Merge branch 'RED-8550-pre-backup' into 'main'

RED-8550: Faulty table recognition and text duplication leads to huge sections See merge request fforesight/layout-parser!105
RED-8550: Faulty table recognition and text duplication leads to huge sections
2024-02-28 16:13:56 +01:00 · 2024-02-28 16:13:56 +01:00 · 2024-02-26 12:32:14 +01:00 · 2024-02-26 12:24:15 +01:00 · 2024-02-23 14:08:57 +01:00 · 2024-02-23 14:02:53 +01:00
52 changed files with 601 additions and 381 deletions
@@ -5,6 +5,7 @@ import java.io.Serializable;
 import java.util.Arrays;
 import java.util.List;
 import java.util.Map;
+import java.util.Set;
 import java.util.stream.Stream;

 import io.swagger.v3.oas.annotations.media.Schema;
@@ -115,6 +116,8 @@ public class DocumentStructure implements Serializable {
        Map<String, String> properties;
        @Schema(description = "All child Entries of this Entry.", example = "[1, 2, 3]")
        List<EntryData> children;
+        @Schema(description = "Describes the origin of the semantic node", example = "[ALGORITHM]")
+        Set<LayoutEngine> engines;


        @Override
@@ -0,0 +1,6 @@
+package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
+
+public enum LayoutEngine {
+    ALGORITHM,
+    AI
+}
@@ -18,11 +18,15 @@ public record LayoutParsingRequest(

        @Schema(description = "Path to the original PDF file.")//
        @NonNull String originFileStorageId,//
+
+
        @Schema(description = "Optional Path to the table extraction file.")//
        Optional<String> tablesFileStorageId,//
        @Schema(description = "Optional Path to the image classification file.")//
        Optional<String> imagesFileStorageId,//

+        @Schema(description = "Optional Path to the the visual layout parsing service file") Optional<String> visualLayoutParsingFileId,//
+
        @Schema(description = "Path where the Document Structure File will be stored.")//
        @NonNull String structureFileStorageId,//
        @Schema(description = "Path where the Research Data File will be stored.")//
@@ -28,15 +28,16 @@ import com.knecon.fforesight.service.layoutparser.processor.model.Classification
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
 import com.knecon.fforesight.service.layoutparser.processor.model.image.ClassifiedImage;
-import com.knecon.fforesight.service.layoutparser.processor.model.table.Cell;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRulings;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.adapter.CvTableParsingAdapter;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.adapter.ImageServiceResponseAdapter;
+import com.knecon.fforesight.service.layoutparser.processor.python_api.adapter.VisualLayoutParsingAdapter;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.image.ImageServiceResponse;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.TableCells;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.TableServiceResponse;
+import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.VisualLayoutParsingResponse;
 import com.knecon.fforesight.service.layoutparser.processor.services.BodyTextFrameService;
 import com.knecon.fforesight.service.layoutparser.processor.services.RulingCleaningService;
 import com.knecon.fforesight.service.layoutparser.processor.services.SectionsBuilderService;
@@ -87,6 +88,7 @@ public class LayoutParsingPipeline {
    RedactManagerBlockificationService redactManagerBlockificationService;
    LayoutGridService layoutGridService;
    ObservationRegistry observationRegistry;
+    VisualLayoutParsingAdapter visualLayoutParsingAdapter;


    public LayoutParsingFinishedEvent parseLayoutAndSaveFilesToStorage(LayoutParsingRequest layoutParsingRequest) throws IOException {
@@ -95,10 +97,15 @@ public class LayoutParsingPipeline {
        log.info("Starting layout parsing for {}", layoutParsingRequest.identifier());

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

+        VisualLayoutParsingResponse visualLayoutParsingResponse = new VisualLayoutParsingResponse();
+        if (layoutParsingRequest.visualLayoutParsingFileId()
+                .isPresent()) {
+            visualLayoutParsingResponse = layoutParsingStorageService.getVisualLayoutParsingFile(layoutParsingRequest.visualLayoutParsingFileId()
+                                                                                                         .get());
+        }

        ImageServiceResponse imageServiceResponse = new ImageServiceResponse();
        if (layoutParsingRequest.imagesFileStorageId()
@@ -115,10 +122,11 @@ public class LayoutParsingPipeline {
        }

        ClassificationDocument classificationDocument = parseLayout(layoutParsingRequest.layoutParsingType(),
-                originFile,
-                imageServiceResponse,
-                tableServiceResponse,
-                layoutParsingRequest.identifier().toString());
+                                                                    originFile,
+                                                                    imageServiceResponse,
+                                                                    tableServiceResponse,
+                                                                    visualLayoutParsingResponse,
+                                                                    layoutParsingRequest.identifier().toString());

        log.info("Building document graph for {}", layoutParsingRequest.identifier());

@@ -206,13 +214,14 @@ public class LayoutParsingPipeline {
                                              File originFile,
                                              ImageServiceResponse imageServiceResponse,
                                              TableServiceResponse tableServiceResponse,
+                                              VisualLayoutParsingResponse visualLayoutParsingResponse,
                                              String identifier) {

        PDDocument originDocument = openDocument(originFile);
        addNumberOfPagesToTrace(originDocument.getNumberOfPages(), Files.size(originFile.toPath()));
        Map<Integer, List<TableCells>> pdfTableCells = cvTableParsingAdapter.buildCvParsedTablesPerPage(tableServiceResponse);
        Map<Integer, List<ClassifiedImage>> pdfImages = imageServiceResponseAdapter.buildClassifiedImagesPerPage(imageServiceResponse);
-
+        Map<Integer, List<ClassifiedImage>> signatures = visualLayoutParsingAdapter.buildExtractedSignaturesPerPage(visualLayoutParsingResponse);
        ClassificationDocument classificationDocument = new ClassificationDocument();
        List<ClassificationPage> classificationPages = new ArrayList<>();

@@ -251,10 +260,8 @@ public class LayoutParsingPipeline {
            PDRectangle cropbox = pdPage.getCropBox();
            CleanRulings cleanRulings = rulingCleaningService.getCleanRulings(pdfTableCells.get(pageNumber), stripper.getRulings());

-            List<Cell> emptyTableCells = TableExtractionService.findCells(cleanRulings.getHorizontal(), cleanRulings.getVertical());
-
            ClassificationPage classificationPage = switch (layoutParsingType) {
-                case REDACT_MANAGER -> redactManagerBlockificationService.blockify(stripper.getTextPositionSequences(), emptyTableCells);
+                case REDACT_MANAGER -> redactManagerBlockificationService.blockify(stripper.getTextPositionSequences(), cleanRulings.getHorizontal(), cleanRulings.getVertical());
                case TAAS -> taasBlockificationService.blockify(stripper.getTextPositionSequences(), cleanRulings.getHorizontal(), cleanRulings.getVertical());
                case DOCUMINE -> docuMineBlockificationService.blockify(stripper.getTextPositionSequences(), cleanRulings.getHorizontal(), cleanRulings.getVertical());
            };
@@ -274,6 +281,14 @@ public class LayoutParsingPipeline {
                imageServiceResponseAdapter.findOcr(classificationPage);
            }

+            if (signatures.containsKey(pageNumber)) {
+                if (classificationPage.getImages() == null || classificationPage.getImages().size() == 0) {
+                    classificationPage.setImages(signatures.get(pageNumber));
+                } else {
+                    classificationPage.getImages().addAll(signatures.get(pageNumber));
+                }
+            }
+
            tableExtractionService.extractTables(cleanRulings, classificationPage);

            buildPageStatistics(classificationPage);
@@ -10,8 +10,6 @@ import java.nio.file.Paths;
 import java.nio.file.StandardOpenOption;
 import java.util.Optional;

-import org.apache.pdfbox.Loader;
-import org.apache.pdfbox.pdmodel.PDDocument;
 import org.springframework.stereotype.Service;

 import com.fasterxml.jackson.databind.ObjectMapper;
@@ -22,6 +20,7 @@ import com.knecon.fforesight.service.layoutparser.internal.api.data.taas.Researc
 import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingRequest;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.image.ImageServiceResponse;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.TableServiceResponse;
+import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.VisualLayoutParsingResponse;
 import com.knecon.fforesight.tenantcommons.TenantContext;

 import io.micrometer.observation.annotation.Observed;
@@ -79,6 +78,14 @@ public class LayoutParsingStorageService {
        }
    }

+    public VisualLayoutParsingResponse getVisualLayoutParsingFile(String storageId) throws IOException {
+
+        try (InputStream inputStream = getObject(storageId)) {
+            VisualLayoutParsingResponse visualLayoutParsingResponse = objectMapper.readValue(inputStream, VisualLayoutParsingResponse.class);
+            return visualLayoutParsingResponse;
+        }
+    }
+

    @Observed(name = "LayoutParsingStorageService", contextualName = "store-document-data")
    public void storeDocumentData(LayoutParsingRequest layoutParsingRequest, DocumentData documentData) {
@@ -10,6 +10,7 @@ import java.util.Set;
 import java.util.stream.Collectors;
 import java.util.stream.Stream;

+import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
@@ -30,6 +31,9 @@ import lombok.experimental.FieldDefaults;
@FieldDefaults(level = AccessLevel.PRIVATE)
 public class Document implements GenericSemanticNode {

+    @Builder.Default
+    Set<LayoutEngine> engines = new HashSet<>(Set.of(LayoutEngine.ALGORITHM));
+
    Set<Page> pages;
    DocumentTree documentTree;
    Integer numberOfPages;
@@ -56,13 +60,15 @@ public class Document implements GenericSemanticNode {

    public List<Section> getMainSections() {

-        return streamChildrenOfType(NodeType.SECTION).map(node -> (Section) node).collect(Collectors.toList());
+        return streamChildrenOfType(NodeType.SECTION).map(node -> (Section) node)
+                .collect(Collectors.toList());
    }


    public Stream<TextBlock> streamTerminalTextBlocksInOrder() {

-        return streamAllNodes().filter(SemanticNode::isLeaf).map(SemanticNode::getLeafTextBlock);
+        return streamAllNodes().filter(SemanticNode::isLeaf)
+                .map(SemanticNode::getLeafTextBlock);
    }


@@ -83,13 +89,16 @@ public class Document implements GenericSemanticNode {
    @Override
    public Headline getHeadline() {

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


    private Stream<SemanticNode> streamAllNodes() {

-        return documentTree.allEntriesInOrder().map(DocumentTree.Entry::getNode);
+        return documentTree.allEntriesInOrder()
+                .map(DocumentTree.Entry::getNode);
    }


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

+import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
@@ -26,6 +27,9 @@ import lombok.experimental.FieldDefaults;
@FieldDefaults(level = AccessLevel.PRIVATE)
 public class Footer implements GenericSemanticNode {

+    @Builder.Default
+    Set<LayoutEngine> engines = new HashSet<>(Set.of(LayoutEngine.ALGORITHM));
+
    List<Integer> treeId;
    TextBlock leafTextBlock;

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

+import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
@@ -26,6 +27,8 @@ import lombok.experimental.FieldDefaults;
@FieldDefaults(level = AccessLevel.PRIVATE)
 public class Header implements GenericSemanticNode {

+    @Builder.Default
+    Set<LayoutEngine> engines = new HashSet<>(Set.of(LayoutEngine.ALGORITHM));
    List<Integer> treeId;
    TextBlock leafTextBlock;

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

+import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
@@ -26,6 +27,8 @@ import lombok.experimental.FieldDefaults;
@FieldDefaults(level = AccessLevel.PRIVATE)
 public class Headline implements GenericSemanticNode {

+    @Builder.Default
+    Set<LayoutEngine> engines = new HashSet<>(Set.of(LayoutEngine.ALGORITHM));
    List<Integer> treeId;
    TextBlock leafTextBlock;

@@ -8,10 +8,12 @@ import java.util.List;
 import java.util.Map;
 import java.util.Set;

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

 import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
@@ -28,6 +30,9 @@ import lombok.experimental.FieldDefaults;
@FieldDefaults(level = AccessLevel.PRIVATE)
 public class Image implements GenericSemanticNode {

+    @Builder.Default
+    Set<LayoutEngine> engines = new HashSet<>(Set.of(LayoutEngine.ALGORITHM));
+
    List<Integer> treeId;
    String id;

@@ -35,8 +40,6 @@ public class Image implements GenericSemanticNode {
    boolean transparent;
    Rectangle2D position;

-    TextBlock leafTextBlock;
-
    boolean redaction;
    boolean ignored;
    @Builder.Default
@@ -67,7 +70,9 @@ public class Image implements GenericSemanticNode {
    @Override
    public TextBlock getTextBlock() {

