Merge branch 'RM-231-bp' into 'release/0.192.x'

RM-231: missing whitespace in name See merge request fforesight/layout-parser!266
RM-231: missing whitespace in name
2025-01-14 13:07:49 +01:00 · 2025-01-14 13:01:09 +01:00 · 2025-01-13 13:22:59 +01:00 · 2024-11-13 10:54:39 +01:00 · 2024-11-13 10:54:39 +01:00 · 2024-11-04 12:51:38 +01:00
52 changed files with 701 additions and 609 deletions
@@ -26,10 +26,11 @@ 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.docstrum.model.Character;
+import com.knecon.fforesight.service.layoutparser.processor.model.table.Ruling;
 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.services.classification.ClassificationService;
-import com.knecon.fforesight.service.layoutparser.processor.services.classification.TableOfContentsClassificationService;
 import com.knecon.fforesight.service.layoutparser.processor.services.mapper.MarkdownMapper;
 import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationDocument;
@@ -45,7 +46,7 @@ import com.knecon.fforesight.service.layoutparser.processor.model.outline.TableO
 import com.knecon.fforesight.service.layoutparser.processor.model.table.Cell;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRulings;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
-import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.adapter.CvTableParsingAdapter;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.adapter.ImageServiceResponseAdapter;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.adapter.VisualLayoutParsingAdapter;
@@ -53,7 +54,6 @@ import com.knecon.fforesight.service.layoutparser.processor.python_api.model.ima
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.TableCells;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.TableServiceResponse;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.VisualLayoutParsingResponse;
-import com.knecon.fforesight.service.layoutparser.processor.services.BodyTextFrameService;
 import com.knecon.fforesight.service.layoutparser.processor.services.RulingCleaningService;
 import com.knecon.fforesight.service.layoutparser.processor.services.SectionsBuilderService;
 import com.knecon.fforesight.service.layoutparser.processor.services.SimplifiedSectionTextService;
@@ -63,9 +63,6 @@ import com.knecon.fforesight.service.layoutparser.processor.services.blockificat
 import com.knecon.fforesight.service.layoutparser.processor.services.blockification.DocstrumBlockificationService;
 import com.knecon.fforesight.service.layoutparser.processor.services.blockification.DocuMineBlockificationService;
 import com.knecon.fforesight.service.layoutparser.processor.services.blockification.RedactManagerBlockificationService;
-import com.knecon.fforesight.service.layoutparser.processor.services.classification.ClarifyndClassificationService;
-import com.knecon.fforesight.service.layoutparser.processor.services.classification.DocuMineClassificationService;
-import com.knecon.fforesight.service.layoutparser.processor.services.classification.RedactManagerClassificationService;
 import com.knecon.fforesight.service.layoutparser.processor.services.factory.DocumentGraphFactory;
 import com.knecon.fforesight.service.layoutparser.processor.services.graphics.Box;
 import com.knecon.fforesight.service.layoutparser.processor.services.graphics.GraphicExtractorService;
@@ -163,9 +160,9 @@ public class LayoutParsingPipeline {
        }

        if (!viewerDocumentFile.equals(originFile)) {
-            viewerDocumentFile.delete();
+            assert !viewerDocumentFile.exists() || viewerDocumentFile.delete();
        }
-        originFile.delete();
+        assert !originFile.exists() || originFile.delete();

        return LayoutParsingFinishedEvent.builder()
                .identifier(layoutParsingRequest.identifier())
@@ -245,12 +242,8 @@ public class LayoutParsingPipeline {
        }

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

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

        long pageCount = originDocument.getNumberOfPages();

@@ -275,25 +268,22 @@ public class LayoutParsingPipeline {
            stripper.setEndPage(pageNumber);
            stripper.setPdpage(pdPage);
            stripper.getText(originDocument);
-            List<TextPositionSequence> words = stripper.getTextPositionSequences();
+            List<Word> words = stripper.getWords();

 //            rotateDirAdjExactly(words, pdPage); // works really well for many highly rotated documents (e.g. VV-331340.pdf), but it decreases the headline performance by 1.3%, so I am leaving it out for now

            if (layoutParsingType.equals(LayoutParsingType.DOCUMINE_OLD)) {
                var lines = TextPositionOperations.groupByLine(new HashSet<>(words));
                classificationDocument.getLayoutDebugLayer().addLineVisualizationsFromNestedTextPosition(lines, pageNumber);
-                words = TextPositionOperations.sortLines(lines);
+                words = TextPositionOperations.sortWords(lines);
            }
            classificationDocument.getLayoutDebugLayer().addTextVisualizations(words, pageNumber);

            PDRectangle pdr = pdPage.getMediaBox();

-            int rotation = pdPage.getRotation();
-            boolean isLandscape = pdr.getWidth() > pdr.getHeight() && (rotation == 0 || rotation == 180) || pdr.getHeight() > pdr.getWidth() && (rotation == 90 || rotation == 270);
-
-            PDRectangle cropbox = pdPage.getCropBox();
-            classificationDocument.getLayoutDebugLayer().addRulingVisualization(stripper.getRulings(), pageNumber);
-            CleanRulings cleanRulings = rulingCleaningService.deduplicateAndStraightenRulings(pdfTableCells.get(pageNumber), stripper.getRulings());
+            List<Ruling> rulings = stripper.getRulings();
+            classificationDocument.getLayoutDebugLayer().addRulingVisualization(rulings, pageNumber);
+            CleanRulings cleanRulings = rulingCleaningService.deduplicateAndStraightenRulings(pdfTableCells.get(pageNumber), rulings);

            PageInformation pageInformation = PageInformation.fromPDPage(pageNumber, pdPage);
            List<Cell> emptyTableCells = TableExtractionService.findCells(cleanRulings.getHorizontals(), cleanRulings.getVerticals(), pageInformation);
@@ -301,7 +291,7 @@ public class LayoutParsingPipeline {

            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.getWords(), false);

            pdfImages.computeIfAbsent(pageNumber, x -> new ArrayList<>())
                    .addAll(graphics.stream()
@@ -313,8 +303,7 @@ public class LayoutParsingPipeline {
                                    .toList());

            ClassificationPage classificationPage = switch (layoutParsingType) {
-                case REDACT_MANAGER_OLD ->
-                        redactManagerBlockificationService.blockify(stripper.getTextPositionSequences(), cleanRulings, classificationDocument.getLayoutDebugLayer());
+                case REDACT_MANAGER_OLD -> redactManagerBlockificationService.blockify(stripper.getWords(), cleanRulings, classificationDocument.getLayoutDebugLayer());
                case DOCUMINE_OLD -> docuMineBlockificationService.blockify(words, cleanRulings);
                case DOCUMINE, REDACT_MANAGER, REDACT_MANAGER_PARAGRAPH_DEBUG, REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH ->
                        docstrumBlockificationService.blockify(words, cleanRulings, true, classificationDocument.getLayoutDebugLayer(), layoutParsingType);
@@ -322,27 +311,9 @@ public class LayoutParsingPipeline {
                        docstrumBlockificationService.blockify(words, cleanRulings, false, classificationDocument.getLayoutDebugLayer(), layoutParsingType);
            };

-            classificationPage.setCleanRulings(cleanRulings);
-            classificationPage.setRotation(rotation);
-            classificationPage.setLandscape(isLandscape);
-            classificationPage.setPageNumber(pageNumber);
-            classificationPage.setPageWidth(cropbox.getWidth());
-            classificationPage.setPageHeight(cropbox.getHeight());
+            updateClassificationPage(pdPage, pdr, classificationPage, cleanRulings, pageNumber, pageInformation);

-            if (layoutParsingType != LayoutParsingType.REDACT_MANAGER_OLD) {
-                List<OutlineObject> outlineObjects = classificationDocument.getOutlineObjectTree().getOutlineObjectsPerPage().getOrDefault(pageNumber, new ArrayList<>());
-
-                OutlineObject notFoundOutlineObject = null;
-                if (lastProcessedOutlineObject != null && !lastProcessedOutlineObject.isFound()) {
-                    lastProcessedOutlineObject.resetPoint();
-                    notFoundOutlineObject = lastProcessedOutlineObject;
-                }
-                if (!outlineObjects.isEmpty()) {
-                    classificationPage.setOutlineObjects(outlineObjects);
-                    lastProcessedOutlineObject = blockificationPostprocessingService.sanitizeOutlineBlocks(classificationPage, notFoundOutlineObject);
-                }
-                classificationDocument.getLayoutDebugLayer().addOutlineObjects(outlineObjects, pageInformation);
-            }
+            blockificationPostprocessingService.findHeadlinesFromOutline(classificationDocument, pageNumber, classificationPage, pageInformation);

            classificationDocument.getLayoutDebugLayer().addMarkedContentVisualizations(stripper.getMarkedContents(), pageNumber);
            // MarkedContent needs to be converted at this point, otherwise it leads to GC Problems in Pdfbox.
@@ -388,13 +359,31 @@ public class LayoutParsingPipeline {
    }


-    private static void rotateDirAdjExactly(List<TextPositionSequence> words, PDPage pdPage) {
+    private static void updateClassificationPage(PDPage pdPage,
+                                                 PDRectangle pdr,
+                                                 ClassificationPage classificationPage,
+                                                 CleanRulings cleanRulings,
+                                                 int pageNumber,
+                                                 PageInformation pageInformation) {
+
+        int rotation = pdPage.getRotation();
+        boolean isLandscape = pdr.getWidth() > pdr.getHeight() && (rotation == 0 || rotation == 180) || pdr.getHeight() > pdr.getWidth() && (rotation == 90 || rotation == 270);
+        classificationPage.setCleanRulings(cleanRulings);
+        classificationPage.setRotation(rotation);
+        classificationPage.setLandscape(isLandscape);
+        classificationPage.setPageNumber(pageNumber);
+        classificationPage.setPageWidth((float) pageInformation.width());
+        classificationPage.setPageHeight((float) pageInformation.height());
+    }
+
+
+    private static void rotateDirAdjExactly(List<Word> words, PDPage pdPage) {

        for (TextDirection dir : TextDirection.values()) {
-
            double averageRotation = words.stream()
-                    .map(TextPositionSequence::getTextPositions)
+                    .map(Word::getCharacters)
                    .flatMap(Collection::stream)
+                    .map(Character::getTextPosition)
                    .filter(pos -> pos.getDir().equals(dir))
                    .mapToDouble(RedTextPosition::getExactDir).average().orElse(0);

@@ -404,7 +393,7 @@ public class LayoutParsingPipeline {

            AffineTransform rotateInstance = AffineTransform.getRotateInstance(averageRotation, pdPage.getMediaBox().getWidth() / 2, pdPage.getMediaBox().getHeight() / 2);

-            for (TextPositionSequence word : words) {
+            for (Word word : words) {
                if (!dir.equals(word.getDir())) {
                    continue;
                }
@@ -455,10 +444,10 @@ public class LayoutParsingPipeline {
        // Collect all statistics for the classificationPage, except from blocks inside tables, as tables will always be added to BodyTextFrame.
        for (AbstractPageBlock textBlock : classificationPage.getTextBlocks()) {
            if (textBlock instanceof TextPageBlock) {
-                if (((TextPageBlock) textBlock).getSequences() == null) {
+                if (((TextPageBlock) textBlock).getWords() == null) {
                    continue;
                }
-                for (TextPositionSequence word : ((TextPageBlock) textBlock).getSequences()) {
+                for (Word word : ((TextPageBlock) textBlock).getWords()) {
                    classificationPage.getTextHeightCounter().add(word.getTextHeight());
                    classificationPage.getFontCounter().add(word.getFont());
                    classificationPage.getFontSizeCounter().add(word.getFontSize());
@@ -18,7 +18,7 @@ import com.knecon.fforesight.service.layoutparser.processor.docstrum.service.Zon
 import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRulings;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.RedTextPosition;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextDirection;
-import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;
 import com.knecon.fforesight.service.layoutparser.processor.visualization.LayoutDebugLayer;

 import lombok.RequiredArgsConstructor;
@@ -35,7 +35,7 @@ public class DocstrumSegmentationService {
    private final ReadingOrderService readingOrderService;


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

        EnumMap<TextDirection, Integer> directionCounts = new EnumMap<>(TextDirection.class);

@@ -78,18 +78,14 @@ public class DocstrumSegmentationService {
    }


-    private List<Zone> computeZones(List<TextPositionSequence> textPositions, CleanRulings rulings, LayoutDebugLayer visualizations, TextDirection direction) {
+    private List<Zone> computeZones(List<Word> textPositions, CleanRulings rulings, LayoutDebugLayer visualizations, TextDirection direction) {

-        List<RedTextPosition> positions = textPositions.stream()
+        List<Character> characters = textPositions.stream()
                .filter(t -> t.getDir() == direction)
-                .map(TextPositionSequence::getTextPositions)
+                .map(Word::getCharacters)
                .flatMap(List::stream)
                .toList();

-        List<Character> characters = positions.stream()
-                .map(Character::new)
-                .collect(Collectors.toList());
-
        nearestNeighbourService.findNearestNeighbors(characters);

        double characterSpacing = spacingService.computeCharacterSpacing(characters);
@@ -1,9 +1,9 @@
 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 static com.knecon.fforesight.service.layoutparser.processor.model.text.Word.BOLD;
+import static com.knecon.fforesight.service.layoutparser.processor.model.text.Word.BOLD_ITALIC;
+import static com.knecon.fforesight.service.layoutparser.processor.model.text.Word.ITALIC;
+import static com.knecon.fforesight.service.layoutparser.processor.model.text.Word.STANDARD;

 import java.util.ArrayList;
 import java.util.Arrays;
@@ -14,7 +14,7 @@ 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.Word;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.FontStyle;

 import lombok.Data;
@@ -24,7 +24,7 @@ import lombok.EqualsAndHashCode;
@EqualsAndHashCode(onlyExplicitlyIncluded = true, callSuper = false)
 public class Line extends TextBoundingBox {

-    private static final double WORD_DISTANCE_MULTIPLIER = 0.18;
+    private static final double WORD_DISTANCE_MULTIPLIER = 0.17;

    @EqualsAndHashCode.Include
    private final double x0;
@@ -36,18 +36,13 @@ public class Line extends TextBoundingBox {
    @EqualsAndHashCode.Include
    private final double y1;

-    private final double height;
-
    private FontStyle fontStyle;

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


    public Line(List<Character> characters, double wordSpacing) {

-        this.characters = characters;
-
        if (characters.size() >= 2) {
            // linear regression
            double sx = 0.0;
@@ -76,20 +71,32 @@ public class Line extends TextBoundingBox {
            this.y0 = character.getY() - dy;
            this.y1 = character.getY() + dy;
        }
-        height = computeHeight();
-        computeWords(wordSpacing * WORD_DISTANCE_MULTIPLIER);
+        this.words = new ArrayList<>();
+        computeWords(characters, wordSpacing * WORD_DISTANCE_MULTIPLIER);
        buildBBox();
        computeFontStyle();
    }


+    public Line(List<Word> words) {
+
+        this.words = words;
+        buildBBox();
+        x0 = getMinX();
+        y0 = getMinY();
+        x1 = getMaxX();
+        y1 = getMaxY();
+        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) {
+        for (Word word : words) {
            switch (word.getFontStyle()) {
                case STANDARD -> fontStyleCounter.get(FontStyle.REGULAR).getAndIncrement();
                case BOLD -> fontStyleCounter.get(FontStyle.BOLD).getAndIncrement();
@@ -100,8 +107,7 @@ public class Line extends TextBoundingBox {
        fontStyle = fontStyleCounter.entrySet()
                .stream()
                .max(Comparator.comparing(entry -> entry.getValue().get()))
-                .map(Map.Entry::getKey)
-                .orElse(FontStyle.REGULAR);
+                .map(Map.Entry::getKey).orElse(FontStyle.REGULAR);
    }


@@ -117,14 +123,6 @@ public class Line extends TextBoundingBox {
    }


-    private double computeHeight() {
-
-        return characters.stream()
-                       .map(Character::getHeight)
-                       .reduce(0d, Double::sum) / characters.size();
-    }
-
-
    public double angularDifference(Line j) {

        double diff = Math.abs(getAngle() - j.getAngle());
@@ -157,19 +155,22 @@ public class Line extends TextBoundingBox {
    }


-    private void computeWords(double wordSpacing) {
+    private void computeWords(List<Character> characters, double wordSpacing) {

-        TextPositionSequence word = new TextPositionSequence();
+        // Imo, the width of space should be scaled with the font size, but it only depends on the median distance between horizontal neighbours.
+        // If there are large differences in fontsize on a page, this might lead to missing spaces for the smaller fonts and too many for larger fonts.
+        // I've just now changed the scaling factor. If you come across this comment with missing whitespaces again, try scaling the fontsize instead of simply changing the factor again.
+        Word word = new Word();
        Character previous = null;
        for (Character current : characters) {
            if (previous != null) {
                double dist = current.getTextPosition().getXDirAdj() - previous.getTextPosition().getXDirAdj() - previous.getTextPosition().getWidthDirAdj();
                if (dist > wordSpacing) {
                    words.add(word);
-                    word = new TextPositionSequence();
+                    word = new Word();
                }
            }
-            word.getTextPositions().add(current.getTextPosition());
+            word.add(current);
            previous = current;
        }
        words.add(word);
@@ -178,9 +179,7 @@ public class Line extends TextBoundingBox {

    private void buildBBox() {

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


@@ -18,7 +18,6 @@ public class Zone extends TextBoundingBox {
    @SuppressWarnings("PMD.ConstructorCallsOverridableMethod")
    public Zone(List<Line> lines) {

