                                .toList());
    }


+
+    public void addLineVisualizationsFromNestedTextPosition(Collection<Set<TextPositionSequence>> lines, int pageNumber) {
+
+        if (!active) {
+            return;
+        }
+        VisualizationsOnPage visualizationsOnPage = getOrCreateVisualizationsOnPage(pageNumber, this.lines);
+        visualizationsOnPage.getColoredRectangles()
+                .addAll(lines.stream()
+                                .map(line -> line.stream()
+                                        .map(BoundingBox::getBBoxPdf)
+                                        .collect(RectangleTransformations.collectBBox()))
+                                .map(line -> new ColoredRectangle(line, LINES_COLOR, 0.5f))
+                                .toList());
+
+    }
+
    public void addTextBlockVisualizations(List<TextPageBlock> textPageBlocks, int page) {

        if (!active) {
@@ -218,7 +176,7 @@ public class LayoutparsingVisualizations {
        VisualizationsOnPage visualizationsOnPage = getOrCreateVisualizationsOnPage(page, zones);
        visualizationsOnPage.getColoredRectangles()
                .addAll(textPageBlocks.stream()
-                                .map(rect -> new ColoredRectangle(rect.getBBoxInitialUserSpace(), ZONES_COLOR, 1))
+                                .map(rect -> new ColoredRectangle(rect.getBBoxPdf(), ZONES_COLOR, 1))
                                .toList());
    }

@@ -282,11 +240,11 @@ public class LayoutparsingVisualizations {
                .flatMap(Collection::stream)
                .forEach(character -> {
                    Color color = ROTATING_CHARACTER_COLOR.get(index.getAndIncrement() % ROTATING_CHARACTER_COLOR.size());
-                    Rectangle2D charBBox = character.getTextPosition().getBBoxInitialUserSpace();
+                    Rectangle2D charBBox = character.getTextPosition().getBBoxPdf();
                    characterVisualizations.getColoredRectangles().add(new ColoredRectangle(charBBox, color, 1));
                    character.getNeighbors()
                            .forEach(neighbor -> {
-                                Rectangle2D neighborBBox = neighbor.getCharacter().getTextPosition().getBBoxInitialUserSpace();
+                                Rectangle2D neighborBBox = neighbor.getCharacter().getTextPosition().getBBoxPdf();
                                Line2D line = new Line2D.Double(new Point2D.Double(charBBox.getCenterX(), charBBox.getCenterY()),
                                                                new Point2D.Double(neighborBBox.getCenterX(), neighborBBox.getCenterY()));
                                neighbourVisualizations.getColoredLines().add(new ColoredLine(line, color, 1));
@@ -296,15 +254,4 @@ public class LayoutparsingVisualizations {
    }