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


@@ -93,11 +98,4 @@ public class Image implements GenericSemanticNode {
        return bBoxPerPage;
    }

-
-    @Override
-    public boolean isLeaf() {
-
-        return true;
-    }
-
 }
@@ -6,6 +6,8 @@ public enum ImageType {
    LOGO,
    FORMULA,
    SIGNATURE,
+
+    SIGNATURE_VISUAL,
    OTHER,
    OCR;

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

+import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
@@ -24,6 +25,9 @@ import lombok.experimental.FieldDefaults;
@FieldDefaults(level = AccessLevel.PRIVATE)
 public class Paragraph implements GenericSemanticNode {

+    @Builder.Default
+    Set<LayoutEngine> engines = new HashSet<>(Set.of(LayoutEngine.ALGORITHM));
+
    List<Integer> treeId;
    TextBlock leafTextBlock;

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

+import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
@@ -27,6 +28,8 @@ import lombok.extern.slf4j.Slf4j;
@FieldDefaults(level = AccessLevel.PRIVATE)
 public class Section implements GenericSemanticNode {

+    @Builder.Default
+    Set<LayoutEngine> engines = new HashSet<>(Set.of(LayoutEngine.ALGORITHM));
    List<Integer> treeId;

    TextBlock textBlock;
@@ -50,7 +53,8 @@ public class Section implements GenericSemanticNode {

    public boolean hasTables() {

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


@@ -58,7 +62,9 @@ public class Section implements GenericSemanticNode {
    public TextBlock getTextBlock() {

        if (textBlock == null) {
-            textBlock = streamAllSubNodes().filter(SemanticNode::isLeaf).map(SemanticNode::getLeafTextBlock).collect(new TextBlockCollector());
+            textBlock = streamAllSubNodes().filter(SemanticNode::isLeaf)
+                    .map(SemanticNode::getLeafTextBlock)
+                    .collect(new TextBlockCollector());
        }
        return textBlock;
    }
@@ -12,6 +12,7 @@ import java.util.Set;
 import java.util.stream.Collectors;
 import java.util.stream.Stream;

+import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.Boundary;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
@@ -335,6 +336,23 @@ public interface SemanticNode {
    }


+    /**
+     * returns the set of layoutengines.
+     *
+     * @return set of layoutengines.
+     */
+    Set<LayoutEngine> getEngines();
+
+
+    /**
+     * adds a layoutengine to the set.
+     */
+    default void addEngine(LayoutEngine engine) {
+
+        getEngines().add(engine);
+    }
+
+
    /**
     * Streams all children located directly underneath this node in the DocumentTree.
     *
@@ -426,6 +444,7 @@ public interface SemanticNode {
    /**
     * TODO: this produces unwanted results for sections spanning multiple columns.
     * Computes the Union of the bounding boxes of all children recursively.
+     *
     * @return The union of the BoundingBoxes of all children
     */
    private Map<Page, Rectangle2D> getBBoxFromChildren() {
@@ -12,6 +12,7 @@ import java.util.Set;
 import java.util.stream.IntStream;
 import java.util.stream.Stream;

+import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
@@ -31,6 +32,8 @@ import lombok.experimental.FieldDefaults;
@FieldDefaults(level = AccessLevel.PRIVATE)
 public class Table implements SemanticNode {

+    @Builder.Default
+    Set<LayoutEngine> engines = new HashSet<>(Set.of(LayoutEngine.ALGORITHM));
    List<Integer> treeId;
    DocumentTree documentTree;

@@ -45,6 +48,7 @@ public class Table implements SemanticNode {
    @EqualsAndHashCode.Exclude
    Map<Page, Rectangle2D> bBoxCache;

+
    /**
     * Streams all entities in this table, that appear in a row, which contains any of the provided strings.
     *
@@ -53,8 +57,7 @@ public class Table implements SemanticNode {
     */
    public Stream<RedactionEntity> streamEntitiesWhereRowContainsStringsIgnoreCase(List<String> strings) {

-        return IntStream.range(0, numberOfRows)
-                .boxed()
+        return IntStream.range(0, numberOfRows).boxed()
                .filter(row -> rowContainsStringsIgnoreCase(row, strings))
                .flatMap(this::streamRow)
                .map(TableCell::getEntities)
@@ -71,8 +74,11 @@ public class Table implements SemanticNode {
     */
    public boolean rowContainsStringsIgnoreCase(Integer row, List<String> strings) {

-        String rowText = streamRow(row).map(TableCell::getTextBlock).collect(new TextBlockCollector()).getSearchText().toLowerCase(Locale.ROOT);
-        return strings.stream().map(String::toLowerCase).allMatch(rowText::contains);
+        String rowText = streamRow(row).map(TableCell::getTextBlock)
+                .collect(new TextBlockCollector()).getSearchText().toLowerCase(Locale.ROOT);
+        return strings.stream()
+                .map(String::toLowerCase)
+                .allMatch(rowText::contains);
    }


@@ -85,9 +91,13 @@ public class Table implements SemanticNode {
     */
    public Stream<RedactionEntity> streamEntitiesWhereRowHasHeaderAndValue(String header, String value) {

-        List<Integer> vertebrateStudyCols = streamHeaders().filter(headerNode -> headerNode.containsString(header)).map(TableCell::getCol).toList();
+        List<Integer> vertebrateStudyCols = streamHeaders().filter(headerNode -> headerNode.containsString(header))
+                .map(TableCell::getCol)
+                .toList();
        return streamTableCells().filter(tableCellNode -> vertebrateStudyCols.stream()
-                .anyMatch(vertebrateStudyCol -> getCell(tableCellNode.getRow(), vertebrateStudyCol).containsString(value))).map(TableCell::getEntities).flatMap(Collection::stream);
+                        .anyMatch(vertebrateStudyCol -> getCell(tableCellNode.getRow(), vertebrateStudyCol).containsString(value)))
+                .map(TableCell::getEntities)
+                .flatMap(Collection::stream);
    }


@@ -100,9 +110,13 @@ public class Table implements SemanticNode {
     */
    public Stream<RedactionEntity> streamEntitiesWhereRowHasHeaderAndAnyValue(String header, List<String> values) {

-        List<Integer> colsWithHeader = streamHeaders().filter(headerNode -> headerNode.containsString(header)).map(TableCell::getCol).toList();
+        List<Integer> colsWithHeader = streamHeaders().filter(headerNode -> headerNode.containsString(header))
+                .map(TableCell::getCol)
+                .toList();
        return streamTableCells().filter(tableCellNode -> colsWithHeader.stream()
-                .anyMatch(colWithHeader -> getCell(tableCellNode.getRow(), colWithHeader).containsAnyString(values))).map(TableCell::getEntities).flatMap(Collection::stream);
+                        .anyMatch(colWithHeader -> getCell(tableCellNode.getRow(), colWithHeader).containsAnyString(values)))
+                .map(TableCell::getEntities)
+                .flatMap(Collection::stream);
    }


@@ -114,12 +128,15 @@ public class Table implements SemanticNode {
     */
    public Stream<RedactionEntity> streamEntitiesWhereRowContainsEntitiesOfType(List<String> types) {

-        List<Integer> rowsWithEntityOfType = IntStream.range(0, numberOfRows)
-                .boxed()
-                .filter(rowNumber -> streamEntityTypesInRow(rowNumber).anyMatch(existingType -> types.stream().anyMatch(typeToCheck -> typeToCheck.equals(existingType))))
+        List<Integer> rowsWithEntityOfType = IntStream.range(0, numberOfRows).boxed()
+                .filter(rowNumber -> streamEntityTypesInRow(rowNumber).anyMatch(existingType -> types.stream()
+                        .anyMatch(typeToCheck -> typeToCheck.equals(existingType))))
                .toList();

-        return rowsWithEntityOfType.stream().flatMap(this::streamRow).map(TableCell::getEntities).flatMap(Collection::stream);
+        return rowsWithEntityOfType.stream()
+                .flatMap(this::streamRow)
+                .map(TableCell::getEntities)
+                .flatMap(Collection::stream);
    }


@@ -131,18 +148,24 @@ public class Table implements SemanticNode {
     */
    public Stream<RedactionEntity> streamEntitiesWhereRowContainsNoEntitiesOfType(List<String> types) {

-        List<Integer> rowsWithNoEntityOfType = IntStream.range(0, numberOfRows)
-                .boxed()
-                .filter(rowNumber -> streamEntityTypesInRow(rowNumber).noneMatch(existingType -> types.stream().anyMatch(typeToCheck -> typeToCheck.equals(existingType))))
+        List<Integer> rowsWithNoEntityOfType = IntStream.range(0, numberOfRows).boxed()
+                .filter(rowNumber -> streamEntityTypesInRow(rowNumber).noneMatch(existingType -> types.stream()
+                        .anyMatch(typeToCheck -> typeToCheck.equals(existingType))))
                .toList();

-        return rowsWithNoEntityOfType.stream().flatMap(this::streamRow).map(TableCell::getEntities).flatMap(Collection::stream);
+        return rowsWithNoEntityOfType.stream()
+                .flatMap(this::streamRow)
+                .map(TableCell::getEntities)
+                .flatMap(Collection::stream);
    }


    private Stream<String> streamEntityTypesInRow(Integer rowNumber) {

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


@@ -159,7 +182,8 @@ public class Table implements SemanticNode {
            throw new IllegalArgumentException(format("row %d, col %d is out of bounds for number of rows of %d and number of cols %d", row, col, numberOfRows, numberOfCols));
        }
        int idx = row * numberOfCols + col;
-        return (TableCell) documentTree.getEntryById(treeId).getChildren().get(idx).getNode();
+        return (TableCell) documentTree.getEntryById(treeId).getChildren()
+                .get(idx).getNode();
    }


@@ -196,7 +220,8 @@ public class Table implements SemanticNode {
     */
    public Stream<TableCell> streamCol(int col) {

-        return IntStream.range(0, numberOfRows).boxed().map(row -> getCell(row, col));
+        return IntStream.range(0, numberOfRows).boxed()
+                .map(row -> getCell(row, col));
    }


@@ -208,9 +233,11 @@ public class Table implements SemanticNode {
     */
    public Stream<TableCell> streamRow(int row) {

-        return IntStream.range(0, numberOfCols).boxed().map(col -> getCell(row, col));
+        return IntStream.range(0, numberOfCols).boxed()
+                .map(col -> getCell(row, col));
    }

+
    /**
     * Streams all TableCells row-wise and filters them with header == true.
     *
@@ -231,7 +258,8 @@ public class Table implements SemanticNode {
     */
    public Stream<TableCell> streamHeadersForCell(int row, int col) {

-        return Stream.concat(streamRow(row), streamCol(col)).filter(TableCell::isHeader);
+        return Stream.concat(streamRow(row), streamCol(col))
+                .filter(TableCell::isHeader);
    }


@@ -304,7 +332,9 @@ public class Table implements SemanticNode {
    public TextBlock getTextBlock() {

        if (textBlock == null) {
-            textBlock = streamAllSubNodes().filter(SemanticNode::isLeaf).map(SemanticNode::getLeafTextBlock).collect(new TextBlockCollector());
+            textBlock = streamAllSubNodes().filter(SemanticNode::isLeaf)
+                    .map(SemanticNode::getLeafTextBlock)
+                    .collect(new TextBlockCollector());
        }
        return textBlock;
    }
@@ -315,6 +345,8 @@ public class Table implements SemanticNode {

        return treeId.toString() + ": " + NodeType.TABLE + ": #cols: " + numberOfCols + ", #rows: " + numberOfRows + ", " + this.getTextBlock().buildSummary();
    }
+
+
    @Override
    public Map<Page, Rectangle2D> getBBox() {

@@ -323,4 +355,5 @@ public class Table implements SemanticNode {
        }
        return bBoxCache;
    }
+
 }
@@ -7,6 +7,7 @@ import java.util.List;
 import java.util.Map;
 import java.util.Set;

+import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
@@ -26,6 +27,8 @@ import lombok.experimental.FieldDefaults;
@FieldDefaults(level = AccessLevel.PRIVATE)
 public class TableCell implements GenericSemanticNode {

+    @Builder.Default
+    Set<LayoutEngine> engines = new HashSet<>(Set.of(LayoutEngine.ALGORITHM));
    List<Integer> treeId;
    int row;
    int col;
@@ -4,18 +4,21 @@ import java.awt.geom.Rectangle2D;

 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.ImageType;

+import lombok.AllArgsConstructor;
 import lombok.Data;
 import lombok.NonNull;
 import lombok.RequiredArgsConstructor;