-        lines.sort(Comparator.comparingDouble(Line::getY0));
        this.lines = lines;
        setToBBoxOfComponents(lines);
    }
@@ -17,7 +17,7 @@ public class LineBuilderService {

    private static final double CHARACTER_SPACING_DISTANCE_MULTIPLIER = 3.5;
    private static final double LINE_SPACING_THRESHOLD_MULTIPLIER = 0.67;
-    private static final double ANGLE_TOLERANCE = Math.PI / 6;
+    private static final double ANGLE_TOLERANCE = Math.toRadians(5);


    public List<Line> buildLines(List<Character> characters, double characterSpacing, double lineSpacing, CleanRulings rulings) {
@@ -1,9 +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.Collection;
 import java.util.HashSet;
 import java.util.List;
 import java.util.Set;
@@ -11,11 +9,12 @@ import java.util.stream.Collectors;

 import org.springframework.stereotype.Service;

-import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Character;
 import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Line;
 import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.UnionFind;
 import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Zone;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRulings;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;
+import com.knecon.fforesight.service.layoutparser.processor.utils.TextPositionOperations;

@Service
 public class ZoneBuilderService {
@@ -31,7 +30,7 @@ public class ZoneBuilderService {

    private static final double MAX_LINE_SIZE_SCALE = 2.5;

-    private static final double ANGLE_TOLERANCE = Math.PI / 6;
+    private static final double ANGLE_TOLERANCE = Math.toRadians(5);

    private static final double MAX_VERTICAL_MERGE_DISTANCE = 0.5;

@@ -114,64 +113,14 @@ public class ZoneBuilderService {

    private Zone mergeLinesInZone(List<Line> lines, double characterSpacing, double lineSpacing) {

-        double maxHorizontalDistance = 0;
-        double minVerticalDistance = 0;
-        double maxVerticalDistance = lineSpacing * MAX_VERTICAL_MERGE_DISTANCE;
+        Set<Word> words = lines.stream()
+                .map(Line::getWords)
+                .flatMap(Collection::stream)
+                .collect(Collectors.toSet());
+        Collection<Set<Word>> groupedLines = TextPositionOperations.groupByLine(words);

-        UnionFind<Line> unionFind = new UnionFind<>(new HashSet<>(lines));
-
-        lines.forEach(outer -> {
-            lines.forEach(inner -> {
-                if (inner == outer) {
-                    return;
-                }
-
-                double horizontalDistance = outer.horizontalDistance(inner);
-                double verticalDistance = outer.verticalDistance(inner);
-
-                if (horizontalDistance <= maxHorizontalDistance && minVerticalDistance <= verticalDistance && verticalDistance <= maxVerticalDistance) {
-
-                    unionFind.union(outer, inner);
-
-                } else if (minVerticalDistance <= verticalDistance
-                           && verticalDistance <= maxVerticalDistance
-                           && Math.abs(horizontalDistance - Math.min(outer.getLength(), inner.getLength())) < 0.1) {
-
-                    boolean characterOverlap = false;
-                    int overlappingCount = 0;
-                    for (Character outerCharacter : outer.getCharacters()) {
-                        for (Character innerCharacter : inner.getCharacters()) {
-                            double characterOverlapDistance = outerCharacter.overlappingDistance(innerCharacter);
-                            if (characterOverlapDistance > 2) {
-                                characterOverlap = true;
-                            }
-                            if (characterOverlapDistance > 0) {
-                                overlappingCount++;
-                            }
-                        }
-                    }
-                    if (!characterOverlap && overlappingCount <= 2) {
-                        unionFind.union(outer, inner);
-                    }
-                }
-
-            });
-        });
-
-        List<Line> outputZone = new ArrayList<>();
-        for (Set<Line> group : unionFind.getGroups()) {
-            List<Character> characters = new ArrayList<>();
-            for (Line line : group) {
-                characters.addAll(line.getCharacters());
-            }
-            characters.sort(Comparator.comparingDouble(Character::getX));
-
-            outputZone.add(new Line(characters, characterSpacing));
-        }
-
-        return new Zone(outputZone.stream()
-                                .sorted(Comparator.comparing(Line::getY0))
-                                .collect(Collectors.toList()));
+        List<Line> sortedLines = TextPositionOperations.sortLines(groupedLines);
+        return new Zone(sortedLines);
    }

 }
@@ -3,7 +3,7 @@ package com.knecon.fforesight.service.layoutparser.processor.model;
 import java.awt.geom.Rectangle2D;
 import java.util.List;

-import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;

 import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
@@ -16,8 +16,8 @@ import lombok.experimental.FieldDefaults;
 public class LineInformation {

    List<Rectangle2D> lineBBox;
-    List<List<TextPositionSequence>> sequencesByLines;
+    List<List<Word>> sequencesByLines;
    List<List<Rectangle2D>> bBoxWithGapsByLines;
-    List<List<List<TextPositionSequence>>> sequencesWithGapsByLines;
+    List<List<List<Word>>> sequencesWithGapsByLines;

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

 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.model.text.Word;

 import lombok.AllArgsConstructor;
 import lombok.Builder;
@@ -15,7 +15,7 @@ import lombok.Getter;
@AllArgsConstructor
 public class PageContents {

-    List<TextPositionSequence> sortedTextPositionSequences;
+    List<Word> sortedWords;
    Rectangle2D cropBox;
    Rectangle2D mediaBox;
    List<Ruling> rulings;
@@ -16,8 +16,8 @@ import lombok.experimental.FieldDefaults;
@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
 public class SectionIdentifier {

-    public static Pattern numericalIdentifierPattern = Pattern.compile("^[\\s]?(\\d+)[\\s.,;]?(\\d+)?[\\s.,;]?(\\d+)?[\\s.,;]?(\\d+)?");
-    public static Pattern alphanumericIdentifierPattern = Pattern.compile("^[\\s]?[A-Za-z][\\s.,;]?(\\d+)[\\s.,;]?(\\d+)?[\\s.,;]?(\\d+)?[\\s.,;]?(\\d+)?[\\s.,;]?");
+    public static Pattern numericalIdentifierPattern = Pattern.compile("^[\\s]?(\\d{1,2})(?:[\\s.,;](\\d{1,2}))?(?:[\\s.,;](\\d{1,2}))?(?:[\\s.,;](\\d{1,2}))?");
+    public static Pattern alphanumericIdentifierPattern = Pattern.compile("^[\\s]?[A-Za-z][\\s.,;]?(\\d{1,2})[\\s.,;]?(\\d{1,2})?[\\s.,;]?(\\d{1,2})?[\\s.,;]?(\\d{1,2})?[\\s.,;]?");

    public enum Format {
        EMPTY,
@@ -9,7 +9,7 @@ import java.util.List;

 import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.BoundingBox;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
-import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;
 import com.knecon.fforesight.service.layoutparser.processor.utils.TextNormalizationUtilities;

 import lombok.Data;
@@ -68,12 +68,12 @@ public class Cell extends BoundingBox {
        StringBuilder sb = new StringBuilder();

        Iterator<TextPageBlock> itty = textBlocks.iterator();
-        TextPositionSequence previous = null;
+        Word previous = null;
        while (itty.hasNext()) {

            TextPageBlock textBlock = itty.next();

-            for (TextPositionSequence word : textBlock.getSequences()) {
+            for (Word word : textBlock.getWords()) {
                if (previous != null) {
                    if (Math.abs(previous.getMaxYDirAdj() - word.getMaxYDirAdj()) > word.getTextHeight()) {
                        sb.append('\n');
@@ -14,16 +14,17 @@ import lombok.experimental.FieldDefaults;
@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
 public class ListIdentifier {

-    public static final Pattern NUMBER_WITH_DOT = Pattern.compile("^\\s*([1-9]+)\\.\\s+");
-    public static final Pattern NUMBER_IN_PARENTHESES = Pattern.compile("^\\s*\\(([1-9]+)\\)\\s+");
+    public static final Pattern NUMBER_WITH_DOT = Pattern.compile("^\\s*([1-9]{1,4})\\.\\s+");
+    public static final Pattern NUMBER_IN_PARENTHESES = Pattern.compile("^\\s*\\(([1-9]{1,4})\\)\\s+");

    enum Format {
-        NUMBER_WITH_DOT, NUMBER_IN_PARENTHESES
+        NUMBER_WITH_DOT,
+        NUMBER_IN_PARENTHESES
    }

    Format format;
    @Getter
-    TextPositionSequence word;
+    Word word;
    @Getter
    int page;
    int representation;
@@ -31,14 +32,14 @@ public class ListIdentifier {

    public static Optional<ListIdentifier> parse(TextPageBlock textPageBlock, int page) {

-        return parse(textPageBlock.getSequences().subList(0, Math.min(5, textPageBlock.getSequences().size())), page);
+        return parse(textPageBlock.getWords().subList(0, Math.min(5, textPageBlock.getWords().size())), page);
    }


-    public static Optional<ListIdentifier> parse(List<TextPositionSequence> sequences, int page) {
+    public static Optional<ListIdentifier> parse(List<Word> sequences, int page) {

        StringBuilder sb = new StringBuilder();
-        for (TextPositionSequence sequence : sequences) {
+        for (Word sequence : sequences) {
            sb.append(sequence.toString());
            sb.append(" ");
        }
@@ -48,17 +49,33 @@ public class ListIdentifier {
        Matcher numberMatcher = NUMBER_WITH_DOT.matcher(text);

        if (numberMatcher.find()) {
-            return Optional.of(new ListIdentifier(Format.NUMBER_WITH_DOT, sequences.get(0), page, Integer.parseInt(numberMatcher.group(1))));
+            Optional<Integer> representation = parseInteger(numberMatcher.group(1));
+            if (representation.isPresent()) {
+                return Optional.of(new ListIdentifier(Format.NUMBER_WITH_DOT, sequences.get(0), page, representation.get()));
+            }
        }

        Matcher parenthesisMatcher = NUMBER_IN_PARENTHESES.matcher(text);
        if (parenthesisMatcher.find()) {
-            return Optional.of(new ListIdentifier(Format.NUMBER_IN_PARENTHESES, sequences.get(0), page, Integer.parseInt(parenthesisMatcher.group(1))));
+            Optional<Integer> representation = parseInteger(parenthesisMatcher.group(1));
+            if (representation.isPresent()) {
+                return Optional.of(new ListIdentifier(Format.NUMBER_IN_PARENTHESES, sequences.get(0), page, representation.get()));
+            }
        }
        return Optional.empty();
    }


+    private static Optional<Integer> parseInteger(String text) {
+
+        try {
+            return Optional.of(Integer.parseInt(text));
+        } catch (NumberFormatException e) {
+            return Optional.empty();
+        }
+    }
+
+
    public static boolean isInOrder(List<ListIdentifier> listIdentifiers) {

        if (listIdentifiers.size() <= 1) {
@@ -10,18 +10,18 @@ import lombok.Getter;
@Getter
 public class SearchableText {

-    private final List<TextPositionSequence> sequences = new ArrayList<>();
+    private final List<Word> sequences = new ArrayList<>();


-    public void add(TextPositionSequence textPositionSequence) {
+    public void add(Word word) {

-        sequences.add(textPositionSequence);
+        sequences.add(word);
    }


-    public void addAll(List<TextPositionSequence> textPositionSequences) {
+    public void addAll(List<Word> words) {

-        sequences.addAll(textPositionSequences);
+        sequences.addAll(words);
    }


@@ -32,10 +32,10 @@ public class SearchableText {
    }


-    public static String buildString(List<TextPositionSequence> sequences) {
+    public static String buildString(List<Word> sequences) {

        StringBuilder sb = new StringBuilder();
-        for (TextPositionSequence word : sequences) {
+        for (Word word : sequences) {
            sb.append(word);
            sb.append(' ');
        }
@@ -25,7 +25,7 @@ import lombok.NoArgsConstructor;
 public class TextPageBlock extends AbstractPageBlock {

    @Builder.Default
-    private List<TextPositionSequence> sequences = new ArrayList<>();
+    private List<Word> words = new ArrayList<>();
    @Builder.Default
    private FrequencyCounters frequencyCounters = new FrequencyCounters();

@@ -41,43 +41,43 @@ public class TextPageBlock extends AbstractPageBlock {
    private boolean changed;


-    public TextPageBlock(List<TextPositionSequence> sequences) {
+    public TextPageBlock(List<Word> words) {

-        this.sequences = new ArrayList<>(sequences);
+        this.words = new ArrayList<>(words);
        this.frequencyCounters = new FrequencyCounters();

-        if (!sequences.isEmpty()) {
-            addToFrequencyCounters(sequences);
+        if (!words.isEmpty()) {
+            addToFrequencyCounters(words);
        }
        calculateBBox();
    }


-    public List<TextPositionSequence> getSequences() {
+    public List<Word> getWords() {

-        return Collections.unmodifiableList(sequences);
+        return Collections.unmodifiableList(words);
    }


    public TextDirection getDir() {

-        return sequences.get(0).getDir();
+        return words.get(0).getDir();
    }


    private void calculateBBox() {

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

-        setToBBoxOfComponents(sequences);
+        setToBBoxOfComponents(words);
    }


@@ -99,8 +99,8 @@ public class TextPageBlock extends AbstractPageBlock {
            throw new IllegalArgumentException("Cannot merge textBlocks on different pages.");
        }

-        List<TextPositionSequence> sequences = textBlocksToMerge.stream()
-                .map(TextPageBlock::getSequences)
+        List<Word> sequences = textBlocksToMerge.stream()
+                .map(TextPageBlock::getWords)
                .flatMap(java.util.Collection::stream)
                .toList();
        sequences = new ArrayList<>(sequences);
@@ -109,9 +109,9 @@ public class TextPageBlock extends AbstractPageBlock {
    }


-    private void addToFrequencyCounters(List<TextPositionSequence> sequences) {
+    private void addToFrequencyCounters(List<Word> sequences) {

-        for (TextPositionSequence wordBlock : sequences) {
+        for (Word wordBlock : sequences) {

            frequencyCounters.getLineHeightFrequencyCounter().add(wordBlock.getTextHeight());
            frequencyCounters.getFontSizeFrequencyCounter().add(wordBlock.getFontSize());
@@ -120,12 +120,12 @@ public class TextPageBlock extends AbstractPageBlock {
            frequencyCounters.getStyleFrequencyCounter().add(wordBlock.getFontStyle());
        }

-        setUnderlined(this.sequences.stream()
-                              .allMatch(TextPositionSequence::isUnderline));
+        setUnderlined(this.words.stream()
+                              .allMatch(Word::isUnderline));
    }


-    public TextPageBlock union(TextPositionSequence r) {
+    public TextPageBlock union(Word r) {

        TextPageBlock union = this.copy();
        union.add(r);
@@ -138,8 +138,8 @@ public class TextPageBlock extends AbstractPageBlock {
    public TextPageBlock union(TextPageBlock r) {

        TextPageBlock union = this.copy();
-        union.addAll(r.getSequences());
-        addToFrequencyCounters(r.getSequences());
+        union.addAll(r.getWords());
+        addToFrequencyCounters(r.getWords());
        calculateBBox();
        return union;
    }
@@ -148,33 +148,33 @@ public class TextPageBlock extends AbstractPageBlock {
    public void add(TextPageBlock textPageBlock) {

        changed = true;
-        sequences.addAll(textPageBlock.getSequences());
-        addToFrequencyCounters(textPageBlock.getSequences());
+        words.addAll(textPageBlock.getWords());
+        addToFrequencyCounters(textPageBlock.getWords());
        calculateBBox();
    }


-    public void add(TextPositionSequence textPositionSequence) {
+    public void add(Word word) {

        changed = true;
-        sequences.add(textPositionSequence);
-        addToFrequencyCounters(List.of(textPositionSequence));
+        words.add(word);
+        addToFrequencyCounters(List.of(word));
        calculateBBox();
    }


-    public void addAll(List<TextPositionSequence> textPositionSequences) {
+    public void addAll(List<Word> words) {

        changed = true;
-        sequences.addAll(textPositionSequences);
-        addToFrequencyCounters(textPositionSequences);
+        this.words.addAll(words);
+        addToFrequencyCounters(words);
        calculateBBox();
    }


    public TextPageBlock copy() {

-        return new TextPageBlock(new ArrayList<>(sequences));
+        return new TextPageBlock(new ArrayList<>(words));
    }


@@ -193,8 +193,8 @@ public class TextPageBlock extends AbstractPageBlock {

            StringBuilder sb = new StringBuilder();

-            TextPositionSequence previous = null;
-            for (TextPositionSequence word : sequences) {
+            Word previous = null;
+            for (Word word : words) {
                if (previous != null) {
                    if (Math.abs(previous.getMaxYDirAdj() - word.getMaxYDirAdj()) > word.getTextHeight()) {
                        sb.append('\n');
@@ -217,8 +217,8 @@ public class TextPageBlock extends AbstractPageBlock {
    public int getNumberOfLines() {

        int numberOfLines = 1;
-        TextPositionSequence previous = null;
-        for (TextPositionSequence word : sequences) {
+        Word previous = null;
+        for (Word word : words) {
            if (previous != null) {
                if (word.getMaxYDirAdj() - previous.getMaxYDirAdj() > word.getTextHeight()) {
                    numberOfLines++;
@@ -270,7 +270,7 @@ public class TextPageBlock extends AbstractPageBlock {
    @Override
    public boolean isEmpty() {