-    private VisualizationsOnPage getOrCreateVisualizationsOnPage(int page, Visualizations visualizations) {
-
-        if (visualizations.getVisualizationsOnPages().containsKey(page - 1)) {
-            return visualizations.getVisualizationsOnPages()
-                    .get(page - 1);
-        }
-        VisualizationsOnPage visualizationsOnPage = VisualizationsOnPage.builder().build();
-        visualizations.getVisualizationsOnPages().put(page - 1, visualizationsOnPage);
-        return visualizationsOnPage;
-    }
-
 }
--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
Andrei Isvoran	b562ffdf10	RED-9942 - Don't skip images on file	2024-08-23 13:03:25 +03:00
Dominique Eifländer	3eb97d614f	Merge branch 'RED-9760-NPE' into 'main' RED-9760: Fixed nullpointer in TextPageBlock See merge request fforesight/layout-parser!194	2024-08-13 13:24:48 +02:00
Dominique Eifländer	81469413b0	RED-9760: Fixed nullpointer in TextPageBlock	2024-08-13 13:18:50 +02:00
Dominique Eifländer	2993676a6f	Merge branch 'RED-9670' into 'main' RED-9760: change compareDouble to something sensible See merge request fforesight/layout-parser!193	2024-08-12 16:02:51 +02:00
Kilian Schüttler	8e115dcd8a	RED-9760: change compareDouble to something sensible	2024-08-12 16:02:50 +02:00
Dominique Eifländer	173911b840	Merge branch 'hotfix-reading-order' into 'main' hotfix: threshold adjustements See merge request fforesight/layout-parser!192	2024-08-12 14:59:20 +02:00
Kilian Schuettler	b0ae00aa02	hotfix: threshold adjustements	2024-08-12 14:52:18 +02:00
Dominique Eifländer	00bf9f279e	Merge branch 'hotfix-reading-order' into 'main' hotfix: use center coordinates See merge request fforesight/layout-parser!191	2024-08-09 15:51:42 +02:00
Kilian Schuettler	d16377a24a	hotfix: line comparison with center coordinates	2024-08-09 15:45:23 +02:00
Dominique Eifländer	81179ee744	Merge branch 'RED-9760-dcold' into 'main' RED-9760: Changed lineSeparation threshold for documine old See merge request fforesight/layout-parser!190	2024-08-09 14:50:39 +02:00
Dominique Eifländer	1953b5924f	RED-9760: Changed lineSeparation threshold for documine old	2024-08-09 14:42:14 +02:00
Kilian Schüttler	6f6e8d5d4e	Merge branch 'hotfix-reading-order' into 'main' hotfix reading order See merge request fforesight/layout-parser!187	2024-08-09 11:49:12 +02:00
Kilian Schüttler	69bcd4f68d	hotfix reading order	2024-08-09 11:49:12 +02:00
Timo Bejan	b900cfaf31	Merge branch 'CLARI-140' into 'main' CLARI-140 - case issue See merge request fforesight/layout-parser!189	2024-08-08 21:49:40 +02:00
Timo Bejan	cdc2081785	CLARI-140 - case issue	2024-08-08 22:40:11 +03:00
Timo Bejan	a9287ec406	Merge branch 'CLARI-139' into 'main' CLAR-139 - fixed outline error for unparsable object See merge request fforesight/layout-parser!188	2024-08-08 16:36:41 +02:00
Timo Bejan	5b6a706c28	CLAR-139 - fixed outline error for unparsable object	2024-08-08 16:20:14 +03:00
Timo Bejan	28d8ad0a3f	Merge branch 'CLARI-128' into 'main' Fixed Index out of bounds exception in blockificationpostprocessingservice -... See merge request fforesight/layout-parser!186	2024-07-30 16:56:21 +02:00
Timo Bejan	0c1583c1be	Fixed Index out of bounds exception in blockificationpostprocessingservice - this could should be documented btw, there are also probably other use-cases where the code doesnt work	2024-07-30 17:45:05 +03:00
Andrei Isvoran	7633566d9b	Merge branch 'RED-9607-fp' into 'main' RED-9607 - Correctly determine text position sequence based on file rotation See merge request fforesight/layout-parser!184	2024-07-25 14:11:00 +02:00
Andrei Isvoran	cc4f09711e	RED-9607 - Correctly determine text position sequence based on file rotation	2024-07-24 16:35:11 +03:00
Kilian Schüttler	370165dc59	Merge branch 'document-data-markdown' into 'main' CLARI: document-data-markdown See merge request fforesight/layout-parser!181	2024-07-18 17:19:44 +02:00
Maverick Studer	8c052c38d7	CLARI: document-data-markdown	2024-07-18 17:19:43 +02:00
Kilian Schüttler	ea18d3d307	Merge branch 'RED-8800' into 'main' RED-8800: adjust coordinates in BE to ignore cropbox See merge request fforesight/layout-parser!179	2024-07-15 17:45:13 +02:00
Kilian Schüttler	2726fc3fe1	RED-8800: adjust coordinates in BE to ignore cropbox	2024-07-15 17:45:13 +02:00
Kilian Schüttler	033279e261	Merge branch 'RED-9353' into 'main' RED-9353: refactor PDFTronViewerDocumentService See merge request fforesight/layout-parser!178	2024-07-15 12:54:17 +02:00
Kilian Schüttler	ec0dd032c9	RED-9353: refactor PDFTronViewerDocumentService	2024-07-15 12:54:17 +02:00
Andrei Isvoran	598fa7f1c7	Merge branch 'RED-9496-shutdown-fp' into 'main' RED-9496 - Implement graceful shutdown See merge request fforesight/layout-parser!176	2024-07-04 14:26:00 +02:00
Andrei Isvoran	65b1f7d179	RED-9496 - Implement graceful shutdown	2024-07-04 14:21:20 +03:00
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
yhampe	9ecf9ca19f	RED-3813: Recategorize same image as experimental feature now writing hash into structure	2024-06-05 14:20:33 +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