@Data
@RequiredArgsConstructor
+@AllArgsConstructor
 public class ClassifiedImage {

    @NonNull
    private Rectangle2D position;
    @NonNull
    private ImageType imageType;
+    private boolean sourceByAi;
    private boolean isAppendedToSection;
    @NonNull
    private boolean hasTransparency;
@@ -1,6 +1,7 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.text;

 import org.apache.pdfbox.text.TextPosition;
+import org.springframework.beans.BeanUtils;

 import com.fasterxml.jackson.annotation.JsonIgnore;

@@ -49,13 +50,7 @@ public class RedTextPosition {
    public static RedTextPosition fromTextPosition(TextPosition textPosition) {

        var pos = new RedTextPosition();
-        pos.setRotation(textPosition.getRotation());
-        pos.setPageHeight(textPosition.getPageHeight());
-        pos.setPageWidth(textPosition.getPageWidth());
-        pos.setUnicode(textPosition.getUnicode());
-        pos.setDir(textPosition.getDir());
-        pos.setWidthOfSpace(textPosition.getWidthOfSpace());
-        pos.setFontSizeInPt(textPosition.getFontSizeInPt());
+        BeanUtils.copyProperties(textPosition, pos);
        pos.setFontName(textPosition.getFont().getName());

        pos.setFontSizeInPt(textPosition.getFontSizeInPt());
@@ -190,12 +190,6 @@ public class TextPositionSequence implements CharSequence {
    }


-    public float getTextHeightNoPadding() {
-
-        return textPositions.get(0).getHeightDir();
-    }
-
-
    public float getTextHeight() {

        return textPositions.get(0).getHeightDir() + HEIGHT_PADDING;
@@ -0,0 +1,92 @@
+package com.knecon.fforesight.service.layoutparser.processor.python_api.adapter;
+
+import java.awt.geom.Rectangle2D;
+import java.util.ArrayList;
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+
+import org.springframework.stereotype.Service;
+
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.ImageType;
+import com.knecon.fforesight.service.layoutparser.processor.model.image.ClassifiedImage;
+import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.VisualLayoutParsingBox;
+import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.VisualLayoutParsingResponse;
+import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.VisualLayoutParsingResult;
+
+import lombok.RequiredArgsConstructor;
+import lombok.extern.slf4j.Slf4j;
+
+@Service
+@RequiredArgsConstructor
+@Slf4j
+public class VisualLayoutParsingAdapter {
+
+    private static String SIGNATURES = "signature";
+
+
+    public Map<Integer, List<VisualLayoutParsingResult>> buildExtractedTablesPerPage(VisualLayoutParsingResponse visualLayoutParsingResponse) {
+
+        Map<Integer, List<VisualLayoutParsingResult>> tableCells = new HashMap<>();
+        visualLayoutParsingResponse.getData()
+                .forEach(tableData -> tableCells.computeIfAbsent(tableData.getPage_idx(), tableCell -> new ArrayList<>()).addAll(convertTableCells(tableData.getBoxes())));
+
+        return tableCells;
+
+    }
+
+
+    public Map<Integer, List<ClassifiedImage>> buildExtractedSignaturesPerPage(VisualLayoutParsingResponse visualLayoutParsingResponse) {
+
+        Map<Integer, List<ClassifiedImage>> signatures = new HashMap<>();
+        if (visualLayoutParsingResponse.getData() != null) {
+            visualLayoutParsingResponse.getData()
+                    .forEach(tableData -> signatures.computeIfAbsent(tableData.getPage_idx() + 1, tableCell -> new ArrayList<>())
+                            .addAll(convertSignatures(tableData.getPage_idx(), tableData.getBoxes())));
+        }
+        return signatures;
+
+    }
+
+
+    public List<VisualLayoutParsingResult> convertTableCells(List<VisualLayoutParsingBox> tableObjects) {
+
+        List<VisualLayoutParsingResult> parsedTableCells = new ArrayList<>();
+
+        tableObjects.stream().forEach(t -> {
+            VisualLayoutParsingResult result = new VisualLayoutParsingResult();
+            result.setX0(t.getBox().getX1());
+            result.setX1(t.getBox().getX2());
+            result.setY0(t.getBox().getY1());
+            result.setY1(t.getBox().getY2());
+            result.setWidth(result.getX1() - result.getX0());
+            result.setHeight(result.getY1() - result.getY0());
+            result.setLabel(t.getLabel());
+            parsedTableCells.add(result);
+        });
+
+        return parsedTableCells;
+
+    }
+
+
+    public List<ClassifiedImage> convertSignatures(int pageNumber, List<VisualLayoutParsingBox> tableObjects) {
+
+        List<ClassifiedImage> signatures = new ArrayList<>();
+
+        tableObjects.stream().forEach(t -> {
+            if (t.getLabel().equals(SIGNATURES)) {
+                ClassifiedImage signature = new ClassifiedImage(new Rectangle2D.Float(t.getBox().getX1(),
+                        t.getBox().getY1(),
+                        t.getBox().getX2() - t.getBox().getX1(),
+                        t.getBox().getY2() - t.getBox().getY1()), ImageType.SIGNATURE, true, false, false, pageNumber);
+
+                signatures.add(signature);
+            }
+        });
+
+        return signatures;
+
+    }
+
+}
@@ -0,0 +1,18 @@
+package com.knecon.fforesight.service.layoutparser.processor.python_api.model.table;
+
+import lombok.AllArgsConstructor;
+import lombok.Builder;
+import lombok.Data;
+import lombok.NoArgsConstructor;
+
+@Data
+@Builder
+@NoArgsConstructor
+@AllArgsConstructor
+public class VisualLayoutParsingBox {
+
+    private VisualLayoutParsingBoxValue box;
+    private String label;
+    private float probability;
+
+}
@@ -0,0 +1,19 @@
+package com.knecon.fforesight.service.layoutparser.processor.python_api.model.table;
+
+import lombok.AllArgsConstructor;
+import lombok.Builder;
+import lombok.Data;
+import lombok.NoArgsConstructor;
+
+@Data
+@Builder
+@NoArgsConstructor
+@AllArgsConstructor
+public class VisualLayoutParsingBoxValue {
+
+    private float x1;
+    private float y1;
+    private float x2;
+    private float y2;
+
+}
@@ -0,0 +1,21 @@
+package com.knecon.fforesight.service.layoutparser.processor.python_api.model.table;
+
+import java.util.List;
+
+
+import lombok.AllArgsConstructor;
+import lombok.Builder;
+import lombok.Data;
+import lombok.NoArgsConstructor;
+
+@Data
+@Builder
+@NoArgsConstructor
+@AllArgsConstructor
+public class VisualLayoutParsingData {
+
+    private int page_idx;
+
+    private List<VisualLayoutParsingBox> boxes;
+
+}
@@ -0,0 +1,23 @@
+package com.knecon.fforesight.service.layoutparser.processor.python_api.model.table;
+
+import java.util.List;
+
+import lombok.AllArgsConstructor;
+import lombok.Builder;
+import lombok.Data;
+import lombok.NoArgsConstructor;
+
+@Data
+@Builder
+@NoArgsConstructor
+@AllArgsConstructor
+public class VisualLayoutParsingResponse {
+
+    private String dossierId;
+    private String fileId;
+    private String targetFileExtension;
+    private String responseFileExtension;
+    private String X_TENANT_ID;
+    private List<VisualLayoutParsingData> data;
+
+}
@@ -0,0 +1,22 @@
+package com.knecon.fforesight.service.layoutparser.processor.python_api.model.table;
+
+import lombok.AllArgsConstructor;
+import lombok.Builder;
+import lombok.Data;
+import lombok.NoArgsConstructor;
+
+@Data
+@Builder
+@NoArgsConstructor
+@AllArgsConstructor
+public class VisualLayoutParsingResult {
+
+    private float x0;
+    private float y0;
+    private float x1;
+    private float y1;
+    private float width;
+    private float height;
+    private String label;
+
+}
@@ -43,6 +43,7 @@ public class SectionsBuilderService {
        for (ClassificationPage page : document.getPages()) {
            List<TextPageBlock> header = new ArrayList<>();
            List<TextPageBlock> footer = new ArrayList<>();
+            List<TextPageBlock> unclassifiedText = new ArrayList<>();
            for (AbstractPageBlock current : page.getTextBlocks()) {

                if (current.getClassification() == null) {
@@ -61,6 +62,11 @@ public class SectionsBuilderService {
                    continue;
                }

+                if (current.getClassification().equals(PageBlockType.OTHER)) {
+                    unclassifiedText.add((TextPageBlock) current);
+                    continue;
+                }
+
                if (prev != null && current.getClassification().isHeadline() && !prev.getClassification().isHeadline() || !document.isHeadlines()) {
                    ClassificationSection chunkBlock = buildTextBlock(chunkWords, lastHeadline);
                    chunkBlock.setHeadline(lastHeadline);
@@ -88,6 +94,9 @@ public class SectionsBuilderService {
            if (!footer.isEmpty()) {
                footers.add(new ClassificationFooter(footer));
            }
+            if (!unclassifiedText.isEmpty()) {
+                unclassifiedTexts.add(new UnclassifiedText(unclassifiedText));
+            }
        }

        ClassificationSection chunkBlock = buildTextBlock(chunkWords, lastHeadline);
@@ -61,8 +61,8 @@ public class DocuMineBlockificationService {
            boolean newLineAfterSplit = prev != null && word.getMinYDirAdj() != prev.getMinYDirAdj() && wasSplitted && splitX1 != word.getMinXDirAdj();
            boolean isSplitByRuling = isSplitByRuling(minX, minY, maxX, maxY, word, horizontalRulingLines, verticalRulingLines);
            boolean splitByDir = prev != null && !prev.getDir().equals(word.getDir());
-            boolean splitByOtherFontAndOtherY = prev != null && Math.abs(prev.getMaxYDirAdj() - word.getMaxYDirAdj()) > word.getTextHeight() * 0.2 && (word.getFontStyle()
-                    .contains("bold") && !prev.getFontStyle().contains("bold") || prev.getFontStyle().contains("bold") && !word.getFontStyle().contains("bold"));
+            boolean splitByOtherFontAndOtherY = prev != null && prev.getMaxYDirAdj() != word.getMaxYDirAdj() && (word.getFontStyle().contains("bold") && !prev.getFontStyle()
+                    .contains("bold") || prev.getFontStyle().contains("bold") && !word.getFontStyle().contains("bold"));

            Matcher matcher = pattern.matcher(chunkWords.stream().collect(Collectors.joining(" ")).toString());
            boolean startsOnSameX = Math.abs(minX - word.getMinXDirAdj()) < 5 && matcher.matches();
@@ -2,7 +2,6 @@ package com.knecon.fforesight.service.layoutparser.processor.services.blockifica

 import static java.util.stream.Collectors.toSet;

-import java.awt.geom.Point2D;
 import java.util.ArrayList;
 import java.util.Comparator;
 import java.util.Iterator;
@@ -14,13 +13,10 @@ import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBl
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
 import com.knecon.fforesight.service.layoutparser.processor.model.FloatFrequencyCounter;
 import com.knecon.fforesight.service.layoutparser.processor.model.Orientation;
-import com.knecon.fforesight.service.layoutparser.processor.model.table.Cell;
-import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRulings;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.Ruling;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.StringFrequencyCounter;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
-import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
 import com.knecon.fforesight.service.layoutparser.processor.utils.RulingTextDirAdjustUtil;