-        return sequences.isEmpty();
+        return words.isEmpty();
    }

 }
@@ -1,34 +0,0 @@
-package com.knecon.fforesight.service.layoutparser.processor.model.text;
-
-import java.util.Comparator;
-import java.util.HashMap;
-
-public class TextPositionSequenceComparator implements Comparator<TextPositionSequence> {
-
-    private HashMap<TextPositionSequence, TextBlockOnPage> lookup;
-
-
-    public TextPositionSequenceComparator(HashMap<TextPositionSequence, TextBlockOnPage> lookup) {
-
-        this.lookup = lookup;
-    }
-
-
-    @Override
-    public int compare(TextPositionSequence number1, TextPositionSequence number2) {
-
-        int page1 = lookup.get(number1).page().getPageNumber();
-        int page2 = lookup.get(number2).page().getPageNumber();
-
-        if (page1 != page2) {
-            return Integer.compare(page1, page2);
-        }
-
-        if (number1.getY() != number2.getY()) {
-            return Double.compare(number1.getY(), number2.getY());
-        }
-
-        return Integer.compare(Integer.parseInt(number1.toString()), Integer.parseInt(number2.toString()));
-    }
-
-}
@@ -0,0 +1,36 @@
+package com.knecon.fforesight.service.layoutparser.processor.model.text;
+
+import java.util.Comparator;
+import java.util.HashMap;
+
+import com.knecon.fforesight.service.layoutparser.processor.services.classification.NumberWord;
+
+public class TocNumberComparator implements Comparator<NumberWord> {
+
+    private HashMap<NumberWord, TextBlockOnPage> lookup;
+
+
+    public TocNumberComparator(HashMap<NumberWord, TextBlockOnPage> lookup) {
+
+        this.lookup = lookup;
+    }
+
+
+    @Override
+    public int compare(NumberWord number1, NumberWord number2) {
+
+        int page1 = lookup.get(number1).page().getPageNumber();
+        int page2 = lookup.get(number2).page().getPageNumber();
+
+        if (page1 != page2) {
+            return Integer.compare(page1, page2);
+        }
+
+        if (number1.word().getY() != number2.word().getY()) {
+            return Double.compare(number1.word().getY(), number2.word().getY());
+        }
+
+        return Integer.compare(number1.number(), number2.number());
+    }
+
+}
@@ -13,6 +13,7 @@ import java.util.stream.Collectors;

 import org.apache.pdfbox.text.TextPosition;

+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Character;
 import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.TextBoundingBox;

 import lombok.AllArgsConstructor;
@@ -27,7 +28,7 @@ import lombok.extern.slf4j.Slf4j;
@NoArgsConstructor
@AllArgsConstructor
@SuppressWarnings("pmd")
-public class TextPositionSequence extends TextBoundingBox implements CharSequence {
+public class Word extends TextBoundingBox implements CharSequence {

    public static final String STANDARD = "standard";
    public static final String BOLD_ITALIC = "bold, italic";
@@ -38,8 +39,7 @@ public class TextPositionSequence extends TextBoundingBox implements CharSequenc
    private int page;

    @Builder.Default
-    private List<RedTextPosition> textPositions = new ArrayList<>();
-
+    private List<Character> characters = new ArrayList<>();
    private boolean isParagraphStart;
    private boolean strikethrough;
    private boolean underline;
@@ -47,10 +47,11 @@ public class TextPositionSequence extends TextBoundingBox implements CharSequenc
    private Integer hashcodeCache;


-    public TextPositionSequence(List<TextPosition> textPositions, int pageNumber, boolean isParagraphStart) {
+    public Word(List<TextPosition> textPositions, int pageNumber, boolean isParagraphStart) {

-        this.textPositions = textPositions.stream()
+        this.characters = textPositions.stream()
                .map(RedTextPosition::fromTextPosition)
+                .map(Character::new)
                .collect(Collectors.toList());
        this.page = pageNumber;
        this.isParagraphStart = isParagraphStart;
@@ -65,9 +66,9 @@ public class TextPositionSequence extends TextBoundingBox implements CharSequenc
    }


-    public TextPositionSequence(List<RedTextPosition> textPositions, int page) {
+    public Word(List<Character> textPositions, int page) {

-        this.textPositions = textPositions;
+        this.characters = new ArrayList<>(textPositions);
        this.page = page;
        calculateBBoxAndHashcode();
    }
@@ -76,7 +77,7 @@ public class TextPositionSequence extends TextBoundingBox implements CharSequenc
    @Override
    public int length() {

-        return textPositions.size();
+        return characters.size();
    }


@@ -98,10 +99,10 @@ public class TextPositionSequence extends TextBoundingBox implements CharSequenc


    @Override
-    public TextPositionSequence subSequence(int start, int end) {
+    public Word subSequence(int start, int end) {

-        var textPositionSequence = new TextPositionSequence();
-        textPositionSequence.textPositions = textPositions.subList(start, end);
+        var textPositionSequence = new Word();
+        textPositionSequence.characters = characters.subList(start, end);
        textPositionSequence.page = page;
        textPositionSequence.dir = dir;
        textPositionSequence.setToBBoxOfComponents(getTextPositions());
@@ -122,53 +123,59 @@ public class TextPositionSequence extends TextBoundingBox implements CharSequenc

    public RedTextPosition textPositionAt(int index) {

-        return textPositions.get(index);
+        return characters.get(index).getTextPosition();
    }


-    public void add(TextPositionSequence textPositionSequence, RedTextPosition textPosition) {
+    public void add(Word word, RedTextPosition textPosition) {

-        this.textPositions.add(textPosition);
-        this.page = textPositionSequence.getPage();
+        this.characters.add(new Character(textPosition));
+        this.page = word.getPage();
+        calculateBBoxAndHashcode();
+    }
+
+
+    public void add(Character current) {
+
+        characters.add(current);
        calculateBBoxAndHashcode();
    }


    public void add(TextPosition textPosition) {

-        this.textPositions.add(RedTextPosition.fromTextPosition(textPosition));
-        calculateBBoxAndHashcode();
+        add(new Character(RedTextPosition.fromTextPosition(textPosition)));
    }


    public double getTextHeightNoPadding() {

-        return textPositions.get(0).getHeightDirAdj();
+        return characters.get(0).getTextPosition().getHeightDirAdj();
    }


    public double getTextHeight() {

-        return textPositions.get(0).getHeightDirAdj() + HEIGHT_PADDING;
+        return characters.get(0).getTextPosition().getHeightDirAdj() + HEIGHT_PADDING;
    }


    public String getFont() {

-        if (textPositions.get(0).getFontName() == null) {
+        if (characters.get(0).getTextPosition().getFontName() == null) {
            return "none";
        }

-        return FONT_CLEANER.matcher(textPositions.get(0).getFontName().toLowerCase(Locale.ROOT)).replaceAll("");
+        return FONT_CLEANER.matcher(characters.get(0).getTextPosition().getFontName().toLowerCase(Locale.ROOT)).replaceAll("");
    }


    public String getFontStyle() {

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

        if (lowercaseFontName.contains(BOLD) && lowercaseFontName.contains(ITALIC)) {
            return BOLD_ITALIC;
@@ -184,13 +191,13 @@ public class TextPositionSequence extends TextBoundingBox implements CharSequenc

    public float getFontSize() {

-        return textPositions.get(0).getFontSizeInPt();
+        return characters.get(0).getTextPosition().getFontSizeInPt();
    }


    public float getSpaceWidth() {

-        return textPositions.get(0).getWidthOfSpace();
+        return characters.get(0).getTextPosition().getWidthOfSpace();
    }


@@ -199,7 +206,7 @@ public class TextPositionSequence extends TextBoundingBox implements CharSequenc
        if (o == this) {
            return true;
        }
-        if (!(o instanceof TextPositionSequence other)) {
+        if (!(o instanceof Word other)) {
            return false;
        }
        if (!other.canEqual((Object) this)) {
@@ -220,7 +227,7 @@ public class TextPositionSequence extends TextBoundingBox implements CharSequenc
    }


-    protected boolean canEqual(final Object other) {return other instanceof TextPositionSequence;}
+    protected boolean canEqual(final Object other) {return other instanceof Word;}


    public int hashCode() {
@@ -244,6 +251,14 @@ public class TextPositionSequence extends TextBoundingBox implements CharSequenc
    }


+    private List<RedTextPosition> getTextPositions() {
+
+        return characters.stream()
+                .map(Character::getTextPosition)
+                .toList();
+    }
+
+
    public void transform(AffineTransform rotateInstance) {

        for (RedTextPosition textPosition : getTextPositions()) {
@@ -23,11 +23,11 @@ public class DividingColumnDetectionService {
    public List<Rectangle2D> detectColumns(PageContents pageContents) {


-        if (pageContents.getSortedTextPositionSequences().size() < 2) {
+        if (pageContents.getSortedWords().size() < 2) {
            return List.of(pageContents.getCropBox());
        }

-        GapInformation linesWithGapInformation = GapDetectionService.findGapsInLines(pageContents.getSortedTextPositionSequences(), pageContents.getCropBox());
+        GapInformation linesWithGapInformation = GapDetectionService.findGapsInLines(pageContents.getSortedWords(), pageContents.getCropBox());

        return detectColumnsFromLines(linesWithGapInformation.getXGaps(), pageContents.getCropBox());
    }
@@ -5,7 +5,7 @@ import java.util.LinkedList;
 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.model.text.Word;

 import lombok.AllArgsConstructor;
 import lombok.experimental.UtilityClass;
@@ -18,23 +18,23 @@ public class GapDetectionService {
    private static final double NEW_LINE_FACTOR = 0.2;


-    public static GapInformation findGapsInLines(List<TextPositionSequence> sortedTextPositionSequences, Rectangle2D mainBodyTextFrame) {
+    public static GapInformation findGapsInLines(List<Word> sortedWords, Rectangle2D mainBodyTextFrame) {

-        if (sortedTextPositionSequences.isEmpty()) {
+        if (sortedWords.isEmpty()) {
            return new GapInformation();
        }

-        final double avgTextPositionHeight = getAvgTextPositionHeight(sortedTextPositionSequences);
+        final double avgTextPositionHeight = getAvgTextPositionHeight(sortedWords);

        XGapsContext xGapContext = XGapsContext.init(mainBodyTextFrame);
        YGapsContext yGapContext = YGapsContext.init(mainBodyTextFrame);

-        var previousTextPosition = sortedTextPositionSequences.get(0);
+        var previousTextPosition = sortedWords.get(0);
        Rectangle2D rectangle = toRectangle2D(previousTextPosition);

        xGapContext.addGapFromLeftEdgeOfMainBody(rectangle);

-        for (TextPositionSequence currentTextPosition : sortedTextPositionSequences.subList(1, sortedTextPositionSequences.size())) {
+        for (Word currentTextPosition : sortedWords.subList(1, sortedWords.size())) {

            double yDifference = Math.abs(currentTextPosition.getMaxYDirAdj() - previousTextPosition.getMaxYDirAdj());
            double xGap = Math.abs(previousTextPosition.getMaxXDirAdj() - currentTextPosition.getXDirAdj());
@@ -59,14 +59,14 @@ public class GapDetectionService {
            }
            previousTextPosition = currentTextPosition;
        }
-        xGapContext.addGapToRightEdgeOfMainBody(toRectangle2D(sortedTextPositionSequences.get(sortedTextPositionSequences.size() - 1)));
+        xGapContext.addGapToRightEdgeOfMainBody(toRectangle2D(sortedWords.get(sortedWords.size() - 1)));
        xGapContext.gapsPerLine.add(xGapContext.gapsInCurrentLine);

        return new GapInformation(xGapContext.gapsPerLine, yGapContext.gapsPerLine);
    }


-    private static Rectangle2D toRectangle2D(TextPositionSequence textPosition) {
+    private static Rectangle2D toRectangle2D(Word textPosition) {

        return mirrorY(textPosition.getBBox());
    }
@@ -87,18 +87,18 @@ public class GapDetectionService {
    }


-    private static void assertAllTextPositionsHaveSameDir(List<TextPositionSequence> textPositionSequences) {
+    private static void assertAllTextPositionsHaveSameDir(List<Word> words) {

-        assert textPositionSequences.stream()
-                .map(TextPositionSequence::getDir)
-                .allMatch(a -> a.equals(textPositionSequences.get(0).getDir()));
+        assert words.stream()
+                .map(Word::getDir)
+                .allMatch(a -> a.equals(words.get(0).getDir()));
    }


-    private static double getAvgTextPositionHeight(List<TextPositionSequence> textPositionSequences) {
+    private static double getAvgTextPositionHeight(List<Word> words) {

-        return textPositionSequences.stream()
-                .mapToDouble(TextPositionSequence::getHeight).average().orElseThrow();
+        return words.stream()
+                .mapToDouble(Word::getHeight).average().orElseThrow();
    }


@@ -7,17 +7,17 @@ import java.util.List;

 import com.knecon.fforesight.service.layoutparser.processor.model.GapInformation;
 import com.knecon.fforesight.service.layoutparser.processor.model.LineInformation;
-import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;

 import lombok.experimental.UtilityClass;

@UtilityClass
 public class InvisibleTableDetectionService {

-    public List<List<Rectangle2D>> detectTable(List<TextPositionSequence> textPositionSequences, Rectangle2D tableBBox) {
+    public List<List<Rectangle2D>> detectTable(List<Word> words, Rectangle2D tableBBox) {

-        LineInformation lineInformation = LineDetectionService.calculateLineInformation(textPositionSequences);
-        GapInformation gaps = GapDetectionService.findGapsInLines(textPositionSequences, tableBBox);
+        LineInformation lineInformation = LineDetectionService.calculateLineInformation(words);
+        GapInformation gaps = GapDetectionService.findGapsInLines(words, tableBBox);
        List<Rectangle2D> gapsAcrossLines = GapsAcrossLinesService.detectXGapsAcrossLines(gaps, tableBBox);
        List<Double> columnXCoords = gapsAcrossLines.stream().map(RectangularShape::getCenterX).toList();
        int colCount = gapsAcrossLines.size();
@@ -7,7 +7,7 @@ import java.util.List;
 import com.knecon.fforesight.service.layoutparser.processor.model.GapInformation;
 import com.knecon.fforesight.service.layoutparser.processor.model.LineInformation;
 import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
-import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;

 import lombok.AllArgsConstructor;
 import lombok.Getter;
@@ -19,37 +19,37 @@ public class LineDetectionService {
    private static final double X_GAP_FACTOR = 1; // multiplied with average text height, determines the minimum distance of gaps in lines


-    public LineInformation calculateLineInformation(List<TextPositionSequence> sortedTextPositionSequences) {
+    public LineInformation calculateLineInformation(List<Word> sortedWords) {

-        if (sortedTextPositionSequences.isEmpty()) {
+        if (sortedWords.isEmpty()) {
            return LineFactory.init().build();
        }

-        return buildLineInformation(sortedTextPositionSequences);
+        return buildLineInformation(sortedWords);
    }


-    public List<List<Rectangle2D>> findLinesWithGaps(List<TextPositionSequence> sortedTextPositionSequences, Rectangle2D mainBodyTextFrame) {
+    public List<List<Rectangle2D>> findLinesWithGaps(List<Word> sortedWords, Rectangle2D mainBodyTextFrame) {

-        return calculateLineInformation(sortedTextPositionSequences).getBBoxWithGapsByLines();
+        return calculateLineInformation(sortedWords).getBBoxWithGapsByLines();
    }


-    public List<List<TextPositionSequence>> orderByLines(List<TextPositionSequence> sortedTextPositionSequences, Rectangle2D mainBodyTextFrame) {
+    public List<List<Word>> orderByLines(List<Word> sortedWords, Rectangle2D mainBodyTextFrame) {

-        return calculateLineInformation(sortedTextPositionSequences).getSequencesByLines();
+        return calculateLineInformation(sortedWords).getSequencesByLines();
    }


-    private static LineInformation buildLineInformation(List<TextPositionSequence> sortedTextPositionSequences) {
+    private static LineInformation buildLineInformation(List<Word> sortedWords) {

-        final double avgTextPositionHeight = getAvgTextPositionHeight(sortedTextPositionSequences);
+        final double avgTextPositionHeight = getAvgTextPositionHeight(sortedWords);

        LineFactory lineFactory = LineFactory.init();

-        var previousTextPosition = sortedTextPositionSequences.get(0);
+        var previousTextPosition = sortedWords.get(0);
        lineFactory.addToCurrentLine(previousTextPosition);
-        for (TextPositionSequence currentTextPosition : sortedTextPositionSequences.subList(1, sortedTextPositionSequences.size())) {
+        for (Word currentTextPosition : sortedWords.subList(1, sortedWords.size())) {
            if (isNewLine(currentTextPosition, previousTextPosition, avgTextPositionHeight) || isSplitByOrientation(currentTextPosition, previousTextPosition)) {
                lineFactory.startNewLine();
            } else if (isXGap(currentTextPosition, previousTextPosition, avgTextPositionHeight)) {
@@ -63,25 +63,25 @@ public class LineDetectionService {
    }


-    private static double getAvgTextPositionHeight(List<TextPositionSequence> textPositionSequences) {
+    private static double getAvgTextPositionHeight(List<Word> words) {