@SuppressWarnings("all")
@@ -35,12 +31,12 @@ public class RedactManagerBlockificationService {
     * This method must use text direction adjusted postions (DirAdj). Where {0,0} is on the upper left. Never try to change this!
     * Rulings (TablePageBlock lines) must be adjusted to the text directions as well, when checking if a block is split by a ruling.
     *
-     * @param textPositions The words of a page.
+     * @param textPositions         The words of a page.
+     * @param horizontalRulingLines Horizontal table lines.
+     * @param verticalRulingLines   Vertical table lines.
     * @return Page object that contains the Textblock and text statistics.
     */
-    public ClassificationPage blockify(List<TextPositionSequence> textPositions, List<Cell> emptyCells) {
-
-        CleanRulings usedRulings = RectangleTransformations.extractRulings(emptyCells);
+    public ClassificationPage blockify(List<TextPositionSequence> textPositions, List<Ruling> horizontalRulingLines, List<Ruling> verticalRulingLines) {

        int indexOnPage = 0;
        List<TextPositionSequence> chunkWords = new ArrayList<>();
@@ -58,7 +54,7 @@ public class RedactManagerBlockificationService {
            boolean splitByX = prev != null && maxX + 50 < word.getMinXDirAdj() && prev.getMinYDirAdj() == word.getMinYDirAdj();
            boolean xIsBeforeFirstX = prev != null && word.getMinXDirAdj() < minX;
            boolean newLineAfterSplit = prev != null && word.getMinYDirAdj() != prev.getMinYDirAdj() && wasSplitted && splitX1 != word.getMinXDirAdj();
-            boolean isSplitByRuling = isSplitByRuling(minX, minY, maxX, maxY, word, usedRulings.getHorizontal(), usedRulings.getVertical());
+            boolean isSplitByRuling = isSplitByRuling(minX, minY, maxX, maxY, word, horizontalRulingLines, verticalRulingLines);
            boolean splitByDir = prev != null && !prev.getDir().equals(word.getDir());

            if (prev != null && (lineSeparation || startFromTop || splitByX || splitByDir || isSplitByRuling)) {
@@ -5,6 +5,7 @@ import java.util.Locale;
 import java.util.regex.Matcher;
 import java.util.regex.Pattern;

+import com.knecon.fforesight.service.layoutparser.processor.utils.MarkedContentUtils;
 import org.springframework.stereotype.Service;

 import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
@@ -12,7 +13,6 @@ import com.knecon.fforesight.service.layoutparser.processor.model.Classification
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
 import com.knecon.fforesight.service.layoutparser.processor.model.PageBlockType;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
-import com.knecon.fforesight.service.layoutparser.processor.utils.MarkedContentUtils;
 import com.knecon.fforesight.service.layoutparser.processor.utils.PositionUtils;

 import lombok.RequiredArgsConstructor;
@@ -23,7 +23,7 @@ import lombok.extern.slf4j.Slf4j;
@RequiredArgsConstructor
 public class DocuMineClassificationService {

-    private static final Pattern pattern = Pattern.compile("^([1-9]\\d?\\.){1,3}\\d{1,2}\\.?\\s[0-9A-Za-z \\[\\]]{2,50}", Pattern.CASE_INSENSITIVE);
+    private static final Pattern pattern = Pattern.compile("^(\\d{1,2}\\.){1,3}\\d{1,2}\\.?\\s[0-9A-Za-z \\[\\]]{2,50}", Pattern.CASE_INSENSITIVE);
    private static final Pattern pattern2 = Pattern.compile("\\p{L}{3,}", Pattern.CASE_INSENSITIVE);
    private static final Pattern pattern3 = Pattern.compile("^(\\d{1,1}\\.){1,3}\\d{1,2}\\.?\\s[a-z]{1,2}\\/[a-z]{1,2}.*");

@@ -64,54 +64,46 @@ public class DocuMineClassificationService {
            return;
        }
        if (MarkedContentUtils.intersects(textBlock, page.getMarkedContentBboxPerType(), MarkedContentUtils.HEADER)
-            || PositionUtils.isOverBodyTextFrame(bodyTextFrame, textBlock, page.getRotation()) && (document.getFontSizeCounter().getMostPopular() == null
-                                                                                                   || textBlock.getHighestFontSize() <= document.getFontSizeCounter()
-                .getMostPopular())) {
+                || PositionUtils.isOverBodyTextFrame(bodyTextFrame, textBlock, page.getRotation()) && (document.getFontSizeCounter()
+                .getMostPopular() == null || textBlock.getHighestFontSize() <= document.getFontSizeCounter().getMostPopular())
+        ) {
            textBlock.setClassification(PageBlockType.HEADER);

        } else if (MarkedContentUtils.intersects(textBlock, page.getMarkedContentBboxPerType(), MarkedContentUtils.FOOTER)
-                   || PositionUtils.isUnderBodyTextFrame(bodyTextFrame, textBlock, page.getRotation()) && (document.getFontSizeCounter().getMostPopular() == null
-                                                                                                           || textBlock.getHighestFontSize() <= document.getFontSizeCounter()
-                .getMostPopular())) {
+                || PositionUtils.isUnderBodyTextFrame(bodyTextFrame, textBlock, page.getRotation()) && (document.getFontSizeCounter()
+                .getMostPopular() == null || textBlock.getHighestFontSize() <= document.getFontSizeCounter().getMostPopular())
+        ) {
            textBlock.setClassification(PageBlockType.FOOTER);
-        } else if (page.getPageNumber() == 1 && (PositionUtils.getHeightDifferenceBetweenChunkWordAndDocumentWord(textBlock, document.getTextHeightCounter().getMostPopular()) > 2.5
-                                                 && textBlock.getHighestFontSize() > document.getFontSizeCounter().getMostPopular() || page.getTextBlocks().size() == 1)) {
+        } else if (page.getPageNumber() == 1 && (PositionUtils.getHeightDifferenceBetweenChunkWordAndDocumentWord(textBlock,
+                document.getTextHeightCounter().getMostPopular()) > 2.5 && textBlock.getHighestFontSize() > document.getFontSizeCounter().getMostPopular() || page.getTextBlocks()
+                .size() == 1)) {
            if (!Pattern.matches("[0-9]+", textBlock.toString())) {
                textBlock.setClassification(PageBlockType.TITLE);
            }
-        } else if (textBlock.getText().length() > 5
-                   && (textBlock.getMostPopularWordHeight() > document.getTextHeightCounter().getMostPopular()
-                       || textBlock.getMostPopularWordFontSize() > document.getFontSizeCounter().getMostPopular())
-                   && PositionUtils.getApproxLineCount(textBlock) < 5.9
-                   && (textBlock.getMostPopularWordStyle().contains("bold") && Character.isDigit(textBlock.toString().charAt(0)) && matcher2.reset().find() && !textBlock.toString()
-                .contains(":")
-                       || textBlock.toString().equals(textBlock.toString().toUpperCase(Locale.ROOT)) && matcher2.reset().find() && !textBlock.toString().contains(":")
-                       || textBlock.toString().startsWith("APPENDIX")
-                       || textBlock.toString().startsWith("FIGURE")
-                       || textBlock.toString().startsWith("TABLE"))
-                   && !textBlock.toString().endsWith(":")
-                   && matcher2.reset().find()) {
+        } else if (textBlock.getText().length() > 5 && (textBlock.getMostPopularWordHeight() > document.getTextHeightCounter()
+                .getMostPopular() || textBlock.getMostPopularWordFontSize() > document.getFontSizeCounter().getMostPopular()) && PositionUtils.getApproxLineCount(textBlock) < 5.9
+
+                && (textBlock.getMostPopularWordStyle().contains("bold") && Character.isDigit(textBlock.toString().charAt(0)) && !matcher2.matches() && !textBlock.toString()
+                .contains(":") || textBlock.toString().equals(textBlock.toString().toUpperCase(Locale.ROOT)) && !matcher2.matches() && !textBlock.toString()
+                .contains(":") || textBlock.toString().startsWith("APPENDIX") || textBlock.toString().startsWith("FIGURE") || textBlock.toString()
+                .startsWith("TABLE")) && !textBlock.toString().endsWith(":") && matcher2.find()) {
            textBlock.setClassification(PageBlockType.getHeadlineType(1));
            document.setHeadlines(true);

-        } else if (matcher.reset().find() && PositionUtils.getApproxLineCount(textBlock) < 2.9 && matcher2.reset().find() && !matcher3.reset().matches()) {
+        } else if (matcher.find() && PositionUtils.getApproxLineCount(textBlock) < 2.9 && matcher2.find() && !matcher3.matches()) {
            textBlock.setClassification(PageBlockType.getHeadlineType(2));
            document.setHeadlines(true);
-        } else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock)
-                   && textBlock.getMostPopularWordFontSize() == document.getFontSizeCounter().getMostPopular()
-                   && textBlock.getMostPopularWordStyle().equals("bold")
-                   && !document.getFontStyleCounter().getMostPopular().equals("bold")) {
+        } else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock) && textBlock.getMostPopularWordFontSize() == document.getFontSizeCounter()
+                .getMostPopular() && textBlock.getMostPopularWordStyle().equals("bold") && !document.getFontStyleCounter().getMostPopular().equals("bold")) {
            textBlock.setClassification(PageBlockType.PARAGRAPH_BOLD);
-        } else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock)
-                   && textBlock.getMostPopularWordFont().equals(document.getFontCounter().getMostPopular())
-                   && textBlock.getMostPopularWordStyle().equals(document.getFontStyleCounter().getMostPopular())
-                   && textBlock.getMostPopularWordFontSize() == document.getFontSizeCounter().getMostPopular()) {
+        } else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock) && textBlock.getMostPopularWordFont()
+                .equals(document.getFontCounter().getMostPopular()) && textBlock.getMostPopularWordStyle()
+                .equals(document.getFontStyleCounter().getMostPopular()) && textBlock.getMostPopularWordFontSize() == document.getFontSizeCounter().getMostPopular()) {
            textBlock.setClassification(PageBlockType.PARAGRAPH);
-        } else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock)
-                   && textBlock.getMostPopularWordFontSize() == document.getFontSizeCounter().getMostPopular()
-                   && textBlock.getMostPopularWordStyle().equals("italic")
-                   && !document.getFontStyleCounter().getMostPopular().equals("italic")
-                   && PositionUtils.getApproxLineCount(textBlock) < 2.9) {
+        } else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock) && textBlock.getMostPopularWordFontSize() == document.getFontSizeCounter()
+                .getMostPopular() && textBlock.getMostPopularWordStyle().equals("italic") && !document.getFontStyleCounter()
+                .getMostPopular()
+                .equals("italic") && PositionUtils.getApproxLineCount(textBlock) < 2.9) {
            textBlock.setClassification(PageBlockType.PARAGRAPH_ITALIC);
        } else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock)) {
            textBlock.setClassification(PageBlockType.PARAGRAPH_UNKNOWN);
@@ -7,12 +7,14 @@ import static java.util.stream.Collectors.toList;
 import java.awt.geom.Rectangle2D;
 import java.util.ArrayList;
 import java.util.HashMap;
+import java.util.HashSet;
 import java.util.LinkedList;
 import java.util.List;
 import java.util.Map;
 import java.util.NoSuchElementException;
 import java.util.Set;

+import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
 import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationDocument;
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationFooter;
@@ -49,14 +51,9 @@ public class DocumentGraphFactory {
        Document documentGraph = new Document();
        Context context = new Context(documentGraph);

-        document.getPages()
-                .forEach(context::buildAndAddPageWithCounter);
-        document.getSections()
-                .stream()
-                .flatMap(section -> section.getImages()
-                        .stream())
-                .forEach(image -> context.getImages().add(image));
-        addSections(document, context, documentGraph);
+        document.getPages().forEach(context::buildAndAddPageWithCounter);
+        document.getSections().stream().flatMap(section -> section.getImages().stream()).forEach(image -> context.getImages().add(image));
+        addSections(document, context);
        addHeaderAndFooterToEachPage(document, context);

        documentGraph.setNumberOfPages(context.pages.size());
@@ -67,10 +64,9 @@ public class DocumentGraphFactory {
    }


-    private void addSections(ClassificationDocument classificationDocument, Context context, Document document) {
+    private void addSections(ClassificationDocument document, Context context) {

-        classificationDocument.getSections()
-                .forEach(section -> SectionNodeFactory.addSection(null, section.getNonEmptyPageBlocks(), section.getImages(), context, document));
+        document.getSections().forEach(section -> SectionNodeFactory.addSection(null, section.getNonEmptyPageBlocks(), section.getImages(), context));
    }


@@ -80,11 +76,9 @@ public class DocumentGraphFactory {

        GenericSemanticNode node;
        if (originalTextBlock.isHeadline()) {
-            node = Headline.builder().documentTree(context.getDocumentTree())
-                    .build();
+            node = Headline.builder().documentTree(context.getDocumentTree()).build();
        } else {
-            node = Paragraph.builder().documentTree(context.getDocumentTree())
-                    .build();
+            node = Paragraph.builder().documentTree(context.getDocumentTree()).build();
        }

        page.getMainBody().add(node);
@@ -99,29 +93,25 @@ public class DocumentGraphFactory {
    }


-    public void addImage(GenericSemanticNode parent, ClassifiedImage image, Context context) {
-
-        Image imageNode = createImage(image, context);
-        List<Integer> treeId = context.getDocumentTree().createNewChildEntryAndReturnId(parent, imageNode);
-        imageNode.setTreeId(treeId);
-        imageNode.setLeafTextBlock(context.textBlockFactory.emptyTextBlock(parent, context, context.getPage(image.getPage())));
-    }
-
-
-    private Image createImage(ClassifiedImage image, Context context) {
+    public void addImage(Section section, ClassifiedImage image, Context context) {

        Rectangle2D position = image.getPosition();
        Page page = context.getPage(image.getPage());
-        Image imageNode = Image.builder()
+        var imageBuilder = Image.builder()
                .id(IdBuilder.buildId(Set.of(page), List.of(position)))
                .imageType(image.getImageType())
                .position(position)
                .transparent(image.isHasTransparency())
                .page(page)
-                .documentTree(context.getDocumentTree())
-                .build();
+                .documentTree(context.getDocumentTree());
+        if (image.isSourceByAi()) {
+            imageBuilder.engines(new HashSet<>(Set.of(LayoutEngine.AI)));
+        }
+        Image imageNode = imageBuilder.build();
        page.getMainBody().add(imageNode);
-        return imageNode;
+
+        List<Integer> treeId = context.getDocumentTree().createNewChildEntryAndReturnId(section, imageNode);
+        imageNode.setTreeId(treeId);
    }


@@ -161,7 +151,10 @@ public class DocumentGraphFactory {

        Page page = context.getPage(textBlocks.get(0).getPage());
        Footer footer = Footer.builder().documentTree(context.getDocumentTree()).build();
-        AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock(TextPositionOperations.mergeTextPositionSequence(textBlocks), footer, context, page);
+        AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock(TextPositionOperations.mergeAndSortTextPositionSequenceByYThenX(textBlocks),
+                footer,
+                context,
+                page);
        List<Integer> tocId = context.getDocumentTree().createNewMainEntryAndReturnId(footer);
        footer.setTreeId(tocId);
        footer.setLeafTextBlock(textBlock);
@@ -173,7 +166,7 @@ public class DocumentGraphFactory {

        Page page = context.getPage(textBlocks.get(0).getPage());
        Header header = Header.builder().documentTree(context.getDocumentTree()).build();
-        AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock(TextPositionOperations.mergeTextPositionSequence(textBlocks), header, 0, page);
+        AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock(TextPositionOperations.mergeAndSortTextPositionSequenceByYThenX(textBlocks), header, 0, page);
        List<Integer> tocId = context.getDocumentTree().createNewMainEntryAndReturnId(header);
        header.setTreeId(tocId);
        header.setLeafTextBlock(textBlock);
@@ -184,8 +177,7 @@ public class DocumentGraphFactory {
    private void addEmptyFooter(int pageIndex, Context context) {

        Page page = context.getPage(pageIndex);
-        Footer footer = Footer.builder().documentTree(context.getDocumentTree())
-                .build();
+        Footer footer = Footer.builder().documentTree(context.getDocumentTree()).build();
        AtomicTextBlock textBlock = context.textBlockFactory.emptyTextBlock(footer, context, page);
        List<Integer> tocId = context.getDocumentTree().createNewMainEntryAndReturnId(footer);
        footer.setTreeId(tocId);
@@ -197,8 +189,7 @@ public class DocumentGraphFactory {
    private void addEmptyHeader(int pageIndex, Context context) {

        Page page = context.getPage(pageIndex);
-        Header header = Header.builder().documentTree(context.getDocumentTree())
-                .build();
+        Header header = Header.builder().documentTree(context.getDocumentTree()).build();
        AtomicTextBlock textBlock = context.textBlockFactory.emptyTextBlock(header, 0, page);
        List<Integer> tocId = context.getDocumentTree().createNewMainEntryAndReturnId(header);
        header.setTreeId(tocId);
@@ -11,7 +11,6 @@ import java.util.Map;
 import java.util.Set;

 import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
 import com.knecon.fforesight.service.layoutparser.processor.model.image.ClassifiedImage;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.TablePageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
@@ -25,46 +24,27 @@ import lombok.experimental.UtilityClass;
@UtilityClass
 public class SectionNodeFactory {

-    public void addSection(GenericSemanticNode parentNode,
-                           List<AbstractPageBlock> pageBlocks,
-                           List<ClassifiedImage> images,
-                           DocumentGraphFactory.Context context,
-                           Document document) {
-
-        // This is for the case where we have images on a page without any text/footer/header.
-        // The pageBlocks list is empty, but we still need to add those images to the document.
-        if (!images.isEmpty() && pageBlocks.isEmpty()) {
-            images.stream()
-                    .distinct()
-                    .forEach(image -> DocumentGraphFactory.addImage(document, image, context));
-            return;
-        }
+    public void addSection(GenericSemanticNode parentNode, List<AbstractPageBlock> pageBlocks, List<ClassifiedImage> images, DocumentGraphFactory.Context context) {

        if (pageBlocks.isEmpty()) {
            return;
        }
-
-        Map<Integer, List<AbstractPageBlock>> blocksPerPage = pageBlocks.stream()
-                .collect(groupingBy(AbstractPageBlock::getPage));
-        Section section = Section.builder().documentTree(context.getDocumentTree())
-                .build();
+        Map<Integer, List<AbstractPageBlock>> blocksPerPage = pageBlocks.stream().collect(groupingBy(AbstractPageBlock::getPage));
+        Section section = Section.builder().documentTree(context.getDocumentTree()).build();

        context.getSections().add(section);
-        blocksPerPage.keySet()
-                .forEach(pageNumber -> addSectionNodeToPageNode(context, section, pageNumber));
+        blocksPerPage.keySet().forEach(pageNumber -> addSectionNodeToPageNode(context, section, pageNumber));

        section.setTreeId(getTreeId(parentNode, context, section));

-        addFirstHeadlineDirectlyToSection(pageBlocks, context, section, document);
+        addFirstHeadlineDirectlyToSection(pageBlocks, context, section);
        if (containsTablesAndTextBlocks(pageBlocks)) {
-            splitPageBlocksIntoSubSections(pageBlocks).forEach(subSectionPageBlocks -> addSection(section, subSectionPageBlocks, emptyList(), context, document));
+            splitPageBlocksIntoSubSections(pageBlocks).forEach(subSectionPageBlocks -> addSection(section, subSectionPageBlocks, emptyList(), context));
        } else {
-            addTablesAndParagraphsAndHeadlinesToSection(pageBlocks, context, section, document);
+            addTablesAndParagraphsAndHeadlinesToSection(pageBlocks, context, section);
        }

-        images.stream()
-                .distinct()
-                .forEach(image -> DocumentGraphFactory.addImage(section, image, context));
+        images.stream().distinct().forEach(image -> DocumentGraphFactory.addImage(section, image, context));
    }


@@ -78,16 +58,16 @@ public class SectionNodeFactory {
    }


-    private void addFirstHeadlineDirectlyToSection(List<AbstractPageBlock> pageBlocks, DocumentGraphFactory.Context context, Section section, Document document) {
+    private void addFirstHeadlineDirectlyToSection(List<AbstractPageBlock> pageBlocks, DocumentGraphFactory.Context context, Section section) {

        if (pageBlocks.get(0).isHeadline()) {
-            addTablesAndParagraphsAndHeadlinesToSection(List.of(pageBlocks.get(0)), context, section, document);
+            addTablesAndParagraphsAndHeadlinesToSection(List.of(pageBlocks.get(0)), context, section);
            pageBlocks.remove(0);
        }
    }


-    private void addTablesAndParagraphsAndHeadlinesToSection(List<AbstractPageBlock> pageBlocks, DocumentGraphFactory.Context context, Section section, Document document) {
+    private void addTablesAndParagraphsAndHeadlinesToSection(List<AbstractPageBlock> pageBlocks, DocumentGraphFactory.Context context, Section section) {

        Set<AbstractPageBlock> alreadyMerged = new HashSet<>();
        List<AbstractPageBlock> remainingBlocks = new LinkedList<>(pageBlocks);
@@ -106,7 +86,7 @@ public class SectionNodeFactory {
            } else if (abstractPageBlock instanceof TablePageBlock tablePageBlock) {
                List<TablePageBlock> tablesToMerge = TableMergingUtility.findConsecutiveTablesWithSameColCountAndSameHeaders(tablePageBlock, remainingBlocks);
                alreadyMerged.addAll(tablesToMerge);
-                TableNodeFactory.addTable(section, tablesToMerge, context, document);
+                TableNodeFactory.addTable(section, tablesToMerge, context);
            } else {
                throw new RuntimeException(format("Unhandled AbstractPageBlockType %s!", abstractPageBlock.getClass()));
            }
@@ -116,9 +96,7 @@ public class SectionNodeFactory {

    private boolean containsTablesAndTextBlocks(List<AbstractPageBlock> pageBlocks) {

-        return pageBlocks.stream()
-                       .anyMatch(pageBlock -> pageBlock instanceof TablePageBlock) && pageBlocks.stream()
-                       .anyMatch(pageBlock -> pageBlock instanceof TextPageBlock);
+        return pageBlocks.stream().anyMatch(pageBlock -> pageBlock instanceof TablePageBlock) && pageBlocks.stream().anyMatch(pageBlock -> pageBlock instanceof TextPageBlock);
    }


@@ -134,9 +112,7 @@ public class SectionNodeFactory {

        List<List<AbstractPageBlock>> splitList = splitIntoCoherentList(pageBlocks);
        movePrecedingHeadlineToTableList(splitList);
-        return splitList.stream()
-                .filter(list -> !list.isEmpty())
-                .toList();
+        return splitList.stream().filter(list -> !list.isEmpty()).toList();
    }


@@ -157,8 +133,7 @@ public class SectionNodeFactory {

    private boolean listIsTablesOnly(List<AbstractPageBlock> abstractPageBlocks) {

-        return abstractPageBlocks.stream()
-                .allMatch(abstractPageBlock -> abstractPageBlock instanceof TablePageBlock);
+        return abstractPageBlocks.stream().allMatch(abstractPageBlock -> abstractPageBlock instanceof TablePageBlock);
    }


@@ -8,7 +8,6 @@ import java.util.Set;
 import java.util.stream.Collectors;

 import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
 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;
@@ -28,26 +27,23 @@ public class TableNodeFactory {
    public final double TABLE_CELL_MERGE_CONTENTS_SIZE_THRESHOLD = 0.05;


-    public void addTable(GenericSemanticNode parentNode, List<TablePageBlock> tablesToMerge, DocumentGraphFactory.Context context, Document document) {
+    public void addTable(GenericSemanticNode parentNode, List<TablePageBlock> tablesToMerge, DocumentGraphFactory.Context context) {

        setPageNumberInCells(tablesToMerge);
-        Set<Page> pages = tablesToMerge.stream()
-                .map(AbstractPageBlock::getPage)
-                .map(context::getPage)
-                .collect(Collectors.toSet());
-        List<List<Cell>> mergedRows = tablesToMerge.stream()
-                .map(TablePageBlock::getRows)
-                .flatMap(Collection::stream)
-                .toList();
+        Set<Page> pages = tablesToMerge.stream().map(AbstractPageBlock::getPage).map(context::getPage).collect(Collectors.toSet());
+        List<List<Cell>> mergedRows = tablesToMerge.stream().map(TablePageBlock::getRows).flatMap(Collection::stream).toList();

-        Table table = Table.builder().documentTree(context.getDocumentTree()).numberOfCols(mergedRows.isEmpty() ? 0 : mergedRows.get(0).size()).numberOfRows(mergedRows.size())
+        Table table = Table.builder()
+                .documentTree(context.getDocumentTree())
+                .numberOfCols(mergedRows.isEmpty() ? 0 : mergedRows.get(0).size())
+                .numberOfRows(mergedRows.size())
                .build();

        pages.forEach(page -> addTableToPage(page, parentNode, table));

        List<Integer> treeId = context.getDocumentTree().createNewChildEntryAndReturnId(parentNode, table);
        table.setTreeId(treeId);
-        addTableCells(mergedRows, table, context, document);
+        addTableCells(mergedRows, table, context);

        ifTableHasNoHeadersSetFirstRowAsHeaders(table);
    }
@@ -67,8 +63,7 @@ public class TableNodeFactory {

    private void setPageNumberInTextBlocksWithPageNumberSetTo0(TablePageBlock table, Cell cell) {

-        cell.getTextBlocks()
-                .stream()//
+        cell.getTextBlocks().stream()//
                .filter(tb -> tb.getPage() == 0)//
                .forEach(tb -> tb.setPage(table.getPage()));
    }
@@ -87,32 +82,28 @@ public class TableNodeFactory {

    private void ifTableHasNoHeadersSetFirstRowAsHeaders(Table table) {

-        if (table.streamHeaders()
-                .findAny().isEmpty()) {
-            table.streamRow(0)
-                    .forEach(tableCellNode -> tableCellNode.setHeader(true));
+        if (table.streamHeaders().findAny().isEmpty()) {
+            table.streamRow(0).forEach(tableCellNode -> tableCellNode.setHeader(true));
        }
    }