-        return textPositionSequences.stream().mapToDouble(TextPositionSequence::getHeight).average().orElseThrow();
+        return words.stream().mapToDouble(Word::getHeight).average().orElseThrow();
    }


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

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


-    private static boolean isSplitByOrientation(TextPositionSequence currentTextPosition, TextPositionSequence previousTextPosition) {
+    private static boolean isSplitByOrientation(Word currentTextPosition, Word previousTextPosition) {

        return !previousTextPosition.getDir().equals(currentTextPosition.getDir());
    }


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

        return Math.abs(previousTextPosition.getYDirAdj() - currentTextPosition.getYDirAdj()) > avgTextPositionHeight;
    }
@@ -96,13 +96,13 @@ public class LineDetectionService {
        List<List<Rectangle2D>> bBoxWithGapsByLines;
        List<Rectangle2D> bBoxWithGapsInCurrentLine;

-        List<List<List<TextPositionSequence>>> sequencesWithGapsByLines;
-        List<List<TextPositionSequence>> sequencesWithGapsInCurrentLine;
+        List<List<List<Word>>> sequencesWithGapsByLines;
+        List<List<Word>> sequencesWithGapsInCurrentLine;

-        List<TextPositionSequence> currentSequencesWithoutGaps;
+        List<Word> currentSequencesWithoutGaps;

-        List<List<TextPositionSequence>> sequencesByLines;
-        List<TextPositionSequence> sequencesInCurrentLine;
+        List<List<Word>> sequencesByLines;
+        List<Word> sequencesInCurrentLine;

        List<List<Rectangle2D>> xGaps;
        List<List<Rectangle2D>> yGaps;
@@ -116,14 +116,14 @@ public class LineDetectionService {
            List<Rectangle2D> bBoxWithGapsInCurrentLine = new LinkedList<>();
            bBoxWithGapsByLines.add(bBoxWithGapsInCurrentLine);

-            List<List<List<TextPositionSequence>>> sequencesWithGapsByLines = new LinkedList<>();
-            List<List<TextPositionSequence>> sequencesWithGapsInCurrentLine = new LinkedList<>();
+            List<List<List<Word>>> sequencesWithGapsByLines = new LinkedList<>();
+            List<List<Word>> sequencesWithGapsInCurrentLine = new LinkedList<>();
            sequencesWithGapsByLines.add(sequencesWithGapsInCurrentLine);
-            List<TextPositionSequence> currentSequencesWithoutGaps = new LinkedList<>();
+            List<Word> currentSequencesWithoutGaps = new LinkedList<>();
            sequencesWithGapsInCurrentLine.add(currentSequencesWithoutGaps);

-            List<List<TextPositionSequence>> sequencesByLines = new LinkedList<>();
-            List<TextPositionSequence> sequencesInCurrentLine = new LinkedList<>();
+            List<List<Word>> sequencesByLines = new LinkedList<>();
+            List<Word> sequencesInCurrentLine = new LinkedList<>();
            sequencesByLines.add(sequencesInCurrentLine);

            return new LineFactory(lineBBox,
@@ -178,13 +178,13 @@ public class LineDetectionService {
        }


-        private Rectangle2D textPositionBBox(List<TextPositionSequence> textPositionSequences) {
+        private Rectangle2D textPositionBBox(List<Word> words) {

-            return RectangleTransformations.rectangle2DBBox(textPositionSequences.stream().map(TextPositionSequence::getBBox).toList());
+            return RectangleTransformations.rectangle2DBBox(words.stream().map(Word::getBBox).toList());
        }


-        public void addToCurrentLine(TextPositionSequence current) {
+        public void addToCurrentLine(Word current) {

            sequencesInCurrentLine.add(current);
            currentSequencesWithoutGaps.add(current);
@@ -13,7 +13,7 @@ import org.apache.pdfbox.pdmodel.PDPage;
 import org.springframework.core.io.ClassPathResource;

 import com.knecon.fforesight.service.layoutparser.processor.model.PageContents;
-import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;
 import com.knecon.fforesight.service.layoutparser.processor.services.parsing.PDFLinesTextStripper;
 import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;

@@ -40,7 +40,7 @@ public class PageContentExtractor {
                stripper.setPdpage(pdPage);
                stripper.getText(pdDocument);

-                Map<Float, List<TextPositionSequence>> sortedTextPositionSequencesPerDir = stripper.getTextPositionSequences()
+                Map<Float, List<Word>> sortedTextPositionSequencesPerDir = stripper.getWords()
                        .stream()
                        .collect(Collectors.groupingBy(textPositionSequence -> textPositionSequence.getDir().getDegrees()));

@@ -57,7 +57,7 @@ public class PageContentExtractor {
    }


-    public List<TextPositionSequence> sortByDirAccordingToPageRotation(Map<Float, List<TextPositionSequence>> sortedTextPositionSequencesPerDir, int rotation) {
+    public List<Word> sortByDirAccordingToPageRotation(Map<Float, List<Word>> sortedTextPositionSequencesPerDir, int rotation) {

        LinkedList<Float> sortedKeys = new LinkedList<>(sortedTextPositionSequencesPerDir.keySet().stream().sorted().toList());

@@ -14,9 +14,9 @@ public class PageInformationService {

    public PageInformation build(PageContents pageContents) {

-        LineInformation lineInformation = LineDetectionService.calculateLineInformation(pageContents.getSortedTextPositionSequences());
+        LineInformation lineInformation = LineDetectionService.calculateLineInformation(pageContents.getSortedWords());
        Rectangle2D mainBodyTextFrame = MainBodyTextFrameExtractionService.calculateMainBodyTextFrame(lineInformation);
-        GapInformation gapInformation = GapDetectionService.findGapsInLines(pageContents.getSortedTextPositionSequences(), mainBodyTextFrame);
+        GapInformation gapInformation = GapDetectionService.findGapsInLines(pageContents.getSortedWords(), mainBodyTextFrame);

        return new PageInformation(pageContents, lineInformation, mainBodyTextFrame, gapInformation);
    }
@@ -5,7 +5,7 @@ import java.util.List;
 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.TextDirection;
-import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;

 import lombok.experimental.UtilityClass;

@@ -17,9 +17,9 @@ public class TextRulingsClassifier {
    private final static double TEXT_BBOX_THRESHOLD_FACTOR = 0.15; // multiplied with text width then subtracted from word width. If ruling covers this width, it is considered as strikethrough/underline.


-    public static void classifyUnderlinedAndStrikethroughText(List<TextPositionSequence> words, CleanRulings cleanRulings) {
+    public static void classifyUnderlinedAndStrikethroughText(List<Word> words, CleanRulings cleanRulings) {

-        for (TextPositionSequence word : words) {
+        for (Word word : words) {
            if (word.getDir().equals(TextDirection.ZERO) || word.getDir().equals(TextDirection.HALF_CIRCLE)) {
                handleHorizontalText(cleanRulings, word);
            } else {
@@ -29,7 +29,7 @@ public class TextRulingsClassifier {
    }


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

        float lowerY = (float) (word.getBBoxPdf().getMinY() + TEXT_BBOX_THRESHOLD_FACTOR * word.getWidth());
        float upperY = (float) (word.getBBoxPdf().getMaxY() - TEXT_BBOX_THRESHOLD_FACTOR * word.getWidth());
@@ -63,7 +63,7 @@ public class TextRulingsClassifier {
    }


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

        float leftX = (float) (word.getBBoxPdf().getMinX() + TEXT_BBOX_THRESHOLD_FACTOR * word.getWidth());
        float rightX = (float) (word.getBBoxPdf().getMaxX() - TEXT_BBOX_THRESHOLD_FACTOR * word.getWidth());
@@ -11,12 +11,14 @@ import org.apache.commons.text.similarity.LevenshteinDistance;
 import org.springframework.stereotype.Service;

 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngineProto.LayoutEngine;
+import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationDocument;
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
 import com.knecon.fforesight.service.layoutparser.processor.model.PageBlockType;
 import com.knecon.fforesight.service.layoutparser.processor.model.SectionIdentifier;
 import com.knecon.fforesight.service.layoutparser.processor.model.outline.OutlineObject;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
-import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;
+import com.knecon.fforesight.service.layoutparser.processor.utils.PageInformation;
 import com.knecon.fforesight.service.layoutparser.processor.utils.TextNormalizationUtilities;

 import lombok.Data;
@@ -27,6 +29,24 @@ public class BlockificationPostprocessingService {
    private static final float STRING_SIMILARITY_THRESHOLD = 0.1f;


+    public void findHeadlinesFromOutline(ClassificationDocument classificationDocument, int pageNumber, ClassificationPage classificationPage, PageInformation pageInformation) {
+
+        OutlineObject lastProcessedOutlineObject = null;
+        List<OutlineObject> outlineObjects = classificationDocument.getOutlineObjectTree().getOutlineObjectsPerPage().getOrDefault(pageNumber, new ArrayList<>());
+
+        OutlineObject notFoundOutlineObject = null;
+        if (lastProcessedOutlineObject != null && !lastProcessedOutlineObject.isFound()) {
+            lastProcessedOutlineObject.resetPoint();
+            notFoundOutlineObject = lastProcessedOutlineObject;
+        }
+        if (!outlineObjects.isEmpty()) {
+            classificationPage.setOutlineObjects(outlineObjects);
+            lastProcessedOutlineObject = sanitizeOutlineBlocks(classificationPage, notFoundOutlineObject);
+        }
+        classificationDocument.getLayoutDebugLayer().addOutlineObjects(outlineObjects, pageInformation);
+    }
+
+
    public OutlineObject sanitizeOutlineBlocks(ClassificationPage classificationPage, OutlineObject notFoundOutlineObject) {

        List<OutlineObject> outlineObjects = classificationPage.getOutlineObjects();
@@ -222,14 +242,14 @@ public class BlockificationPostprocessingService {
            headline = sectionIdentifier + headline;
        }

-        WordSequenceResult wordSequenceResult = findWordSequence(blockToSplit.getSequences(), headline);
+        WordSequenceResult wordSequenceResult = findWordSequence(blockToSplit.getWords(), headline);
        if (wordSequenceResult.inSequence.isEmpty() && !headline.equals(title)) {
-            wordSequenceResult = findWordSequence(blockToSplit.getSequences(), title);
+            wordSequenceResult = findWordSequence(blockToSplit.getWords(), title);
        }

        boolean modifiedBlockToSplit = false;
        if (!wordSequenceResult.inSequence.isEmpty()) {
-            blockToSplit.setSequences(wordSequenceResult.inSequence);
+            blockToSplit.setWords(wordSequenceResult.inSequence);
            blockToSplit.recalculateBBox();
            modifiedBlockToSplit = true;
        }
@@ -250,19 +270,19 @@ public class BlockificationPostprocessingService {
    }


-    private static WordSequenceResult findWordSequence(List<TextPositionSequence> textPositionSequences, String text) {
+    private static WordSequenceResult findWordSequence(List<Word> words, String text) {

        String target = sanitizeString(text);
-        List<TextPositionSequence> inSequence = new ArrayList<>();
-        List<TextPositionSequence> preSequence = new ArrayList<>();
-        List<TextPositionSequence> postSequence = new ArrayList<>();
+        List<Word> inSequence = new ArrayList<>();
+        List<Word> preSequence = new ArrayList<>();
+        List<Word> postSequence = new ArrayList<>();
        StringBuilder currentSequence = new StringBuilder();

        if (target.isBlank()) {
            return new WordSequenceResult();
        }

-        for (TextPositionSequence sequence : textPositionSequences) {
+        for (Word sequence : words) {

            currentSequence.append(sanitizeString(sequence.toString()));
            inSequence.add(sequence);
@@ -274,10 +294,10 @@ public class BlockificationPostprocessingService {
                    int index = 0;
                    String toRemove = currentSequence.substring(0, currentSequence.length() - target.length());

-                    TextPositionSequence next = inSequence.get(index);
+                    Word next = inSequence.get(index);
                    while (currentSequence.length() - next.length() >= target.length()) {

-                        TextPositionSequence removed = inSequence.remove(index);
+                        Word removed = inSequence.remove(index);
                        currentSequence.delete(0, removed.toString().length());
                        preSequence.add(removed);

@@ -306,7 +326,7 @@ public class BlockificationPostprocessingService {
                }

                if (currentSequence.toString().equals(target)) {
-                    postSequence.addAll(textPositionSequences.subList(textPositionSequences.indexOf(sequence) + 1, textPositionSequences.size()));
+                    postSequence.addAll(words.subList(words.indexOf(sequence) + 1, words.size()));
                    return new WordSequenceResult(inSequence, preSequence, postSequence);
                }
            }
@@ -316,10 +336,10 @@ public class BlockificationPostprocessingService {
    }


-    private static SplitSequenceResult splitSequence(TextPositionSequence sequence, String toRemove) {
+    private static SplitSequenceResult splitSequence(Word sequence, String toRemove) {

-        TextPositionSequence in = null;
-        TextPositionSequence out;
+        Word in = null;
+        Word out;

        String currentSequence = sequence.toString().toLowerCase(Locale.ROOT);
        int index = currentSequence.indexOf(toRemove);
@@ -329,17 +349,17 @@ public class BlockificationPostprocessingService {

        if (index > 0) {
            in = createSubSequence(sequence, 0, index);
-        } else if (endIndex < sequence.getTextPositions().size()) {
-            in = createSubSequence(sequence, endIndex, sequence.getTextPositions().size());
+        } else if (endIndex < sequence.length()) {
+            in = createSubSequence(sequence, endIndex, sequence.length());
        }

        return new SplitSequenceResult(in, out);
    }


-    private static TextPositionSequence createSubSequence(TextPositionSequence sequence, int start, int end) {
+    private static Word createSubSequence(Word sequence, int start, int end) {

-        TextPositionSequence newSeq = new TextPositionSequence(new ArrayList<>(sequence.getTextPositions().subList(start, end)), sequence.getPage());
+        Word newSeq = new Word(new ArrayList<>(sequence.getCharacters().subList(start, end)), sequence.getPage());
        newSeq.setParagraphStart(sequence.isParagraphStart());
        return newSeq;
    }
@@ -354,10 +374,10 @@ public class BlockificationPostprocessingService {
            List<TextPageBlock> mergedBlocks = new ArrayList<>();
            for (TextPageBlock textPageBlock : blocksToMerge.subList(1, blocksToMerge.size())) {

-                if (firstBlock != null && !firstBlock.getSequences().isEmpty()) {
+                if (firstBlock != null && !firstBlock.getWords().isEmpty()) {

                    if (textPageBlock.getDir() == firstBlock.getDir()) {
-                        firstBlock.addAll(textPageBlock.getSequences());
+                        firstBlock.addAll(textPageBlock.getWords());
                        mergedBlocks.add(textPageBlock);
                    }
                }
@@ -496,12 +516,12 @@ public class BlockificationPostprocessingService {

    public static class WordSequenceResult {

-        public List<TextPositionSequence> inSequence;
-        public List<TextPositionSequence> preSequence;
-        public List<TextPositionSequence> postSequence;
+        public List<Word> inSequence;
+        public List<Word> preSequence;
+        public List<Word> postSequence;


-        public WordSequenceResult(List<TextPositionSequence> inSequence, List<TextPositionSequence> preSequence, List<TextPositionSequence> postSequence) {
+        public WordSequenceResult(List<Word> inSequence, List<Word> preSequence, List<Word> postSequence) {

            this.inSequence = inSequence;
            this.preSequence = preSequence;
@@ -522,7 +542,7 @@ public class BlockificationPostprocessingService {

    }

-    public record SplitSequenceResult(TextPositionSequence in, TextPositionSequence out) {
+    public record SplitSequenceResult(Word in, Word out) {

    }

@@ -14,7 +14,7 @@ import com.knecon.fforesight.service.layoutparser.processor.model.Classification
 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.model.text.Word;
 import com.knecon.fforesight.service.layoutparser.processor.visualization.LayoutDebugLayer;

 import lombok.RequiredArgsConstructor;
@@ -30,7 +30,7 @@ public class DocstrumBlockificationService {
    static final float THRESHOLD = 1f;


-    public ClassificationPage blockify(List<TextPositionSequence> textPositions,
+    public ClassificationPage blockify(List<Word> textPositions,
                                       CleanRulings rulings,
                                       boolean xyOrder,
                                       LayoutDebugLayer visualizations,
@@ -72,16 +72,16 @@ public class DocstrumBlockificationService {
        List<AbstractPageBlock> abstractPageBlocks = new ArrayList<>();
        zones.forEach(zone -> {

-            List<TextPositionSequence> textPositionSequences = new ArrayList<>();
+            List<Word> words = new ArrayList<>();
            zone.getLines()
                    .forEach(line -> {
                        line.getWords()
                                .forEach(word -> {
-                                    textPositionSequences.add(new TextPositionSequence(word.getTextPositions(), word.getPage()));
+                                    words.add(new Word(word.getCharacters(), word.getPage()));
                                });
                    });

-            abstractPageBlocks.add(buildTextBlock(textPositionSequences, 0));
+            abstractPageBlocks.add(buildTextBlock(words, 0));
        });

        return abstractPageBlocks;
@@ -102,7 +102,7 @@ public class DocstrumBlockificationService {
            }
            TextPageBlock current = (TextPageBlock) block;

-            if (previous != null && !previous.getSequences().isEmpty()) {
+            if (previous != null && !previous.getWords().isEmpty()) {

                if (current.getDir() != previous.getDir() || usedRulings.lineBetween(current, previous)) {
                    previous = current;
@@ -182,8 +182,8 @@ public class DocstrumBlockificationService {

    private TextPageBlock combineBlocksAndResetIterator(TextPageBlock previous, TextPageBlock current, ListIterator<AbstractPageBlock> itty, boolean toDuplicate) {