-    private void addTableCells(List<List<Cell>> rows, Table table, DocumentGraphFactory.Context context, Document document) {
+    private void addTableCells(List<List<Cell>> rows, Table table, DocumentGraphFactory.Context context) {

        for (int rowIndex = 0; rowIndex < rows.size(); rowIndex++) {
            for (int colIndex = 0; colIndex < rows.get(rowIndex).size(); colIndex++) {
-                addTableCell(rows.get(rowIndex)
-                                     .get(colIndex), rowIndex, colIndex, table, context, document);
+                addTableCell(rows.get(rowIndex).get(colIndex), rowIndex, colIndex, table, context);
            }
        }
    }


    @SuppressWarnings("PMD.UnusedPrivateMethod") // PMD actually flags this wrong
-    private void addTableCell(Cell cell, int rowIndex, int colIndex, Table tableNode, DocumentGraphFactory.Context context, Document document) {
+    private void addTableCell(Cell cell, int rowIndex, int colIndex, Table tableNode, DocumentGraphFactory.Context context) {

        Page page = context.getPage(cell.getPageNumber());

-        TableCell tableCell = TableCell.builder().documentTree(context.getDocumentTree()).row(rowIndex).col(colIndex).header(cell.isHeaderCell()).bBox(cell.getBounds2D())
-                .build();
+        TableCell tableCell = TableCell.builder().documentTree(context.getDocumentTree()).row(rowIndex).col(colIndex).header(cell.isHeaderCell()).bBox(cell.getBounds2D()).build();
        page.getMainBody().add(tableCell);

        List<Integer> treeId = context.getDocumentTree().createNewTableChildEntryAndReturnId(tableNode, tableCell);
@@ -122,26 +113,16 @@ public class TableNodeFactory {
        if (cell.getTextBlocks().isEmpty()) {
            tableCell.setLeafTextBlock(context.getTextBlockFactory().emptyTextBlock(tableNode, context, page));
        } else if (cell.getTextBlocks().size() == 1) {
-            textBlock = context.getTextBlockFactory()
-                    .buildAtomicTextBlock(cell.getTextBlocks()
-                                                  .get(0).getSequences(), tableCell, context, page);
+            textBlock = context.getTextBlockFactory().buildAtomicTextBlock(cell.getTextBlocks().get(0).getSequences(), tableCell, context, page);
            tableCell.setLeafTextBlock(textBlock);
        } else if (firstTextBlockIsHeadline(cell)) {
-            SectionNodeFactory.addSection(tableCell,
-                                          cell.getTextBlocks()
-                                                  .stream()
-                                                  .map(tb -> (AbstractPageBlock) tb)
-                                                  .toList(),
-                                          emptyList(),
-                                          context,
-                                          document);
+            SectionNodeFactory.addSection(tableCell, cell.getTextBlocks().stream().map(tb -> (AbstractPageBlock) tb).toList(), emptyList(), context);
        } else if (cellAreaIsSmallerThanPageAreaTimesThreshold(cell, page)) {
            List<TextPositionSequence> sequences = TextPositionOperations.mergeAndSortTextPositionSequenceByYThenX(cell.getTextBlocks());
            textBlock = context.getTextBlockFactory().buildAtomicTextBlock(sequences, tableCell, context, page);
            tableCell.setLeafTextBlock(textBlock);
        } else {
-            cell.getTextBlocks()
-                    .forEach(tb -> DocumentGraphFactory.addParagraphOrHeadline(tableCell, tb, context, emptyList()));
+            cell.getTextBlocks().forEach(tb -> DocumentGraphFactory.addParagraphOrHeadline(tableCell, tb, context, emptyList()));
        }
    }

@@ -154,8 +135,7 @@ public class TableNodeFactory {

    private boolean firstTextBlockIsHeadline(Cell cell) {

-        return cell.getTextBlocks()
-                .get(0).isHeadline();
+        return cell.getTextBlocks().get(0).isHeadline();
    }

 }
@@ -80,6 +80,7 @@ public class DocumentDataMapper {
                .treeId(toPrimitiveIntArray(entry.getTreeId()))
                .children(entry.getChildren().stream().map(DocumentDataMapper::toEntryData).toList())
                .type(entry.getType())
+                .engines(entry.getNode().getEngines())
                .atomicBlockIds(atomicTextBlocks)
                .pageNumbers(entry.getNode().getPages().stream().map(Page::getNumber).map(Integer::longValue).toArray(Long[]::new))
                .properties(properties)
@@ -17,6 +17,7 @@ import java.util.stream.Stream;

 import org.springframework.stereotype.Service;

+import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
 import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Page;
@@ -63,43 +64,53 @@ public class LayoutGridService {
    @Observed(name = "ViewerDocumentService", contextualName = "create-viewer-document")
    public void addLayoutGrid(File originFile, Document document, File destinationFile, boolean layerVisibilityDefaultValue) {

-        LayoutGrid layoutGrid = createLayoutGrid(document);
+        Visualizations layoutGrid = this.addLayoutGrid(document, layerVisibilityDefaultValue, false);
+        Visualizations visualLayoutGrid = this.addLayoutGrid(document, layerVisibilityDefaultValue, true);

-        viewerDocumentService.addVisualizationsOnPage(originFile,
-                destinationFile,
-                Visualizations.builder()
-                        .layer(ContentStreams.KNECON_LAYOUT)
-                        .visualizationsOnPages(layoutGrid.getVisualizationsPerPages())
-                        .layerVisibilityDefaultValue(layerVisibilityDefaultValue)
-                        .build());
+        viewerDocumentService.addVisualizationsOnPage(originFile, destinationFile, List.of(layoutGrid, visualLayoutGrid));
    }


-    private LayoutGrid createLayoutGrid(Document document) {
+    @SneakyThrows
+    @Observed(name = "ViewerDocumentService", contextualName = "create-viewer-document")
+    public Visualizations addLayoutGrid(Document document, boolean layerVisibilityDefaultValue, boolean visualParsingGrid) {
+
+        LayoutGrid layoutGrid = createLayoutGrid(document, visualParsingGrid);
+
+        return Visualizations.builder()
+                .layer(visualParsingGrid ? ContentStreams.KNECON_VISUAL_PARSING : ContentStreams.KNECON_LAYOUT)
+                .visualizationsOnPages(layoutGrid.getVisualizationsPerPages())
+                .layerVisibilityDefaultValue(layerVisibilityDefaultValue)
+                .build();
+    }
+
+
+    private LayoutGrid createLayoutGrid(Document document, boolean visualParsingGrid) {

        LayoutGrid layoutGrid = new LayoutGrid(document.getNumberOfPages());
-        document.streamAllSubNodes().forEach(semanticNode -> {
-            Color color = switch (semanticNode.getType()) {
-                case PARAGRAPH -> PARAGRAPH_COLOR;
-                case TABLE -> TABLE_COLOR;
-                case SECTION -> SECTION_COLOR;
-                case HEADLINE -> HEADLINE_COLOR;
-                case HEADER, FOOTER -> HEADER_COLOR;
-                case IMAGE -> IMAGE_COLOR;
-                default -> null;
-            };
-            if (isNotSectionOrTableCellOrDocument(semanticNode)) {
-                addAsRectangle(semanticNode, layoutGrid, color);
-            }
-            if (semanticNode.getType().equals(NodeType.SECTION)) {
-                addSection(semanticNode, layoutGrid, color);
-            }
-            if (semanticNode.getType().equals(NodeType.TABLE)) {
-                Table table = (Table) semanticNode;
-                addInnerTableLines(table, layoutGrid);
-            }
-        });
-
+        document.streamAllSubNodes()
+                .filter(node -> (node.getEngines().contains(LayoutEngine.AI) && visualParsingGrid) || (node.getEngines().contains(LayoutEngine.ALGORITHM) && !visualParsingGrid))
+                .forEach(semanticNode -> {
+                    Color color = switch (semanticNode.getType()) {
+                        case PARAGRAPH -> PARAGRAPH_COLOR;
+                        case TABLE -> TABLE_COLOR;
+                        case SECTION -> SECTION_COLOR;
+                        case HEADLINE -> HEADLINE_COLOR;
+                        case HEADER, FOOTER -> HEADER_COLOR;
+                        case IMAGE -> IMAGE_COLOR;
+                        default -> null;
+                    };
+                    if (isNotSectionOrTableCellOrDocument(semanticNode)) {
+                        addAsRectangle(semanticNode, layoutGrid, color);
+                    }
+                    if (semanticNode.getType().equals(NodeType.SECTION)) {
+                        addSection(semanticNode, layoutGrid, color);
+                    }
+                    if (semanticNode.getType().equals(NodeType.TABLE)) {
+                        Table table = (Table) semanticNode;
+                        addInnerTableLines(table, layoutGrid);
+                    }
+                });
        return layoutGrid;
    }

@@ -110,13 +121,11 @@ public class LayoutGridService {
            return;
        }
        for (Page page : table.getPages()) {
-
            Optional<Integer> optionalFirstRowOnPage = table.streamCol(0).filter(tableCell -> tableCell.isOnPage(page.getNumber())).map(TableCell::getRow).findFirst();
            if (optionalFirstRowOnPage.isEmpty()) {
                continue;
            }
            int firstRowOnPage = optionalFirstRowOnPage.get();
-
            Stream<Double> xStream = switch (page.getRotation()) {
                case 90 -> streamBBoxOfCellsOnPage(table.streamCol(0), page).map(RectangularShape::getMinX);
                case 180 -> streamBBoxOfCellsOnPage(table.streamRow(firstRowOnPage), page).map(RectangularShape::getMaxX);
@@ -125,7 +134,6 @@ public class LayoutGridService {
            };
            List<Double> xs = xStream.collect(Collectors.toList());
            xs.remove(0);
-
            Stream<Double> yStream = switch (page.getRotation()) {
                case 90 -> streamBBoxOfCellsOnPage(table.streamRow(firstRowOnPage), page).map(RectangularShape::getMinY);
                case 180 -> streamBBoxOfCellsOnPage(table.streamCol(0), page).map(RectangularShape::getMinY);
@@ -135,7 +143,7 @@ public class LayoutGridService {
            List<Double> ys = yStream.collect(Collectors.toList());
            ys.remove(0);

-            Rectangle2D tableBBox = table.getBBox().get(page);
+            Rectangle2D tableBBox = table.getBBox().get(table.getFirstPage());
            List<ColoredLine> coloredLines = layoutGrid.getVisualizationsPerPages().get(page.getNumber() - 1).getColoredLines();
            xs.forEach(x -> {
                Line2D line = new Line2D.Double(new Point2D.Double(x, tableBBox.getMaxY()), new Point2D.Double(x, tableBBox.getMinY()));
@@ -3,10 +3,8 @@ package com.knecon.fforesight.service.layoutparser.processor.utils;
 import static java.lang.String.format;

 import java.awt.geom.Area;
-import java.awt.geom.Point2D;
 import java.awt.geom.Rectangle2D;
 import java.awt.geom.RectangularShape;
-import java.util.ArrayList;
 import java.util.Collections;
 import java.util.LinkedList;
 import java.util.List;
@@ -22,8 +20,6 @@ import org.apache.pdfbox.pdmodel.common.PDRectangle;
 import com.iqser.red.service.persistence.service.v1.api.shared.model.redactionlog.Point;
 import com.iqser.red.service.persistence.service.v1.api.shared.model.redactionlog.Rectangle;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.AtomicTextBlock;
-import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRulings;
-import com.knecon.fforesight.service.layoutparser.processor.model.table.Ruling;

 import lombok.AllArgsConstructor;
 import lombok.NoArgsConstructor;
@@ -153,27 +149,7 @@ public class RectangleTransformations {
                previousRectangle = currentRectangle;
            }
        }
-        return rectangleListsWithGaps.stream()
-                .map(RectangleTransformations::rectangle2DBBox)
-                .toList();
-    }
-
-
-    public static CleanRulings extractRulings(List<? extends Rectangle2D.Float> rectangles) {
-
-        // Underlined or strikethrough are also in rulings but we dont want to split blocks with them so we use cells.
-        List<Ruling> horizontalRulings = new ArrayList<>();
-        List<Ruling> verticalRulings = new ArrayList<>();
-
-        rectangles.forEach(rectangle -> {
-            horizontalRulings.add(new Ruling(new Point2D.Float(rectangle.x, rectangle.y), new Point2D.Float(rectangle.x + rectangle.width, rectangle.y)));
-            horizontalRulings.add(new Ruling(new Point2D.Float(rectangle.x, rectangle.y + rectangle.height),
-                                             new Point2D.Float(rectangle.x + rectangle.width, rectangle.y + rectangle.height)));
-            verticalRulings.add(new Ruling(new Point2D.Float(rectangle.x, rectangle.y), new Point2D.Float(rectangle.x, rectangle.y + rectangle.height)));
-            verticalRulings.add(new Ruling(new Point2D.Float(rectangle.x + rectangle.width, rectangle.y),
-                                           new Point2D.Float(rectangle.x + rectangle.width, rectangle.y + rectangle.height)));
-        });
-        return CleanRulings.builder().vertical(verticalRulings).horizontal(horizontalRulings).build();
+        return rectangleListsWithGaps.stream().map(RectangleTransformations::rectangle2DBBox).toList();
    }