-        previous.addAll(current.getSequences());
-        previous = buildTextBlock(previous.getSequences(), 0);
+        previous.addAll(current.getWords());
+        previous = buildTextBlock(previous.getWords(), 0);
        previous.setToDuplicate(toDuplicate);
        if (current.getClassification() != null && previous.getClassification() == null) {
            previous.setClassification(current.getClassification());
@@ -283,8 +283,8 @@ public class DocstrumBlockificationService {
                if (current.getDir() == inner.getDir() && current.intersects(inner, yThreshold, xThreshold)) {

                    boolean toDuplicate = current.isToDuplicate() || inner.isToDuplicate();
-                    current.addAll(inner.getSequences());
-                    current = buildTextBlock(current.getSequences(), 0);
+                    current.addAll(inner.getWords());
+                    current = buildTextBlock(current.getWords(), 0);

                    current.setToDuplicate(toDuplicate);
                    blocks.set(i, null);
@@ -301,7 +301,7 @@ public class DocstrumBlockificationService {
    }


-    public static TextPageBlock buildTextBlock(List<TextPositionSequence> wordBlockList, int indexOnPage) {
+    public static TextPageBlock buildTextBlock(List<Word> wordBlockList, int indexOnPage) {

        return new TextPageBlock(wordBlockList);
    }
@@ -16,13 +16,14 @@ 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.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.model.text.Word;

@SuppressWarnings("all")
@Service
 public class DocuMineBlockificationService {

    static final float THRESHOLD = 1f;
+    public static final double FONT_SIZE_CHANGE_RATIO = 0.15;

    Pattern pattern = Pattern.compile("^(\\d{1,2}\\.){1,3}\\d{1,2}\\.?\\s[0-9A-Za-z ()-]{2,50}", Pattern.CASE_INSENSITIVE);

@@ -36,9 +37,9 @@ public class DocuMineBlockificationService {
     * @param cleanRulings  All rulings on a page
     * @return Page object that contains the Textblock and text statistics.
     */
-    public ClassificationPage blockify(List<TextPositionSequence> textPositions, CleanRulings cleanRulings) {
+    public ClassificationPage blockify(List<Word> textPositions, CleanRulings cleanRulings) {

-        List<TextPositionSequence> chunkWords = new ArrayList<>();
+        List<Word> chunkWords = new ArrayList<>();
        List<AbstractPageBlock> textPageBlocks = new ArrayList<>();

        CleanRulings usedRulings = cleanRulings.withoutTextRulings();
@@ -47,11 +48,11 @@ public class DocuMineBlockificationService {
        double maxX = 0;
        double minY = 1000;
        double maxY = 0;
-        TextPositionSequence prev = null;
+        Word prev = null;

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

            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();
@@ -60,11 +61,7 @@ public class DocuMineBlockificationService {
            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")
-                                                    || Math.abs(prev.getFontSize() - word.getFontSize()) >= 1
-                                                    || Math.abs(word.getTextHeight() - prev.getTextHeight()) > 0.8);
+            boolean splitByOtherFontAndOtherY = prev != null && Math.abs(prev.getMaxYDirAdj() - word.getMaxYDirAdj()) > word.getTextHeight() * 0.2 && isFontChange(word, prev);

            Matcher matcher = pattern.matcher(chunkWords.stream()
                                                      .collect(Collectors.joining(" ")).toString());
@@ -127,6 +124,15 @@ public class DocuMineBlockificationService {
    }


+    private static boolean isFontChange(Word word, Word prev) {
+
+        return word.getFontStyle().contains("bold") && !prev.getFontStyle().contains("bold")
+               || prev.getFontStyle().contains("bold") && !word.getFontStyle().contains("bold")
+               || Math.abs(prev.getFontSize() - word.getFontSize()) >= FONT_SIZE_CHANGE_RATIO * Math.min(prev.getFontSize(), word.getFontSize())
+               || Math.abs(word.getTextHeight() - prev.getTextHeight()) >= FONT_SIZE_CHANGE_RATIO * Math.min(prev.getTextHeight(), word.getTextHeight());
+    }
+
+
    public void mergeblocks(ClassificationPage page, CleanRulings usedRulings, float xThreshold, float yThreshold) {

        var blocks = page.getTextBlocks();
@@ -169,8 +175,8 @@ public class DocuMineBlockificationService {
                        .equals(inner.getClassification()))) {

                    boolean toDuplicate = current.isToDuplicate() || inner.isToDuplicate();
-                    current.addAll(inner.getSequences());
-                    current = buildTextBlock(current.getSequences(), 0);
+                    current.addAll(inner.getWords());
+                    current = buildTextBlock(current.getWords(), 0);
                    current.setClassification(inner.getClassification());
                    current.setToDuplicate(toDuplicate);
                    blocks.set(i, null);
@@ -193,7 +199,7 @@ public class DocuMineBlockificationService {
    }


-    public static TextPageBlock buildTextBlock(List<TextPositionSequence> wordBlockList, int indexOnPage) {
+    public static TextPageBlock buildTextBlock(List<Word> wordBlockList, int indexOnPage) {

        return new TextPageBlock(wordBlockList);
    }
@@ -11,7 +11,7 @@ import com.knecon.fforesight.service.layoutparser.processor.model.Classification
 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.TextPageBlock;
-import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;
 import com.knecon.fforesight.service.layoutparser.processor.visualization.LayoutDebugLayer;

@SuppressWarnings("all")
@@ -30,20 +30,20 @@ public class RedactManagerBlockificationService {
     * @param visualizations
     * @return Page object that contains the Textblock and text statistics.
     */
-    public ClassificationPage blockify(List<TextPositionSequence> textPositions, CleanRulings cleanRulings, LayoutDebugLayer visualizations) {
+    public ClassificationPage blockify(List<Word> textPositions, CleanRulings cleanRulings, LayoutDebugLayer visualizations) {

        CleanRulings usedRulings = cleanRulings.withoutTextRulings();

        int indexOnPage = 0;
-        List<TextPositionSequence> chunkWords = new ArrayList<>();
+        List<Word> chunkWords = new ArrayList<>();
        List<AbstractPageBlock> chunkBlockList = new ArrayList<>();

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

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

            boolean lineSeparation = word.getYDirAdj() - maxY > word.getHeight() * 1.25;
            boolean startFromTop = prev != null && word.getYDirAdj() < prev.getYDirAdj() - prev.getTextHeight();
@@ -81,7 +81,7 @@ public class ClarifyndClassificationService {
                   && (textBlock.getMostPopularWordStyle().equals("bold")
                       || !document.getFontStyleCounter().getCountPerValue().containsKey("bold")
                          && textBlock.getMostPopularWordFontSize() > document.getFontSizeCounter().getMostPopular() + 1)
-                   && textBlock.getSequences().get(0).getTextPositions().get(0).getFontSizeInPt() >= textBlock.getMostPopularWordFontSize()) {
+                   && textBlock.getWords().get(0).getFontSize()>= textBlock.getMostPopularWordFontSize()) {

            PageBlockType headlineType = HeadlineClassificationService.headlineClassByFontSize(textBlock, headlineFontSizes);
            headlineClassificationService.classifyHeadline(textBlock, headlineType);
@@ -91,7 +91,7 @@ public class ClarifyndClassificationService {
                   && textBlock.getMostPopularWordStyle().equals("bold")
                   && !document.getFontStyleCounter().getMostPopular().equals("bold")
                   && PositionUtils.getApproxLineCount(textBlock) < 2.9
-                   && textBlock.getSequences().get(0).getTextPositions().get(0).getFontSizeInPt() >= textBlock.getMostPopularWordFontSize()) {
+                   && textBlock.getWords().get(0).getFontSize() >= textBlock.getMostPopularWordFontSize()) {

            PageBlockType headlineType = HeadlineClassificationService.headlineClassByFontSize(textBlock, headlineFontSizes);
            headlineClassificationService.classifyHeadline(textBlock, headlineType);
@@ -15,7 +15,6 @@ import java.util.List;
 import java.util.Locale;
 import java.util.Map;
 import java.util.regex.Matcher;
-import java.util.regex.Pattern;

 import org.springframework.stereotype.Service;

@@ -153,6 +152,16 @@ public class DocuMineClassificationService {
                   && !headlineWithSlashesMatches) {

            setAsHeadline(headlineClassificationService, textBlock, document, headlineFontSizes);
+//        } else if (textBlock.getMostPopularWordFont().contains("bold")
+//                   && greaterOrEqualFontThanPageAverage(textBlock, page)
+//                   && textBlock.getWords().size() <= 6
+//                   && PositionUtils.getApproxLineCount(textBlock) < 2.9
+//                   && isAtLeast3Characters
+//                   && charCount > textBlock.getText().length() * 0.75
+//                   && !textBlock.getText().contains(":")
+//                   && textBlock.getWidth() < page.getBodyTextFrame().getWidth() * 0.7) {
+//
+//            setAsHeadline(headlineClassificationService, textBlock, document, headlineFontSizes);
        } else if (!listIdentifiers.isEmpty()) {

            textBlock.setClassification(PageBlockType.LIST_ITEM);
@@ -11,7 +11,7 @@ import org.springframework.stereotype.Service;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.AbstractBlockOnPage;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.ListIdentifier;
 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.model.text.Word;

@Service
 public class ListItemClassificationService {
@@ -71,7 +71,7 @@ public class ListItemClassificationService {

        List<ListIdentifier> result = new LinkedList<>();
        if (block.block() instanceof TextPageBlock textBlock) {
-            List<TextPositionSequence> sequences = textBlock.getSequences();
+            List<Word> sequences = textBlock.getWords();
            for (int i = 0; i < sequences.size(); i++) {

                if (i != 0 && sequences.get(i - 1).getXDirAdj() < sequences.get(i).getXDirAdj()) {
@@ -79,8 +79,8 @@ public class ListItemClassificationService {
                    continue;
                }

-                TextPositionSequence sequence = sequences.get(i);
-                List<TextPositionSequence> wordsAtStartOfLine = new ArrayList<>(3);
+                Word sequence = sequences.get(i);
+                List<Word> wordsAtStartOfLine = new ArrayList<>(3);
                int end = Math.min(sequences.size(), i + 3);
                for (int j = i; j < end; j++) {
                    if (sequences.get(j).intersectsYDirAdj(sequence, 2)) {
@@ -0,0 +1,7 @@
+package com.knecon.fforesight.service.layoutparser.processor.services.classification;
+
+import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;
+
+public record NumberWord(Word word, int number) {
+
+}
@@ -11,7 +11,6 @@ import com.knecon.fforesight.service.layoutparser.processor.model.Classification
 import com.knecon.fforesight.service.layoutparser.processor.model.PageBlockType;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.ImageType;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
-import com.knecon.fforesight.service.layoutparser.processor.utils.MarkedContentUtils;
 import com.knecon.fforesight.service.layoutparser.processor.utils.PositionUtils;

 import lombok.RequiredArgsConstructor;
@@ -95,7 +94,7 @@ public class RedactManagerClassificationService {
                   && (textBlock.getMostPopularWordStyle().equals("bold")
                       || !document.getFontStyleCounter().getCountPerValue().containsKey("bold")
                          && textBlock.getMostPopularWordFontSize() > document.getFontSizeCounter().getMostPopular() + 1)
-                   && textBlock.getSequences().get(0).getTextPositions().get(0).getFontSizeInPt() >= textBlock.getMostPopularWordFontSize()) {
+                   && textBlock.getWords().get(0).getFontSize() >= textBlock.getMostPopularWordFontSize()) {

            PageBlockType headlineType = HeadlineClassificationService.headlineClassByFontSize(textBlock, headlineFontSizes);
            headlineClassificationService.classifyHeadline(textBlock, headlineType);
@@ -105,7 +104,7 @@ public class RedactManagerClassificationService {
                   && textBlock.getMostPopularWordStyle().equals("bold")
                   && !document.getFontStyleCounter().getMostPopular().equals("bold")
                   && PositionUtils.getApproxLineCount(textBlock) < 2.9
-                   && textBlock.getSequences().get(0).getTextPositions().get(0).getFontSizeInPt() >= textBlock.getMostPopularWordFontSize()) {
+                   && textBlock.getWords().get(0).getFontSize() >= textBlock.getMostPopularWordFontSize()) {

            PageBlockType headlineType = HeadlineClassificationService.headlineClassByFontSize(textBlock, headlineFontSizes);
            headlineClassificationService.classifyHeadline(textBlock, headlineType);
@@ -14,6 +14,7 @@ import java.util.List;
 import java.util.Locale;
 import java.util.Map;
 import java.util.Set;
+import java.util.regex.Matcher;
 import java.util.stream.Collectors;

 import org.springframework.stereotype.Service;
@@ -23,10 +24,11 @@ import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBl
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationDocument;
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
 import com.knecon.fforesight.service.layoutparser.processor.model.PageBlockType;
+import com.knecon.fforesight.service.layoutparser.processor.model.SectionIdentifier;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextBlockOnPage;
 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.model.text.TextPositionSequenceComparator;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.TocNumberComparator;
 import com.knecon.fforesight.service.layoutparser.processor.utils.TextNormalizationUtilities;
 import com.knecon.fforesight.service.layoutparser.processor.visualization.LayoutDebugLayer;

@@ -69,11 +71,14 @@ public class TableOfContentsClassificationService {

    private int identifyTOCItems(int start, List<TextBlockOnPage> textBlocks, ClassificationDocument document) {

+        if (start >= textBlocks.size()) {
+            return start;
+        }
        ClassificationPage startPage = textBlocks.get(start).page();
        List<TextBlockOnPage> initialLookAhead = textBlocks.subList(start, Math.min(start + SURROUNDING_BLOCKS_RADIUS, textBlocks.size()));
-        HashMap<TextPositionSequence, TextBlockOnPage> lookup = new HashMap<>();
-        List<TextPositionSequence> numbers = extractNumbers(initialLookAhead, lookup, document.getPages().size());
-        TocNumberFinder tocNumberFinder = new TocNumberFinder(numbers, lookup);
+        HashMap<NumberWord, TextBlockOnPage> numberToBlockLookup = new HashMap<>();
+        List<NumberWord> numbers = extractNumbers(initialLookAhead, numberToBlockLookup, document.getPages().size());
+        TocNumberFinder tocNumberFinder = new TocNumberFinder(numbers, numberToBlockLookup);

        int lastCandidate = start;
        for (int i = start; i < Math.min(lastCandidate + SURROUNDING_BLOCKS_RADIUS, textBlocks.size()); i++) {
@@ -93,28 +98,28 @@ public class TableOfContentsClassificationService {
                break;
            }

-            List<TextPositionSequence> numbersFromBlock = extractNumbers(textBlockOnPage, lookup, document.getPages().size());
+            List<NumberWord> numbersFromBlock = extractNumbers(textBlockOnPage, numberToBlockLookup, document.getPages().size());

-            List<TextPositionSequence> currentRightmostCluster = tocNumberFinder.getCurrentRightmostCluster();
+            List<NumberWord> currentRightmostCluster = tocNumberFinder.getCurrentRightmostCluster();

            if (currentRightmostCluster.size() < MINIMUM_MATCHES) {
                log.debug("No numbers indicating a table of contents here.");
                return start;
            }

-            if (anyIntersection(currentRightmostCluster, numbersFromBlock, lookup)) {
+            if (anyIntersection(currentRightmostCluster, numbersFromBlock, numberToBlockLookup)) {
                lastCandidate = i;
                numbersFromBlock.forEach(tocNumberFinder::add);
            }
        }

-        addVisualization(document.getLayoutDebugLayer(), tocNumberFinder, lookup);
-
        Set<TextBlockOnPage> blocksWithNumberInCluster = tocNumberFinder.getCurrentRightmostCluster()
                .stream()
-                .map(lookup::get)
+                .map(numberToBlockLookup::get)
                .collect(Collectors.toSet());

+        addVisualization(document.getLayoutDebugLayer(), tocNumberFinder, numberToBlockLookup, blocksWithNumberInCluster, textBlocks.get(start - 1));
+
        int lastConfirmed = start;
        for (int i = start; i < lastCandidate + 1; i++) {
            TextBlockOnPage textBlockOnPage = textBlocks.get(i);
@@ -132,18 +137,22 @@ public class TableOfContentsClassificationService {
    }


-    private static void addVisualization(LayoutDebugLayer layoutDebugLayer, TocNumberFinder tocNumberFinder, Map<TextPositionSequence, TextBlockOnPage> lookup) {
+    private static void addVisualization(LayoutDebugLayer layoutDebugLayer,
+                                         TocNumberFinder tocNumberFinder,
+                                         Map<NumberWord, TextBlockOnPage> lookup,
+                                         Set<TextBlockOnPage> blocksWithNumberInCluster,
+                                         TextBlockOnPage startingHeadline) {

        tocNumberFinder.getCurrentRightmostCluster()
                .stream()
                .collect(Collectors.groupingBy(key -> lookup.get(key).page().getPageNumber()))
                .forEach((pageNumber, number) -> layoutDebugLayer.addTocPages(number, pageNumber));
+        layoutDebugLayer.addTocBlocks(blocksWithNumberInCluster);
+        layoutDebugLayer.addTocBlocks(Set.of(startingHeadline));
    }