@@ -23,10 +23,4 @@ public class TextPositionOperations {
        return sequence;
    }

-
-    public static List<TextPositionSequence> mergeTextPositionSequence(List<TextPageBlock> textBlocks) {
-
-        return textBlocks.stream().flatMap(tb -> tb.getSequences().stream()).collect(Collectors.toList());
-    }
-
 }
@@ -28,13 +28,15 @@ import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPosit
 *
 * @author Ben Litchfield
 */
-public class TextPositionSequenceComparator implements Comparator<TextPositionSequence> {
-
+public class TextPositionSequenceComparator implements Comparator<TextPositionSequence>
+{
    @Override
-    public int compare(TextPositionSequence pos1, TextPositionSequence pos2) {
+    public int compare(TextPositionSequence pos1, TextPositionSequence pos2)
+    {
        // only compare text that is in the same direction
        int cmp1 = Float.compare(pos1.getDir().getDegrees(), pos2.getDir().getDegrees());
-        if (cmp1 != 0) {
+        if (cmp1 != 0)
+        {
            return cmp1;
        }

@@ -46,19 +48,25 @@ public class TextPositionSequenceComparator implements Comparator<TextPositionSe
        float pos2YBottom = pos2.getMaxYDirAdj();

        // note that the coordinates have been adjusted so 0,0 is in upper left
-        float pos1YTop = pos1YBottom - pos1.getTextHeightNoPadding();
-        float pos2YTop = pos2YBottom - pos2.getTextHeightNoPadding();
+        float pos1YTop = pos1YBottom - pos1.getTextHeight();
+        float pos2YTop = pos2YBottom - pos2.getTextHeight();

        float yDifference = Math.abs(pos1YBottom - pos2YBottom);

        // we will do a simple tolerance comparison
-        if (yDifference < .1 || pos2YBottom >= pos1YTop && pos2YBottom <= pos1YBottom || pos1YBottom >= pos2YTop && pos1YBottom <= pos2YBottom) {
+        if (yDifference < .1 ||
+                pos2YBottom >= pos1YTop && pos2YBottom <= pos1YBottom ||
+                pos1YBottom >= pos2YTop && pos1YBottom <= pos2YBottom)
+        {
            return Float.compare(x1, x2);
-        } else if (pos1YBottom < pos2YBottom) {
+        }
+        else if (pos1YBottom < pos2YBottom)
+        {
            return -1;
-        } else {
+        }
+        else
+        {
            return 1;
        }
    }
-
 }
@@ -27,6 +27,7 @@ import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Do
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.image.ImageServiceResponse;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.TableServiceResponse;
+import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.VisualLayoutParsingResponse;
 import com.knecon.fforesight.service.layoutparser.processor.services.factory.DocumentGraphFactory;
 import com.knecon.fforesight.service.layoutparser.processor.services.mapper.TaasDocumentDataMapper;
 import com.knecon.fforesight.service.layoutparser.processor.utils.DrawingOptions;
@@ -51,6 +52,7 @@ public class BdrJsonBuildTest extends AbstractTest {
                file,
                new ImageServiceResponse(),
                new TableServiceResponse(),
+                new VisualLayoutParsingResponse(),
                file.toString()));
    }

@@ -33,6 +33,7 @@ import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Do
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.SemanticNode;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.image.ImageServiceResponse;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.TableServiceResponse;
+import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.VisualLayoutParsingResponse;
 import com.knecon.fforesight.service.layoutparser.processor.services.factory.DocumentGraphFactory;
 import com.knecon.fforesight.tenantcommons.TenantsClient;

@@ -98,6 +99,7 @@ public class HeadlinesGoldStandardIntegrationTest {
                pdfFileResource.getFile(),
                new ImageServiceResponse(),
                new TableServiceResponse(),
+                new VisualLayoutParsingResponse(),
                filePath));

        var foundHeadlines = documentGraph.streamAllSubNodes()
@@ -28,20 +28,7 @@ public class LayoutparserEnd2EndTest extends AbstractTest {
        prepareStorage("files/bdr/Wie weiter bei Kristeneinrichtungen.pdf");
        LayoutParsingRequest layoutParsingRequest = buildDefaultLayoutParsingRequest(LayoutParsingType.REDACT_MANAGER);
        LayoutParsingFinishedEvent finishedEvent = layoutParsingPipeline.parseLayoutAndSaveFilesToStorage(layoutParsingRequest);
-        Arrays.stream(finishedEvent.message().split("\n"))
-                .forEach(log::info);
-    }
-
-
-    @Test
-    @SneakyThrows
-    public void testLayoutParserEndToEnd_RED_8747() {
-
-        prepareStorage("files/SinglePages/MergedEntities.pdf");
-        LayoutParsingRequest layoutParsingRequest = buildDefaultLayoutParsingRequest(LayoutParsingType.REDACT_MANAGER);
-        LayoutParsingFinishedEvent finishedEvent = layoutParsingPipeline.parseLayoutAndSaveFilesToStorage(layoutParsingRequest);
-        Arrays.stream(finishedEvent.message().split("\n"))
-                .forEach(log::info);
+        Arrays.stream(finishedEvent.message().split("\n")).forEach(log::info);
    }

 }
@@ -16,6 +16,7 @@ import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsi
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.image.ImageServiceResponse;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.TableServiceResponse;
+import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.VisualLayoutParsingResponse;
 import com.knecon.fforesight.service.layoutparser.processor.services.factory.DocumentGraphFactory;
 import com.knecon.fforesight.service.layoutparser.processor.services.mapper.DocumentDataMapper;
 import com.knecon.fforesight.service.layoutparser.server.utils.BuildDocumentTest;
@@ -58,6 +59,7 @@ public class DocumentGraphJsonWritingTest extends BuildDocumentTest {
                filename.toFile(),
                new ImageServiceResponse(),
                new TableServiceResponse(),
+                new VisualLayoutParsingResponse(),
                filename.toFile().toString()));

        DocumentData documentData = DocumentDataMapper.toDocumentData(documentGraph);
@@ -12,6 +12,7 @@ import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsi
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.image.ImageServiceResponse;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.TableServiceResponse;
+import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.VisualLayoutParsingResponse;
 import com.knecon.fforesight.service.layoutparser.processor.services.factory.DocumentGraphFactory;
 import com.knecon.fforesight.service.layoutparser.processor.services.visualization.LayoutGridService;
 import com.knecon.fforesight.service.layoutparser.server.utils.BuildDocumentTest;
@@ -56,7 +57,7 @@ public class ViewerDocumentTest extends BuildDocumentTest {
                                                                       documentFile,
                                                                       new ImageServiceResponse(),
                                                                       tableResponse,
-                                                                       Path.of(fileName).getFileName().toFile().toString());
+                                                                       new VisualLayoutParsingResponse(),Path.of(fileName).getFileName().toFile().toString());
        ViewerDocumentService viewerDocumentService = new ViewerDocumentService(null);
        LayoutGridService layoutGridService = new LayoutGridService(viewerDocumentService);
        Document document = DocumentGraphFactory.buildDocumentGraph(classificationDocument);
@@ -1,7 +1,6 @@
 package com.knecon.fforesight.service.layoutparser.server.segmentation;

 import static org.assertj.core.api.Assertions.assertThat;
-import static org.junit.jupiter.api.Assertions.assertTrue;

 import java.awt.geom.Rectangle2D;
 import java.io.File;
@@ -26,20 +25,15 @@ import com.fasterxml.jackson.databind.ObjectMapper;
 import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingType;
 import com.knecon.fforesight.service.layoutparser.processor.LayoutParsingPipeline;
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationDocument;
-import com.knecon.fforesight.service.layoutparser.processor.model.PageContents;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.ImageType;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
 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.python_api.model.image.ImageServiceResponse;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.TableServiceResponse;
-import com.knecon.fforesight.service.layoutparser.processor.services.PageContentExtractor;
+import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.VisualLayoutParsingResponse;
 import com.knecon.fforesight.service.layoutparser.processor.services.SectionsBuilderService;
 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.server.utils.AbstractTest;

 import lombok.SneakyThrows;
@@ -66,6 +60,7 @@ public class PdfSegmentationServiceTest extends AbstractTest {
                                                                                          originDocument,
                                                                                          new ImageServiceResponse(),
                                                                                          tableServiceResponse,
+                                                                                          new VisualLayoutParsingResponse(),
                                                                                          "document");

        redactManagerClassificationService.classifyDocument(classificationDocument);
@@ -759,54 +754,6 @@ public class PdfSegmentationServiceTest extends AbstractTest {
    }


-    @Test
-    public void testMergedEntities_Page26() throws IOException {
-
-        ClassPathResource pdfFileResource = new ClassPathResource("files/SinglePages/MergedEntities.pdf");
-
-        ClassificationDocument document = buildClassificationDocument(pdfFileResource.getFile());
-
-        validateTableSize(document, 1);
-
-        validateTable(document, 0, 6, 6, 5, 0);
-
-    }
-
-
-    @Test
-    public void testHeaderAndFooter() throws IOException {
-
-        String fileName = "files/SinglePages/Page1_54 Fludioxonil - EU AIR3 - Document E1 - Listing of Community and Member States MRLs.pdf";
-        String textToSearch = "Annex to Regulation 283/2013 Annex to Regulation 284/2013";
-        ClassPathResource pdfFileResource = new ClassPathResource(fileName);
-
-        List<PageContents> textPositionPerPage = PageContentExtractor.getSortedPageContents(fileName);
-        var textPositions = textPositionPerPage.stream()
-                .flatMap(t -> t.getSortedTextPositionSequences()
-                        .stream()
-                        .map(TextPositionSequence::toString))
-                .collect(Collectors.joining(" "));
-        assertThat(textPositions.contains(textToSearch)).isFalse();
-
-        ClassificationDocument classificationDocument = buildClassificationDocument(pdfFileResource.getFile());
-
-        assertThat(classificationDocument.getHeaders()
-                           .get(0).getTextBlocks().size()).isEqualTo(3);
-        assertThat(classificationDocument.getHeaders()
-                           .get(0).getTextBlocks()
-                           .get(0).getSequences().size()).isEqualTo(8);
-        assertThat(classificationDocument.getHeaders()
-                           .get(0).getTextBlocks()
-                           .get(0).toString()).isEqualTo(textToSearch);
-
-        Document document = DocumentGraphFactory.buildDocumentGraph(classificationDocument);
-
-        TextBlock leafTextBlock = document.getFirstPage().getHeader().getLeafTextBlock();
-        assertTrue(leafTextBlock.getSearchText().contains(textToSearch));
-
-    }
-
-
    @SneakyThrows
    private void toHtml(ClassificationDocument document, String filename) {