-    private static boolean anyIntersection(Collection<TextPositionSequence> numbers1,
-                                           Collection<TextPositionSequence> numbers2,
-                                           Map<TextPositionSequence, TextBlockOnPage> lookup) {
+    private static boolean anyIntersection(Collection<NumberWord> numbers1, Collection<NumberWord> numbers2, Map<NumberWord, TextBlockOnPage> lookup) {

        return numbers1.stream()
                .anyMatch(numberFromCluster -> numbers2.stream()
@@ -151,9 +160,9 @@ public class TableOfContentsClassificationService {
    }


-    private static List<TextPositionSequence> extractNumbers(List<TextBlockOnPage> textBlocks, Map<TextPositionSequence, TextBlockOnPage> lookup, int numberOfPages) {
+    private static List<NumberWord> extractNumbers(List<TextBlockOnPage> textBlocks, Map<NumberWord, TextBlockOnPage> lookup, int numberOfPages) {

-        List<TextPositionSequence> blocks = new LinkedList<>();
+        List<NumberWord> blocks = new LinkedList<>();
        for (TextBlockOnPage textBlock : textBlocks) {
            blocks.addAll(extractNumbers(textBlock, lookup, numberOfPages));
        }
@@ -161,30 +170,40 @@ public class TableOfContentsClassificationService {
    }


-    private static List<TextPositionSequence> extractNumbers(TextBlockOnPage textBlock, Map<TextPositionSequence, TextBlockOnPage> lookup, int numberOfPages) {
+    private static List<NumberWord> extractNumbers(TextBlockOnPage textBlock, Map<NumberWord, TextBlockOnPage> lookup, int numberOfPages) {

-        List<TextPositionSequence> blocks = new LinkedList<>();
+        List<NumberWord> blocks = new LinkedList<>();
        TextPageBlock block = textBlock.textBlock();
-        List<TextPositionSequence> sequences = block.getSequences();
-        for (int i = 0; i < sequences.size(); i++) {
+        List<Word> words = block.getWords();
+        for (int i = 0; i < words.size(); i++) {

-            TextPositionSequence word = sequences.get(i);
-
-            if (!NUMERIC.matcher(word).matches()) {
+            Word word = words.get(i);
+            if (!wordIsEndOfLine(i, words)) {
                continue;
            }

-            if (AMOUNT_PATTERN.matcher(getSurroundingString(i, sequences)).matches()) {
+            if (AMOUNT_PATTERN.matcher(getSurroundingString(i, words)).matches()) {
+                continue;
+            }
+
+            Matcher matcher = SectionIdentifier.numericalIdentifierPattern.matcher(word.toString());
+            if (matcher.find() && matcher.group(2) != null) {
+                continue;
+            }
+
+            Matcher numberFinder = NUMERIC.matcher(word);
+            if (!numberFinder.find() || word.length() > 5) {
                continue;
            }

            try {
-                int pageNumber = Integer.parseInt(word.toString());
+                int pageNumber = Integer.parseInt(numberFinder.group());
                if (0 >= pageNumber || pageNumber > numberOfPages) {
                    continue;
                }
-                lookup.put(word, textBlock);
-                blocks.add(word);
+                NumberWord numberWord = new NumberWord(word, pageNumber);
+                lookup.put(numberWord, textBlock);
+                blocks.add(numberWord);
            } catch (NumberFormatException e) {
                log.debug("That wasn't a number! Should not happen, due to numeric check beforehand.");
            }
@@ -193,23 +212,34 @@ public class TableOfContentsClassificationService {
    }


-    private static CharSequence getSurroundingString(int i, List<TextPositionSequence> sequences) {
+    private static boolean wordIsEndOfLine(int i, List<Word> words) {
+
+        if (i == words.size() - 1) {
+            return true;
+        }
+        Word word = words.get(i);
+        Word nextWord = words.get(i + 1);
+        return !nextWord.rightOf(word);
+    }
+
+
+    private static CharSequence getSurroundingString(int i, List<Word> sequences) {

        int end = Math.min(i + 5, sequences.size());
        return sequences.subList(i, end)
                .stream()
-                .map(TextPositionSequence::toString)
+                .map(Word::toString)
                .collect(Collectors.joining(" "));
    }


-    private static boolean matches(TextPositionSequence number1, TextPositionSequence number2, Map<TextPositionSequence, TextBlockOnPage> lookup) {
+    private static boolean matches(NumberWord number1, NumberWord number2, Map<NumberWord, TextBlockOnPage> lookup) {

-        if (number1.getDir() != number2.getDir()) {
+        if (number1.word().getDir() != number2.word().getDir()) {
            return false;
        }

-        return number1.intersectsXDirAdj(number2, INTERSECTION_TOLERANCE);
+        return number1.word().intersectsXDirAdj(number2.word(), INTERSECTION_TOLERANCE);
    }


@@ -247,11 +277,11 @@ public class TableOfContentsClassificationService {

    private static class TocNumberFinder {

-        final UnionFind<TextPositionSequence> numberClusters;
-        final HashMap<TextPositionSequence, TextBlockOnPage> lookup;
+        final UnionFind<NumberWord> numberClusters;
+        final HashMap<NumberWord, TextBlockOnPage> lookup;


-        TocNumberFinder(List<TextPositionSequence> blocks, HashMap<TextPositionSequence, TextBlockOnPage> lookup) {
+        TocNumberFinder(List<NumberWord> blocks, HashMap<NumberWord, TextBlockOnPage> lookup) {

            this.numberClusters = new UnionFind<>(new HashSet<>(blocks));
            for (int i = 0; i < blocks.size(); i++) {
@@ -265,14 +295,14 @@ public class TableOfContentsClassificationService {
        }


-        public void add(TextPositionSequence number) {
+        public void add(NumberWord number) {

            if (numberClusters.getElements().contains(number)) {
                return;
            }

            numberClusters.addElement(number);
-            for (TextPositionSequence element : numberClusters.getElements()) {
+            for (NumberWord element : numberClusters.getElements()) {
                if (matches(number, element, lookup)) {
                    numberClusters.union(element, number);
                }
@@ -280,73 +310,100 @@ public class TableOfContentsClassificationService {
        }


-        public List<TextPositionSequence> getCurrentRightmostCluster() {
+        public List<NumberWord> getCurrentRightmostCluster() {

            return numberClusters.getGroups()
                    .stream()
                    .filter(cluster -> cluster.size() > MINIMUM_MATCHES)
                    .map(cluster -> cluster.stream()
-                            .sorted(new TextPositionSequenceComparator(lookup))
+                            .sorted(new TocNumberComparator(lookup))
                            .toList())
                    .map(this::removeOutliers)
-//                    .map(this::filterByMinimumDensity)
+                    .map(this::removeOnNonConsecutivePages)
+                    .map(this::filterByWordNearTopOfPage)
                    .filter(cluster -> cluster.size() > MINIMUM_MATCHES)
-                    .max(Comparator.comparingDouble(cluster -> cluster.get(0).getBBox().getMaxX())).orElse(Collections.emptyList());
+                    .max(Comparator.comparingDouble(cluster -> cluster.get(0).word().getBBox().getMaxX())).orElse(Collections.emptyList());
        }

-// does not seem to be doing much, ideally instead of using the height of the blocks i would like to use the height, beginning from the MainBody top,
-// but as the MainBody is often wrong, this results in some numbers being discarded even though they are correct.
-//        private List<TextPositionSequence> filterByMinimumDensity(List<TextPositionSequence> numbers) {
-//
-//            Map<ClassificationPage, List<TextPositionSequence>> clustersPerPage = numbers.stream()
-//                    .collect(Collectors.groupingBy(number -> lookup.get(number).page()));
-//
-//            List<TextPositionSequence> result = new ArrayList<>(numbers.size());
-//            clustersPerPage.keySet()
-//                    .stream()
-//                    .sorted(Comparator.comparingInt(ClassificationPage::getPageNumber))
-//                    .forEach(page -> {
-//                        var numbersOnPage = clustersPerPage.get(page);
-//
-//                        double height = numbersOnPage.stream()
-//                                .map(BoundingBox::getBBox)
-//                                .collect(RectangleTransformations.collectBBox()).getHeight();
-//
-//                        double count = numbersOnPage.size();
-//
-//                        if ((count / height) >= (DENSITY_THRESHOLD_COUNT / page.getPageHeight())) {
-//                            result.addAll(numbers);
-//                        }
-//                    });
-//            return result;
-//        }

+        private List<NumberWord> removeOnNonConsecutivePages(List<NumberWord> numbers) {

-        public List<TextPositionSequence> removeOutliers(List<TextPositionSequence> numbers) {
-
-            List<TextPositionSequence> result = new ArrayList<>();
+            List<NumberWord> result = new ArrayList<>();

            result.add(numbers.get(0));

+            for (int i = 1; i < numbers.size(); i++) {
+                int prev = getPageNumber(numbers, i - 1);
+                int curr = getPageNumber(numbers, i);
+
+                if (Math.abs(prev - curr) > 1) {
+                    break;
+                } else {
+                    result.add(numbers.get(i));
+                }
+            }
+            return result;
+        }
+
+
+        private int getPageNumber(List<NumberWord> numbers, int i) {
+
+            return lookup.get(numbers.get(i)).page().getPageNumber();
+        }
+
+
+        private List<NumberWord> filterByWordNearTopOfPage(List<NumberWord> numbers) {
+
+            List<NumberWord> result = new ArrayList<>();
+
+            result.add(numbers.get(0));
+
+            for (int i = 1; i < numbers.size(); i++) {
+                NumberWord prev = numbers.get(i - 1);
+                NumberWord curr = numbers.get(i);
+                ClassificationPage prevPage = lookup.get(prev).page();
+                ClassificationPage currPage = lookup.get(curr).page();
+                if (prevPage.equals(currPage)) {
+                    result.add(curr);
+                } else if (curr.word().getBBox().getMinY() < currPage.getPageHeight() * 0.33) {
+                    result.add(curr);
+                }
+            }
+            return result;
+        }
+
+
+        public List<NumberWord> removeOutliers(List<NumberWord> numbers) {
+
+            List<NumberWord> confirmedClusterNumbers = new ArrayList<>();
+
+            confirmedClusterNumbers.add(numbers.get(0));
+
            for (int i = 1; i < numbers.size() - 1; i++) {
                int prev = getNumberAsInt(numbers, i - 1);
                int curr = getNumberAsInt(numbers, i);
                int next = getNumberAsInt(numbers, i + 1);

                if (!(curr <= prev || curr >= next) || !isBetterWithout(numbers, i)) {
-                    result.add(numbers.get(i));
+                    confirmedClusterNumbers.add(numbers.get(i));
                }
            }
-            if (getNumberAsInt(numbers, numbers.size() - 1) >= getNumberAsInt(numbers, Math.max(0, numbers.size() - 2))) {
-                result.add(numbers.get(numbers.size() - 1));
+            if (getNumberAsInt(numbers, numbers.size() - 1) >= getLatestNumber(confirmedClusterNumbers)) {
+                confirmedClusterNumbers.add(numbers.get(numbers.size() - 1));
            }

-            return result;
+            return confirmedClusterNumbers;
+        }
+
+
+        private static int getLatestNumber(List<NumberWord> confirmedClusterNumbers) {
+
+            return confirmedClusterNumbers.get(confirmedClusterNumbers.size() - 1).number();
        }


        // Helper method to check if removing the current number results in a better order
-        public static boolean isBetterWithout(List<TextPositionSequence> numbers, int i) {
+        public static boolean isBetterWithout(List<NumberWord> numbers, int i) {

            if (i == 0 || i == numbers.size() - 1) {
                return false;
@@ -362,9 +419,9 @@ public class TableOfContentsClassificationService {
    }


-    private static int getNumberAsInt(List<TextPositionSequence> numbers, int i) {
+    private static int getNumberAsInt(List<NumberWord> numbers, int i) {

-        return Integer.parseInt(numbers.get(i).toString());
+        return numbers.get(i).number();
    }

 }
@@ -16,7 +16,7 @@ import java.util.Optional;
 import java.util.Set;
 import java.util.stream.Collectors;

-import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
+import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeTypeProto;
 import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingType;
 import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationDocument;
@@ -82,9 +82,9 @@ public class DocumentGraphFactory {

        documentGraph.streamAllSubNodes()
                .filter(SemanticNode::isLeaf)
-                .filter(node -> !node.getType().equals(NodeType.HEADER))
-                .filter(node -> !node.getType().equals(NodeType.FOOTER))
-                .filter(node -> !node.getType().equals(NodeType.IMAGE))
+                .filter(node -> !node.getType().equals(NodeTypeProto.NodeType.HEADER))
+                .filter(node -> !node.getType().equals(NodeTypeProto.NodeType.FOOTER))
+                .filter(node -> !node.getType().equals(NodeTypeProto.NodeType.IMAGE))
                .map(SemanticNode::getTextBlock)
                .map(TextBlock::getAtomicTextBlocks)
                .flatMap(Collection::stream)
@@ -97,12 +97,12 @@ public class DocumentGraphFactory {
        for (TableOfContentItem tocItem : classificationDocument.getTableOfContents()) {
            GenericSemanticNode parent = tocItem.getParent() == null ? null : tocItem.getParent().getSection();
            Optional<GenericSemanticNode> section = SectionNodeFactory.addSection(layoutParsingType,
-                                                                                   parent,
-                                                                                   tocItem.getChildren().isEmpty(),
-                                                                                   tocItem.getNonEmptySectionBlocks(),
-                                                                                   tocItem.getImages(),
-                                                                                   context,
-                                                                                   document);
+                                                                                  parent,
+                                                                                  tocItem.getChildren().isEmpty(),
+                                                                                  tocItem.getNonEmptySectionBlocks(),
+                                                                                  tocItem.getImages(),
+                                                                                  context,
+                                                                                  document);
            tocItem.setSection(section.orElse(null));
        }
    }
@@ -133,9 +133,9 @@ public class DocumentGraphFactory {

        if (node instanceof DuplicatedParagraph duplicatedParagraph) {
            AtomicTextBlock unsortedTextBlock = context.textBlockFactory.buildAtomicTextBlock(textBlocks.stream()
-                                                                                                       .flatMap(tb -> tb.getSequences()
-                                                                                                               .stream())
-                                                                                                       .collect(Collectors.toList()), node, context, page);
+                                                                                                      .flatMap(tb -> tb.getWords()
+                                                                                                              .stream())
+                                                                                                      .collect(Collectors.toList()), node, context, page);
            duplicatedParagraph.setUnsortedLeafTextBlock(unsortedTextBlock);
        }

@@ -10,9 +10,10 @@ import java.util.List;
 import java.util.Locale;
 import java.util.Objects;

+import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Character;
 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.TextPositionSequence;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;

 import lombok.experimental.UtilityClass;

@@ -28,22 +29,22 @@ public class SearchTextWithTextPositionFactory {
    public static final double LINEBREAK_DELTA_TOLERANCE = 1.5;


-    public SearchTextWithTextPositionDto buildSearchTextToTextPositionDto(List<TextPositionSequence> sequences) {
+    public SearchTextWithTextPositionDto buildSearchTextToTextPositionDto(List<Word> sequences) {

        if (sequences.isEmpty() || sequences.stream()
-                .allMatch(sequence -> sequence.getTextPositions().isEmpty())) {
+                .allMatch(sequence -> sequence.getCharacters().isEmpty())) {
            return SearchTextWithTextPositionDto.empty();
        }

        Context context = new Context();

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

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

-                currentTextPosition = word.getTextPositions().get(i);
+                currentTextPosition = word.getCharacters().get(i).getTextPosition();
                if (isLineBreak(currentTextPosition, previousTextPosition)) {
                    removeHyphenLinebreaks(context);
                    context.lineBreaksStringIdx.add(context.stringIdx);
@@ -66,8 +67,9 @@ public class SearchTextWithTextPositionFactory {
        }

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

@@ -16,7 +16,7 @@ import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Ta
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.Cell;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.TablePageBlock;
-import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;
 import com.knecon.fforesight.service.layoutparser.processor.utils.TextPositionOperations;

 import lombok.experimental.UtilityClass;
@@ -115,7 +115,7 @@ public class TableNodeFactory {
        if (cell.getTextBlocks().isEmpty()) {
            tableCell.setLeafTextBlock(context.getTextBlockFactory().emptyTextBlock(tableNode, context, page));
        } else if (cell.getTextBlocks().size() == 1) {
-            textBlock = context.getTextBlockFactory().buildAtomicTextBlock(cell.getTextBlocks().get(0).getSequences(), tableCell, context, page);
+            textBlock = context.getTextBlockFactory().buildAtomicTextBlock(cell.getTextBlocks().get(0).getWords(), tableCell, context, page);
            tableCell.setLeafTextBlock(textBlock);
        } else if (firstTextBlockIsHeadline(cell)) {
            SectionNodeFactory.addSection(layoutParsingType,
@@ -129,7 +129,7 @@ public class TableNodeFactory {
                                          context,
                                          document);
        } else if (cellAreaIsSmallerThanPageAreaTimesThreshold(cell, page)) {
-            List<TextPositionSequence> sequences = TextPositionOperations.mergeAndSort(cell.getTextBlocks());
+            List<Word> sequences = TextPositionOperations.mergeAndSort(cell.getTextBlocks());
            textBlock = context.getTextBlockFactory().buildAtomicTextBlock(sequences, tableCell, context, page);
            tableCell.setLeafTextBlock(textBlock);
        } else {
@@ -2,7 +2,7 @@ package com.knecon.fforesight.service.layoutparser.processor.services.factory;

 import java.util.List;

-import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;
 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.textblock.AtomicTextBlock;
@@ -17,14 +17,14 @@ public class TextBlockFactory {
    long textBlockIdx;


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

        Integer numberOnPage = context.getAndIncrementTextBlockNumberOnPage(page);
        return buildAtomicTextBlock(sequences, parent, numberOnPage, page);
    }


-    public AtomicTextBlock buildAtomicTextBlock(List<TextPositionSequence> sequences, SemanticNode parent, Integer numberOnPage, Page page) {
+    public AtomicTextBlock buildAtomicTextBlock(List<Word> sequences, SemanticNode parent, Integer numberOnPage, Page page) {

        SearchTextWithTextPositionDto searchTextWithTextPositionDto = SearchTextWithTextPositionFactory.buildSearchTextToTextPositionDto(sequences);
        int offset = stringOffset;
@@ -11,7 +11,7 @@ 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.model.text.Word;
 import com.knecon.fforesight.service.layoutparser.processor.utils.PageInformation;

 import lombok.RequiredArgsConstructor;
@@ -33,10 +33,10 @@ public class GraphicExtractorService {
                                                PDPage pdPage,
                                                int pageNumber,
                                                CleanRulings cleanRulings,
-                                                List<TextPositionSequence> textPositionSequences,
+                                                List<Word> words,
                                                boolean graphicsRaster) {

-        List<Box> characterBBoxes = getCharacterBBoxes(textPositionSequences);
+        List<Box> characterBBoxes = getCharacterBBoxes(words);
        List<Box> classifiedRulingsBoxes = getLineBBoxesOfAllClassifiedRulings(cleanRulings);

        GraphicBBDetector graphicBBDetector = new GraphicBBDetector(pdPage, true);
@@ -63,9 +63,9 @@ public class GraphicExtractorService {
    }


-    private List<Box> getCharacterBBoxes(List<TextPositionSequence> textPositionSequences) {
+    private List<Box> getCharacterBBoxes(List<Word> words) {

-        return textPositionSequences.stream()
+        return words.stream()
                .map(BoundingBox::getBBoxPdf)
                .map(Box::new)
                .collect(Collectors.toList());
@@ -40,7 +40,7 @@ import org.apache.pdfbox.text.TextPosition;

 import com.knecon.fforesight.service.layoutparser.processor.model.table.Ruling;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.RedTextPosition;
-import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;

 import lombok.Getter;
 import lombok.Setter;
@@ -52,7 +52,7 @@ import lombok.extern.slf4j.Slf4j;
 public class PDFLinesTextStripper extends PDFTextStripper {

    private final static Set<String> DOT_LIKE_CHARACTERS = Set.of(".", "·", "•", "․", "‧", "∙", "⋅", "・", "．", "･", "…", "⸱", "﹒", "ꞏ");
-    private final List<TextPositionSequence> textPositionSequences = new ArrayList<>();
+    private final List<Word> words = new ArrayList<>();
    private final List<Ruling> rulings = new ArrayList<>();
    private final List<Ruling> graphicsPath = new ArrayList<>();
    @Setter
@@ -246,10 +246,10 @@ public class PDFLinesTextStripper extends PDFTextStripper {
                direction = textPositions.get(i).getDir();
            }

-            if (!textPositionSequences.isEmpty()) {
-                previous = textPositionSequences.get(textPositionSequences.size() - 1)
-                        .getTextPositions()
-                        .get(textPositionSequences.get(textPositionSequences.size() - 1).getTextPositions().size() - 1);
+            if (!words.isEmpty()) {
+                previous = words.get(words.size() - 1)
+                        .getCharacters()
+                        .get(words.get(words.size() - 1).getCharacters().size() - 1).getTextPosition();
            }

            if (i == 0 && (textPositions.get(i).getUnicode().equals(" ") || textPositions.get(i).getUnicode().equals("\u00A0") || textPositions.get(i).getUnicode().equals("\t"))) {
@@ -259,7 +259,7 @@ public class PDFLinesTextStripper extends PDFTextStripper {

            if (textPositions.get(i).getDir() != direction && startIndex != i) {
                List<TextPosition> sublist = textPositions.subList(startIndex, i);
-                textPositionSequences.add(new TextPositionSequence(sublist, pageNumber, i == textPositions.size() - 1 && isParagraphStart));
+                words.add(new Word(sublist, pageNumber, i == textPositions.size() - 1 && isParagraphStart));
                startIndex = i;
                direction = textPositions.get(i).getDir();
            }
@@ -268,7 +268,7 @@ public class PDFLinesTextStripper extends PDFTextStripper {
            if (checkIfCurrentPositionIsToTheRightOfPreviousPosition(i, textPositions)) {
                List<TextPosition> sublist = textPositions.subList(startIndex, i);
                if (checkIfSequenceContainsOnlyWhitespaces(sublist)) {
-                    textPositionSequences.add(new TextPositionSequence(sublist, pageNumber, i == textPositions.size() - 1 && isParagraphStart));
+                    words.add(new Word(sublist, pageNumber, i == textPositions.size() - 1 && isParagraphStart));
                }
                startIndex = i;
            }
@@ -276,7 +276,7 @@ public class PDFLinesTextStripper extends PDFTextStripper {
            if (textPositions.get(i).getRotation() == 0 && i > 0 && textPositions.get(i).getX() > textPositions.get(i - 1).getEndX() + 1) {
                List<TextPosition> sublist = textPositions.subList(startIndex, i);
                if (checkIfSequenceContainsOnlyWhitespaces(sublist)) {
-                    textPositionSequences.add(new TextPositionSequence(sublist, pageNumber, i == textPositions.size() - 1 && isParagraphStart));
+                    words.add(new Word(sublist, pageNumber, i == textPositions.size() - 1 && isParagraphStart));
                }
                startIndex = i;
            }
@@ -290,22 +290,22 @@ public class PDFLinesTextStripper extends PDFTextStripper {
                    // Remove false sequence ends (whitespaces)
                    if (checkIfGapSizeBetweenCharactersSmallerThanMaximum(previous, sublist, 0.01f)) {
                        for (TextPosition t : sublist) {
-                            textPositionSequences.get(textPositionSequences.size() - 1).add(t);
+                            words.get(words.size() - 1).add(t);
                        }
                    } else {
-                        textPositionSequences.add(new TextPositionSequence(sublist, pageNumber, i == textPositions.size() - 1 && isParagraphStart));
+                        words.add(new Word(sublist, pageNumber, i == textPositions.size() - 1 && isParagraphStart));
                    }
                }
                startIndex = i + 1;
            }
            if (isDottedLineFollowedByWord(textPositions, i, startIndex)) {
                List<TextPosition> sublist = textPositions.subList(startIndex, i);
-                textPositionSequences.add(new TextPositionSequence(sublist, pageNumber, i == textPositions.size() - 1 && isParagraphStart));
+                words.add(new Word(sublist, pageNumber, i == textPositions.size() - 1 && isParagraphStart));
                startIndex = i;
            }
            if (isWordFollowedByDottedLine(textPositions, i, startIndex)) {
                List<TextPosition> sublist = textPositions.subList(startIndex, i - 2);
-                textPositionSequences.add(new TextPositionSequence(sublist, pageNumber, i == textPositions.size() - 1 && isParagraphStart));
+                words.add(new Word(sublist, pageNumber, i == textPositions.size() - 1 && isParagraphStart));
                startIndex = i - 2;
            }
        }
@@ -324,10 +324,10 @@ public class PDFLinesTextStripper extends PDFTextStripper {
                && sublist.get(0).getYDirAdj() == previous.getYDirAdj()
                && sublist.get(0).getXDirAdj() - (previous.getXDirAdj() + previous.getWidthDirAdj()) < 0.01) {
                for (TextPosition t : sublist) {
-                    textPositionSequences.get(textPositionSequences.size() - 1).add(t);
+                    words.get(words.size() - 1).add(t);
                }
            } else {
-                textPositionSequences.add(new TextPositionSequence(sublist, pageNumber, isParagraphStart));
+                words.add(new Word(sublist, pageNumber, isParagraphStart));
            }
        }

@@ -392,7 +392,7 @@ public class PDFLinesTextStripper extends PDFTextStripper {
    @Override
    public String getText(PDDocument doc) throws IOException {

-        textPositionSequences.clear();
+        words.clear();
        rulings.clear();
        graphicsPath.clear();
        path_x = 0.0f;
@@ -13,7 +13,7 @@ import org.apache.pdfbox.text.TextPosition;

 import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.BoundingBox;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
-import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;

 import lombok.experimental.UtilityClass;

@@ -48,7 +48,7 @@ public class MarkedContentUtils {

        return markedContentByYPosition.values()
                .stream()
-                .map(textPositions -> new TextPositionSequence(textPositions, 0, true).getBBoxPdf())
+                .map(textPositions -> new Word(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());
    }
@@ -89,7 +89,7 @@ public class MarkedContentUtils {
                    .filter(content -> content instanceof TextPosition)
                    .map(content -> (TextPosition) content)
                    .filter(content -> !content.getUnicode().equals(" "))
-                    .map(textPositions -> new TextPositionSequence(List.of(textPositions), 0, true))
+                    .map(textPositions -> new Word(List.of(textPositions), 0, true))
                    .map(BoundingBox::getBBoxPdf)
                    .collect(Collectors.toList());
        }
@@ -11,12 +11,13 @@ 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.Line;
 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 com.knecon.fforesight.service.layoutparser.processor.model.text.Word;

 import lombok.experimental.UtilityClass;

@@ -36,33 +37,42 @@ public class TextPositionOperations {
                    .thenComparing(TextBoundingBox::getXDirAdj, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, X_THRESHOLD));


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

        var sequences = textBlocks.stream()
-                .flatMap(tb -> tb.getSequences()
+                .flatMap(tb -> tb.getWords()
                        .stream())
                .collect(Collectors.toSet());
        return sortUsingLineDetection(sequences);
    }


-    public List<TextPositionSequence> sort(List<TextPositionSequence> sequences) {
+    public List<Word> sort(List<Word> sequences) {

        return sortUsingLineDetection(new HashSet<>(sequences));
    }


-    private List<TextPositionSequence> sortUsingLineDetection(Set<TextPositionSequence> sequences) {
+    private List<Word> sortUsingLineDetection(Set<Word> sequences) {

-        return sortLines(groupByLine(sequences));
+        return sortWords(groupByLine(sequences));
    }


-    public List<TextPositionSequence> sortLines(Collection<Set<TextPositionSequence>> lines) {
+    public List<Word> sortWords(Collection<Set<Word>> lines) {

-        List<List<TextPositionSequence>> lineBlocks = new ArrayList<>();
-        for (Set<TextPositionSequence> line : lines) {
-            List<TextPositionSequence> sortedLine = sortByXDirAdj(line);
+        return sortLines(lines).stream()
+                .map(Line::getWords)
+                .flatMap(Collection::stream)
+                .collect(Collectors.toList());
+    }
+
+
+    public List<Line> sortLines(Collection<Set<Word>> lines) {
+
+        List<List<Word>> lineBlocks = new ArrayList<>();
+        for (Set<Word> line : lines) {
+            List<Word> sortedLine = sortByXDirAdj(line);
            if (!sortedLine.isEmpty()) {
                lineBlocks.add(sortedLine);
            }
@@ -70,35 +80,41 @@ public class TextPositionOperations {
        // need to use old sorting, since COMPARATOR_DIR_ADJ is not transitive
        QuickSort.sort(lineBlocks, Comparator.comparing(line -> line.get(0), COMPARATOR_DIR_ADJ));

-        List<TextPositionSequence> list = new ArrayList<>();
-        for (List<TextPositionSequence> textPositionSequences : lineBlocks) {
-            list.addAll(textPositionSequences);
+        List<Line> list = new ArrayList<>();
+        for (List<Word> lineBlock : lineBlocks) {
+            list.add(new Line(lineBlock));
        }
        return list;
    }


-    private List<TextPositionSequence> sortByXDirAdj(Set<TextPositionSequence> line) {
+    private List<Word> sortByXDirAdj(Set<Word> line) {

        return line.stream()
-                .sorted(Comparator.comparing(TextPositionSequence::getXDirAdj))
+                .sorted(Comparator.comparing(Word::getXDirAdj))
                .toList();
    }


-    public Collection<Set<TextPositionSequence>> groupByLine(Set<TextPositionSequence> sequences) {
+    public Collection<Set<Word>> groupByLine(Set<Word> sequences) {

        double maxLineDistance = sequences.stream()
-                                         .map(TextPositionSequence::getBBoxDirAdj)
+                                         .map(Word::getBBoxDirAdj)
                                         .mapToDouble(RectangularShape::getHeight).average().orElse(10) * MAX_LINE_HEIGHT_FACTOR;
        double maxXGap = sequences.stream()
-                                 .map(TextPositionSequence::getBBoxDirAdj)
+                                 .map(Word::getBBoxDirAdj)
                                 .mapToDouble(RectangularShape::getWidth).average().orElse(75) * MAX_WORD_DISTANCE_FACTOR;

-        UnionFind<TextPositionSequence> unionFind = new UnionFind<>(sequences);
+        return groupByLine(sequences, maxLineDistance, maxXGap);
+    }

-        for (TextPositionSequence sequence : sequences) {
-            for (TextPositionSequence sequence2 : sequences) {
+
+    public Collection<Set<Word>> groupByLine(Set<Word> sequences, double maxLineDistance, double maxXGap) {
+
+        UnionFind<Word> unionFind = new UnionFind<>(sequences);
+
+        for (Word sequence : sequences) {
+            for (Word sequence2 : sequences) {

                if (sequence.equals(sequence2)) { // || unionFind.inSameSet(sequence, sequence2)) doing this is actually slower than not doing it
                    continue;
@@ -144,10 +160,10 @@ public class TextPositionOperations {
    }


-    public List<TextPositionSequence> merge(List<TextPageBlock> textBlocks) {
+    public List<Word> merge(List<TextPageBlock> textBlocks) {

        return textBlocks.stream()
-                .map(TextPageBlock::getSequences)
+                .map(TextPageBlock::getWords)
                .flatMap(Collection::stream)
                .collect(Collectors.toList());
    }
@@ -27,7 +27,9 @@ import com.knecon.fforesight.service.layoutparser.processor.model.outline.Outlin
 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.ListIdentifier;
-import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.TextBlockOnPage;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;
+import com.knecon.fforesight.service.layoutparser.processor.services.classification.NumberWord;
 import com.knecon.fforesight.service.layoutparser.processor.utils.CoordinateTransforms;
 import com.knecon.fforesight.service.layoutparser.processor.utils.MarkedContentUtils;
 import com.knecon.fforesight.service.layoutparser.processor.utils.PageInformation;
@@ -58,14 +60,14 @@ public class LayoutDebugLayer extends LayoutDebugLayerConfig {
    Map<Integer, AtomicInteger> outlineObjectsWithoutPointsPerPage = new HashMap<>();


-    public void addTextVisualizations(List<TextPositionSequence> textPositionSequences, int pageNumber) {
+    public void addTextVisualizations(List<Word> words, int pageNumber) {

        if (!active) {
            return;
        }
        VisualizationsOnPage visualizationsOnPage = getOrCreateVisualizationsOnPage(pageNumber, this.words);
        visualizationsOnPage.getColoredRectangles()
-                .addAll(textPositionSequences.stream()
+                .addAll(words.stream()
                                .map(BoundingBox::getBBoxPdf)
                                .map(rect -> new ColoredRectangle(rect, WORDS_COLOR, 1))
                                .toList());
@@ -188,7 +190,7 @@ public class LayoutDebugLayer extends LayoutDebugLayerConfig {
    }


-    public void addLineVisualizationsFromNestedTextPosition(Collection<Set<TextPositionSequence>> lines, int pageNumber) {
+    public void addLineVisualizationsFromNestedTextPosition(Collection<Set<Word>> lines, int pageNumber) {

        if (!active) {
            return;
@@ -273,7 +275,9 @@ public class LayoutDebugLayer extends LayoutDebugLayerConfig {
        AtomicInteger index = new AtomicInteger(0);
        zones.forEach(zone -> zone.getLines()
                .stream()
-                .map(Line::getCharacters)
+                .map(Line::getWords)
+                .flatMap(Collection::stream)
+                .map(Word::getCharacters)
                .flatMap(Collection::stream)
                .forEach(character -> {
                    Color color = getRotatingColor(index);
@@ -291,7 +295,7 @@ public class LayoutDebugLayer extends LayoutDebugLayerConfig {
    }


-    public void addTocPages(List<TextPositionSequence> numbers, int page) {
+    public void addTocPages(List<NumberWord> numbers, int page) {

        if (!active) {
            return;
@@ -300,13 +304,10 @@ public class LayoutDebugLayer extends LayoutDebugLayerConfig {
        VisualizationsOnPage visualizationsOnPage = getOrCreateVisualizationsOnPage(page, this.tocPages);
        visualizationsOnPage.getColoredRectangles()
                .addAll(numbers.stream()
+                                .map(NumberWord::word)
                                .map(BoundingBox::getBBoxPdf)
                                .map(line -> new ColoredRectangle(line, LINES_COLOR, LINE_WIDTH))
                                .toList());
-        visualizationsOnPage.getColoredRectangles()
-                .add(new ColoredRectangle(numbers.stream()
-                                                  .map(BoundingBox::getBBoxPdf)
-                                                  .collect(RectangleTransformations.collectBBox()), LINES_COLOR, LINE_WIDTH));
    }