@@ -25,6 +25,7 @@ import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Ta
 import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRulings;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.image.ImageServiceResponse;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.TableServiceResponse;
+import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.VisualLayoutParsingResponse;
 import com.knecon.fforesight.service.layoutparser.processor.services.PageContentExtractor;
 import com.knecon.fforesight.service.layoutparser.processor.services.RulingCleaningService;
 import com.knecon.fforesight.service.layoutparser.processor.services.factory.DocumentGraphFactory;
@@ -102,11 +103,13 @@ public class RulingCleaningServiceTest extends BuildDocumentTest {
                filename.toFile(),
                new ImageServiceResponse(),
                new TableServiceResponse(),
+                new VisualLayoutParsingResponse(),
                filename.toFile().toString()));
        Document documentGraphAfter = DocumentGraphFactory.buildDocumentGraph(layoutParsingPipeline.parseLayout(LayoutParsingType.REDACT_MANAGER,
                filename.toFile(),
                new ImageServiceResponse(),
                new TableServiceResponse(),
+                new VisualLayoutParsingResponse(),
                filename.toFile().toString()));
        DocumentData documentDataBefore = DocumentDataMapper.toDocumentData(documentGraphBefore);
        DocumentData documentDataAfter = DocumentDataMapper.toDocumentData(documentGraphAfter);
@@ -20,6 +20,7 @@ import org.springframework.context.annotation.Import;
 import org.springframework.context.annotation.Primary;
 import org.springframework.core.io.ClassPathResource;
 import org.springframework.test.context.junit.jupiter.SpringExtension;
+import org.xmlunit.builder.Input;

 import com.iqser.red.commons.jackson.ObjectMapperFactory;
 import com.iqser.red.storage.commons.service.StorageService;
@@ -51,6 +52,8 @@ public abstract class AbstractTest {
    private RabbitTemplate rabbitTemplate;

    protected final static String ORIGIN_FILE_ID = "origin";
+
+    protected final static String VISUAL_LAYOUT_FILE = "visual";
    protected final static String TABLE_FILE_ID = "table";
    protected final static String IMAGE_FILE_ID = "image";
    protected final static String STRUCTURE_FILE_ID = "structure";
@@ -96,7 +99,7 @@ public abstract class AbstractTest {
    @SneakyThrows
    protected LayoutParsingRequest prepareStorage(String file) {

-        return prepareStorage(file, "cv_table_parsing_response/empty.json", "image_service_response/empty.json");
+        return prepareStorage(file, "cv_table_parsing_response/empty.json", "image_service_response/empty.json","visual_layout_parsing_response/empty.json");
    }


@@ -116,6 +119,7 @@ public abstract class AbstractTest {
                .originFileStorageId(ORIGIN_FILE_ID)
                .tablesFileStorageId(Optional.of(TABLE_FILE_ID))
                .imagesFileStorageId(Optional.of(IMAGE_FILE_ID))
+                .visualLayoutParsingFileId(Optional.of(VISUAL_LAYOUT_FILE))
                .structureFileStorageId(STRUCTURE_FILE_ID)
                .textBlockFileStorageId(TEXT_FILE_ID)
                .positionBlockFileStorageId(POSITION_FILE_ID)
@@ -136,6 +140,16 @@ public abstract class AbstractTest {
        return prepareStorage(pdfFileResource.getInputStream(), cvServiceResponseFileResource.getInputStream(), imageInfoFileResource.getInputStream());
    }

+    @SneakyThrows
+    protected LayoutParsingRequest prepareStorage(String file, String cvServiceResponseFile, String imageInfoFile, String visualLayoutParsingResponseFile) {
+
+        ClassPathResource pdfFileResource = new ClassPathResource(file);
+        ClassPathResource cvServiceResponseFileResource = new ClassPathResource(cvServiceResponseFile);
+        ClassPathResource imageInfoFileResource = new ClassPathResource(imageInfoFile);
+        ClassPathResource visualLayoutParsingResponseResource = new ClassPathResource(visualLayoutParsingResponseFile);
+
+        return prepareStorage(pdfFileResource.getInputStream(), cvServiceResponseFileResource.getInputStream(), imageInfoFileResource.getInputStream(), visualLayoutParsingResponseResource.getInputStream());
+    }

    @SneakyThrows
    protected LayoutParsingRequest prepareStorage(InputStream fileStream, InputStream cvServiceResponseFileStream, InputStream imageInfoStream) {
@@ -147,6 +161,17 @@ public abstract class AbstractTest {
        return buildDefaultLayoutParsingRequest(LayoutParsingType.REDACT_MANAGER);
    }

+    @SneakyThrows
+    protected LayoutParsingRequest prepareStorage(InputStream fileStream, InputStream cvServiceResponseFileStream, InputStream imageInfoStream, InputStream visualLayoutParsingResponseFileStream) {
+
+        storageService.storeObject(TenantContext.getTenantId(), IMAGE_FILE_ID, imageInfoStream);
+        storageService.storeObject(TenantContext.getTenantId(), TABLE_FILE_ID, cvServiceResponseFileStream);
+        storageService.storeObject(TenantContext.getTenantId(), ORIGIN_FILE_ID, fileStream);
+        storageService.storeObject(TenantContext.getTenantId(),VISUAL_LAYOUT_FILE,visualLayoutParsingResponseFileStream );
+
+        return buildDefaultLayoutParsingRequest(LayoutParsingType.REDACT_MANAGER);
+    }
+

    @AfterEach
    public void cleanupStorage() {
@@ -10,6 +10,7 @@ import com.knecon.fforesight.service.layoutparser.processor.LayoutParsingPipelin
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationDocument;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
 import com.knecon.fforesight.service.layoutparser.processor.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 lombok.SneakyThrows;
@@ -25,7 +26,7 @@ public abstract class BuildDocumentTest extends AbstractTest {

        File fileResource = new ClassPathResource(filename).getFile();
        prepareStorage(filename);
-        return layoutParsingPipeline.parseLayout(layoutParsingType, fileResource, layoutParsingStorageService.getImagesFile(IMAGE_FILE_ID), new TableServiceResponse(), filename);
+        return layoutParsingPipeline.parseLayout(layoutParsingType, fileResource, layoutParsingStorageService.getImagesFile(IMAGE_FILE_ID), new TableServiceResponse(), new VisualLayoutParsingResponse(),filename);
    }


@@ -0,0 +1,8 @@
+{
+  "dossierId": "123",
+  "fileId": "123",
+  "targetFileExtension": "ORIGIN.pdf.gz",
+  "responseFileExtension": "EXTRACTED_TABLES.json.gz",
+  "data": [],
+  "X_TENANT_ID": ""
+}
@@ -12,6 +12,8 @@ public class ContentStreams {

    public static Identifier KNECON_LAYOUT = new Identifier("Layout grid", COSName.getPDFName("KNECON_LAYOUT"), true);

+    public static Identifier KNECON_VISUAL_PARSING = new Identifier("Layout grid - visual", COSName.getPDFName("KNECON_VISUAL_PARSING"), true);
+
    public static Identifier KNECON_OCR = new Identifier("OCR", COSName.getPDFName("KNECON_OCR"), false);

    public static Identifier KNECON_OCR_TEXT_DEBUG = new Identifier("OCR Text", COSName.getPDFName("KNECON_OCR_TEXT_DEBUG"), true);
@@ -24,7 +26,14 @@ public class ContentStreams {

    public static Identifier ESCAPE_END = new Identifier("escape start", COSName.getPDFName("ESCAPE_END"), false);

-    public static List<Identifier> allContentStreams = List.of(KNECON_LAYOUT, KNECON_OCR, KNECON_OCR_BBOX_DEBUG, KNECON_OCR_TEXT_DEBUG, OTHER, ESCAPE_START, ESCAPE_END);
+    public static List<Identifier> allContentStreams = List.of(KNECON_LAYOUT,
+                                                               KNECON_VISUAL_PARSING,
+                                                               KNECON_OCR,
+                                                               KNECON_OCR_BBOX_DEBUG,
+                                                               KNECON_OCR_TEXT_DEBUG,
+                                                               OTHER,
+                                                               ESCAPE_START,
+                                                               ESCAPE_END);

    public record Identifier(String name, COSName cosName, boolean optionalContent) {

@@ -17,7 +17,6 @@ import java.util.stream.Collectors;

 import org.apache.pdfbox.Loader;
 import org.apache.pdfbox.cos.COSName;
-import org.apache.pdfbox.pdfwriter.compress.CompressParameters;
 import org.apache.pdfbox.pdmodel.PDDocument;
 import org.apache.pdfbox.pdmodel.PDDocumentCatalog;
 import org.apache.pdfbox.pdmodel.PDPage;
@@ -127,8 +126,8 @@ public class ViewerDocumentService {
                pdDocument = openPDDocument(tmpFile.toFile());
            }
        }
-
        observedIncrementalSave(pdDocument, destinationFile);
+
        pdDocument.close();
        assert tmpFile.toFile().delete();
    }
@@ -283,12 +282,10 @@ public class ViewerDocumentService {

    @SneakyThrows
    private void observedIncrementalSave(PDDocument pdDocument, File outputFile) {
-    /*
-    Sometimes the viewer document is corrupted after saving and missing the content streams on a random page, for the files we viewed it did not seem to happen with incrementalSave. It might only be a timing issue though
-     */
+
        Observation.createNotStarted("ViewerDocumentService", registry).contextualName("incremental-save").observe(() -> {
            try (var out = new FileOutputStream(outputFile)) {
-                pdDocument.save(out, CompressParameters.NO_COMPRESSION);
+                pdDocument.save(out);
            } catch (IOException e) {
                throw new RuntimeException(e);
            }
Author	SHA1	Message	Date
Kilian Schüttler	c324d3815e	Merge branch 'RED-8550-pre-backup' into 'main' RED-8550: Faulty table recognition and text duplication leads to huge sections See merge request fforesight/layout-parser!105	2024-02-28 16:13:56 +01:00
Maverick Studer	74f55a5cbf	RED-8550: Faulty table recognition and text duplication leads to huge sections	2024-02-28 16:13:56 +01:00
Kilian Schüttler	e7bf607663	Merge branch 'hotfix' into 'main' hotfix: double viewerdoc writes in rare cases lead to some contentstreams not being written See merge request fforesight/layout-parser!104	2024-02-26 12:32:14 +01:00
Kilian Schuettler	f4d789311c	hotfix: double viewerdoc writes in rare cases lead to some contentstreams not being written	2024-02-26 12:24:15 +01:00
Yannik Hampe	9817eae897	Merge branch 'RED-8481-fix' into 'main' Red 8481 fix See merge request fforesight/layout-parser!103	2024-02-23 14:08:57 +01:00
yhampe	477f6af886	RED-8481: Use visual layout parsing to detect signatures added a new layer for visual parsing results checkstyle	2024-02-23 14:02:53 +01:00
yhampe	2c171b6a9e	RED-8481: Use visual layout parsing to detect signatures added a new layer for visual parsing results codestyle	2024-02-23 13:55:11 +01:00
yhampe	71477dabde	RED-8481: Use visual layout parsing to detect signatures added a new layer for visual parsing results codestyle	2024-02-23 12:46:51 +01:00
yhampe	a927cbd9dc	RED-8481: Use visual layout parsing to detect signatures added a new layer for visual parsing results fixed tests	2024-02-23 12:38:05 +01:00
yhampe	a1521877d7	RED-8481: Use visual layout parsing to detect signatures added a new layer for visual parsing results added a source label to image properties to enable rules	2024-02-23 12:20:11 +01:00
Maverick Studer	f4b6386e1c	Merge branch 'RED-8550-testing' into 'main' RED-8550: Faulty table recognition and text duplication leads to huge sections See merge request fforesight/layout-parser!102	2024-02-21 13:54:30 +01:00
Maverick Studer	1d64028158	RED-8550: Faulty table recognition and text duplication leads to huge sections	2024-02-21 13:54:30 +01:00
Yannik Hampe	0979a267d4	Merge branch 'RED-8481' into 'main' RED-8481: Use visual layout parsing to detect signatures See merge request fforesight/layout-parser!101	2024-02-15 13:07:06 +01:00
yhampe	cc77d19500	RED-8481: Use visual layout parsing to detect signatures addressed review comments	2024-02-15 13:01:30 +01:00
yhampe	fa048b2fe0	RED-8481: Use visual layout parsing to detect signatures addressed review comments	2024-02-15 12:19:26 +01:00
yhampe	bdf1161c91	RED-8481: Use visual layout parsing to detect signatures addressed review comments	2024-02-15 12:12:23 +01:00
yhampe	b4a225144d	RED-8481: Use visual layout parsing to detect signatures working on failing tests	2024-02-15 10:16:07 +01:00
yhampe	903b1c1fd4	RED-8481: Use visual layout parsing to detect signatures fixed failing tests because of null pointer	2024-02-15 09:27:07 +01:00
yhampe	c3e7582ee3	RED-8481: Use visual layout parsing to detect signatures fixed failing tests because of null pointer	2024-02-14 12:33:36 +01:00
yhampe	cfc5db45cd	RED-8481: Use visual layout parsing to detect signatures fixed failing tests because of null pointer	2024-02-14 12:24:32 +01:00
yhampe	fbd0196719	RED-8481: Use visual layout parsing to detect signatures implemented visuallayoutparsingresult	2024-02-14 12:16:37 +01:00