@@ -330,8 +331,10 @@ public class LayoutDebugLayer extends LayoutDebugLayerConfig {

    private void addOutlineObject(OutlineObject outlineObject, PageInformation pageInformation) {

+        if (!active) {
+            return;
+        }
        int rectSize = 5;
-
        Point2D point2D;
        if (outlineObject.getPoint().isPresent()) {
            point2D = outlineObject.getPoint().get();
@@ -355,10 +358,25 @@ public class LayoutDebugLayer extends LayoutDebugLayerConfig {

    public void addListIdentifiers(List<ListIdentifier> listIdentifiers) {

+        if (!active) {
+            return;
+        }
        for (ListIdentifier listIdentifier : listIdentifiers) {
            getOrCreateVisualizationsOnPage(listIdentifier.getPage(), this.listIdentifiers).getColoredRectangles()
                    .add(new ColoredRectangle(listIdentifier.getWord().getBBoxPdf(), WORDS_COLOR, LINE_WIDTH));
        }
    }

+
+    public void addTocBlocks(Set<TextBlockOnPage> blocksWithNumberInCluster) {
+
+        if (!active) {
+            return;
+        }
+        for (TextBlockOnPage textBlockOnPage : blocksWithNumberInCluster) {
+            getOrCreateVisualizationsOnPage(textBlockOnPage.page().getPageNumber(), this.tocBlocks).getColoredRectangles()
+                    .add(new ColoredRectangle(textBlockOnPage.textBlock().getBBoxPdf(), TOC_COLOR, LINE_WIDTH));
+        }
+    }
+
 }
@@ -83,4 +83,11 @@ class SectionIdentifierTest {
        assertEquals(List.of(4, 1, 2, 4), identifier.getIdentifiers());
    }

+    @Test
+    void testFalsePositive111() {
+        SectionIdentifier identifier = SectionIdentifier.fromSearchText("111: Headline");
+        assertEquals(SectionIdentifier.Format.NUMERICAL, identifier.getFormat());
+        assertEquals(1, identifier.level());
+    }
+
 }
@@ -26,7 +26,7 @@ import com.knecon.fforesight.service.layoutparser.processor.model.Classification
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.AtomicTextBlock;
 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.model.text.Word;
 import com.knecon.fforesight.service.layoutparser.processor.services.factory.DocumentGraphFactory;
 import com.knecon.fforesight.service.layoutparser.processor.utils.TextPositionOperations;
 import com.knecon.fforesight.service.layoutparser.server.PDFNetInitializer;
@@ -387,7 +387,7 @@ public class DocumentReadingOrderTest extends BuildDocumentTest {
                for (AbstractPageBlock abstractBlock : classificationDocumentPage.getTextBlocks()) {

                    if (abstractBlock instanceof TextPageBlock textBlock) {
-                        for (TextPositionSequence sequence : TextPositionOperations.mergeAndSort(List.of(textBlock))) {
+                        for (Word sequence : TextPositionOperations.mergeAndSort(List.of(textBlock))) {

                            float stringWidth;
                            try {
@@ -32,7 +32,7 @@ import com.knecon.fforesight.service.layoutparser.processor.model.graph.textbloc
 import com.knecon.fforesight.service.layoutparser.processor.model.image.ClassifiedImage;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.Cell;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.TablePageBlock;
-import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;
 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;
@@ -105,9 +105,9 @@ public class PdfSegmentationServiceTest extends AbstractTest {

        List<PageContents> textPositionPerPage = PageContentExtractor.getSortedPageContents(fileName);
        var textPositions = textPositionPerPage.stream()
-                .flatMap(t -> t.getSortedTextPositionSequences()
+                .flatMap(t -> t.getSortedWords()
                        .stream()
-                        .map(TextPositionSequence::toString))
+                        .map(Word::toString))
                .collect(Collectors.joining(" "));
        assertThat(textPositions.contains(textToSearch)).isFalse();

@@ -117,7 +117,7 @@ public class PdfSegmentationServiceTest extends AbstractTest {
                           .get(0).getTextBlocks().size()).isEqualTo(3);
        assertThat(classificationDocument.getHeaders()
                           .get(0).getTextBlocks()
-                           .get(0).getSequences().size()).isEqualTo(8);
+                           .get(0).getWords().size()).isEqualTo(8);
        assertThat(classificationDocument.getHeaders()
                           .get(0).getTextBlocks()
                           .get(0).toString()).contains(textToSearch);
@@ -36,7 +36,7 @@ class GapAcrossLinesDetectionServiceTest {
        System.out.println("start column detection");
        start = System.currentTimeMillis();
        for (PageInformation pageInformation : pageInformations) {
-            GapInformation gapInformation = GapDetectionService.findGapsInLines(pageInformation.getPageContents().getSortedTextPositionSequences(),  pageInformation.getMainBodyTextFrame());
+            GapInformation gapInformation = GapDetectionService.findGapsInLines(pageInformation.getPageContents().getSortedWords(), pageInformation.getMainBodyTextFrame());
            columnsPerPage.add(GapsAcrossLinesService.detectXGapsAcrossLines(gapInformation, pageInformation.getMainBodyTextFrame()));
        }
        System.out.printf("Finished column detection in %d ms%n", System.currentTimeMillis() - start);
@@ -10,7 +10,7 @@ import java.util.stream.Collectors;
 import org.junit.jupiter.api.Test;

 import com.knecon.fforesight.service.layoutparser.processor.model.PageInformation;
-import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;
 import com.knecon.fforesight.service.layoutparser.processor.services.InvisibleTableDetectionService;
 import com.knecon.fforesight.service.layoutparser.processor.services.PageInformationService;
 import com.knecon.fforesight.service.layoutparser.processor.services.PageContentExtractor;
@@ -34,18 +34,18 @@ class InvisibleTableDetectionServiceTest {
                .collect(Collectors.toList());

        int pageNumber = 1;
-        Rectangle2D tableBBox = pageContents.get(0).getPageContents().getSortedTextPositionSequences().subList(45, 152)
+        Rectangle2D tableBBox = pageContents.get(0).getPageContents().getSortedWords().subList(45, 152)
                .stream()
-                .map(TextPositionSequence::getBBox)
+                .map(Word::getBBox)
                .map(this::mirrorY)
                .collect(RectangleTransformations.collectBBox());

-        List<TextPositionSequence> textPositionSequences = pageContents.get(0).getPageContents().getSortedTextPositionSequences()
+        List<Word> words = pageContents.get(0).getPageContents().getSortedWords()
                .stream()
                .filter(textPositionSequence -> tableBBox.contains(mirrorY(textPositionSequence.getBBox())))
                .toList();

-        var table = InvisibleTableDetectionService.detectTable(textPositionSequences, tableBBox);
+        var table = InvisibleTableDetectionService.detectTable(words, tableBBox);

        PdfDraw.drawRectanglesPerPage(fileName,
                                      List.of(table.stream()
@@ -6,7 +6,7 @@ import java.util.List;
 import org.junit.jupiter.api.Test;

 import com.knecon.fforesight.service.layoutparser.processor.model.PageContents;
-import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;
 import com.knecon.fforesight.service.layoutparser.processor.services.PageContentExtractor;
 import com.knecon.fforesight.service.layoutparser.server.utils.visualizations.PdfDraw;

@@ -25,9 +25,9 @@ class PageContentExtractorTest {

        PdfDraw.drawRectanglesPerPageNumberedByLine(fileName,
                textPositionPerPage.stream()
-                        .map(t -> t.getSortedTextPositionSequences()
+                        .map(t -> t.getSortedWords()
                                .stream()
-                                .map(TextPositionSequence::getBBoxPdf)
+                                .map(Word::getBBoxPdf)
                                .map(List::of)
                                .toList())
                        .toList(), tmpFileName);
@@ -11,7 +11,7 @@ import org.junit.jupiter.api.Test;
 import com.knecon.fforesight.service.layoutparser.processor.model.PageContents;
 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.model.text.Word;
 import com.knecon.fforesight.service.layoutparser.processor.services.PageContentExtractor;
 import com.knecon.fforesight.service.layoutparser.processor.services.RulingCleaningService;
 import com.knecon.fforesight.service.layoutparser.processor.services.TextRulingsClassifier;
@@ -32,16 +32,16 @@ public class RulingsClassifierTest {
        for (PageContents pageContent : pageContents) {
            CleanRulings cleanRulings = rulingCleaningService.deduplicateAndStraightenRulings(Collections.emptyList(), pageContent.getRulings());
            RectangularIntersectionFinder.find(cleanRulings.getHorizontals(), cleanRulings.getVerticals());
-            TextRulingsClassifier.classifyUnderlinedAndStrikethroughText(pageContent.getSortedTextPositionSequences(), cleanRulings);
+            TextRulingsClassifier.classifyUnderlinedAndStrikethroughText(pageContent.getSortedWords(), cleanRulings);

-            assertTrue(pageContent.getSortedTextPositionSequences()
+            assertTrue(pageContent.getSortedWords()
                               .stream()
                               .filter(word -> word.toString().equals("Underlined"))
-                               .allMatch(TextPositionSequence::isUnderline));
-            assertTrue(pageContent.getSortedTextPositionSequences()
+                               .allMatch(Word::isUnderline));
+            assertTrue(pageContent.getSortedWords()
                               .stream()
                               .filter(word -> word.toString().equals("Striketrough"))
-                               .allMatch(TextPositionSequence::isStrikethrough));
+                               .allMatch(Word::isStrikethrough));

            assertEquals(4,
                         cleanRulings.buildAll()
@@ -70,7 +70,7 @@ public class RulingsClassifierTest {
        for (PageContents pageContent : pageContents) {
            CleanRulings cleanRulings = rulingCleaningService.deduplicateAndStraightenRulings(Collections.emptyList(), pageContent.getRulings());
            RectangularIntersectionFinder.find(cleanRulings.getHorizontals(), cleanRulings.getVerticals());
-            TextRulingsClassifier.classifyUnderlinedAndStrikethroughText(pageContent.getSortedTextPositionSequences(), cleanRulings);
+            TextRulingsClassifier.classifyUnderlinedAndStrikethroughText(pageContent.getSortedWords(), cleanRulings);

            assertEquals(30, cleanRulings.getHorizontals().size());
            assertEquals(30, cleanRulings.getTableLines().getHorizontals().size());
@@ -1,8 +1,6 @@
 package com.knecon.fforesight.service.layoutparser.server.utils;

-import java.awt.geom.Rectangle2D;
 import java.io.File;
-import java.io.FileOutputStream;
 import java.nio.file.Path;
 import java.util.Map;

@@ -12,27 +10,11 @@ import org.springframework.core.io.ClassPathResource;
 import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingRequest;
 import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingType;
 import com.knecon.fforesight.service.layoutparser.processor.LayoutParsingPipeline;
-import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationDocument;
-import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
-import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
-import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.TableServiceResponse;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.VisualLayoutParsingResponse;
 import com.knecon.fforesight.service.layoutparser.processor.services.factory.DocumentGraphFactory;
-import com.knecon.fforesight.service.viewerdoc.model.Standard14EmbeddableFont;
-import com.pdftron.common.Matrix2D;
-import com.pdftron.pdf.ColorPt;
-import com.pdftron.pdf.ColorSpace;
-import com.pdftron.pdf.Element;
-import com.pdftron.pdf.ElementBuilder;
-import com.pdftron.pdf.ElementWriter;
-import com.pdftron.pdf.Font;
-import com.pdftron.pdf.GState;
-import com.pdftron.pdf.PDFDoc;
-import com.pdftron.pdf.Page;
-import com.pdftron.sdf.SDFDoc;

 import lombok.SneakyThrows;

@@ -57,6 +57,7 @@ public record LayerIdentifier(String name, String markedContentName) {
    public static final LayerIdentifier OUTLINE_OBJECTS = new LayerIdentifier("Outline Positions", "OUTLINE_OBJECTS");
    public static final LayerIdentifier SENTENCES = new LayerIdentifier("Sentences", "SENTENCES");
    public static final LayerIdentifier TOC_PAGES = new LayerIdentifier("TOC pages", "TOC_PAGES");
+    public static final LayerIdentifier TOC_BLOCKS = new LayerIdentifier("TOC blocks", "TOC_BLOCKS");
    public static final LayerIdentifier LIST_IDENTIFIERS = new LayerIdentifier("List identifiers", "LIST_IDENTIFIERS");

    // Visual layout parser
@@ -22,6 +22,7 @@ public class LayoutDebugLayerConfig extends AbstractLayerGroup {

    protected static final Color WORDS_COLOR = new Color(68, 84, 147);
    protected static final Color LINES_COLOR = new Color(152, 45, 179);
+    protected static final Color TOC_COLOR = new Color(33, 159, 144);
    protected static final Color ZONES_COLOR = new Color(131, 38, 38);

    protected static final Color RULINGS_COLOR = new Color(21, 221, 174);
@@ -46,9 +47,9 @@ public class LayoutDebugLayerConfig extends AbstractLayerGroup {
                                                                          new Color(0, 188, 212),
                                                                          new Color(121, 85, 72));

-    protected final Visualizations words = Visualizations.builder().layer(LayerIdentifier.WORDS).visibleByDefault(true).build();
+    protected final Visualizations words = Visualizations.builder().layer(LayerIdentifier.WORDS).build();
    protected final Visualizations lines = Visualizations.builder().layer(LayerIdentifier.LINES).build();
-    protected final Visualizations zones = Visualizations.builder().layer(LayerIdentifier.ZONES).build();
+    protected final Visualizations zones = Visualizations.builder().layer(LayerIdentifier.ZONES).visibleByDefault(true).build();
    protected final Visualizations mainBody = Visualizations.builder().layer(LayerIdentifier.MAIN_BODY).build();
    protected final Visualizations clean_rulings = Visualizations.builder().layer(LayerIdentifier.CLEAN_RULINGS).build();
    protected final Visualizations rulings = Visualizations.builder().layer(LayerIdentifier.RULINGS).build();
@@ -59,6 +60,7 @@ public class LayoutDebugLayerConfig extends AbstractLayerGroup {
    protected final Visualizations outlineObjects = Visualizations.builder().layer(LayerIdentifier.OUTLINE_OBJECTS).build();
    protected final Visualizations sentences = Visualizations.builder().layer(LayerIdentifier.SENTENCES).build();
    protected final Visualizations tocPages = Visualizations.builder().layer(LayerIdentifier.TOC_PAGES).build();
+    protected final Visualizations tocBlocks = Visualizations.builder().layer(LayerIdentifier.TOC_BLOCKS).build();
    protected final Visualizations listIdentifiers = Visualizations.builder().layer(LayerIdentifier.LIST_IDENTIFIERS).build();


@@ -77,9 +79,9 @@ public class LayoutDebugLayerConfig extends AbstractLayerGroup {
                       markedContent, //
                       outlineObjects, //
                       tocPages, //
+                       tocBlocks, //
                       listIdentifiers //
        );
    }

-
 }
Author	SHA1	Message	Date
Kilian Schüttler	e44fc3b536	Merge branch 'RM-231-bp' into 'release/0.192.x' RM-231: missing whitespace in name See merge request fforesight/layout-parser!266	2025-01-14 13:07:49 +01:00
Kilian Schuettler	237034b696	RM-231: missing whitespace in name	2025-01-14 13:01:09 +01:00
Kilian Schuettler	85432a2511	RED-10714: fix IndexOutOfBoundsException (cherry picked from commit `0b6a292c75`)	2025-01-13 13:22:59 +01:00
Kilian Schüttler	a821570065	Merge branch 'RED-9139-bp' into 'main' RED-9139: more robust TOC detection See merge request fforesight/layout-parser!254	2024-11-13 10:54:39 +01:00
Kilian Schüttler	7ee1f9e360	RED-9139: more robust TOC detection	2024-11-13 10:54:39 +01:00
Kilian Schüttler	f9b25c8157	Merge branch 'RED-10249' into 'main' RED-10249: regex found incorrectly due to wrong text sorting See merge request fforesight/layout-parser!252	2024-11-04 12:51:38 +01:00
Kilian Schüttler	c90874da7a	RED-10249: regex found incorrectly due to wrong text sorting	2024-11-04 12:51:37 +01:00
Kilian Schüttler	4683c696a5	Merge branch 'RED-10247' into 'main' RED-10247: dictionary entry not found in footer due to wrong text sorting See merge request fforesight/layout-parser!251	2024-10-25 18:30:35 +02:00
Kilian Schuettler	95c02ce3cf	RED-10247: dictionary entry not found in footer due to wrong text sorting	2024-10-25 17:18:14 +02:00
Kilian Schüttler	b2d62e32fe	Merge branch 'RED-10270-fp' into 'main' RED-10270: fix NumberFormatException See merge request fforesight/layout-parser!248	2024-10-24 17:14:47 +02:00
Kilian Schuettler	65c1f03ea3	RED-10270: fix NumberFormatException	2024-10-24 10:59:05 +02:00
Kilian Schüttler	2219519a2b	Merge branch 'RED-10127' into 'main' RED-10127: rename TextPositionSequence to Word See merge request fforesight/layout-parser!244	2024-10-18 12:20:15 +02:00
Kilian Schüttler	af05218e37	RED-10127: rename TextPositionSequence to Word	2024-10-18 12:20:15 +02:00