Merge branch 'RED-8715' into 'main'

RED-8715: Improve NearestNeighbor Algorithm in LayoutParser

See merge request fforesight/layout-parser!114

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

@@ -25,4 +25,5 @@ dependencies {
     implementation("com.fasterxml.jackson.datatype:jackson-datatype-jsr310:${jacksonVersion}")
     implementation("org.springframework.boot:spring-boot-starter-web:3.1.3")
     implementation("org.jgrapht:jgrapht-core:1.5.2")
+    implementation("org.tinspin:tinspin-indexes:2.1.3")
 }

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

@@ -47,6 +47,7 @@ import com.knecon.fforesight.service.layoutparser.processor.services.TableExtrac
 import com.knecon.fforesight.service.layoutparser.processor.services.blockification.DocstrumBlockificationService;
 import com.knecon.fforesight.service.layoutparser.processor.services.blockification.DocuMineBlockificationService;
 import com.knecon.fforesight.service.layoutparser.processor.services.blockification.RedactManagerBlockificationService;
+import com.knecon.fforesight.service.layoutparser.processor.services.classification.ClarifyndClassificationService;
 import com.knecon.fforesight.service.layoutparser.processor.services.classification.DocuMineClassificationService;
 import com.knecon.fforesight.service.layoutparser.processor.services.classification.RedactManagerClassificationService;
 import com.knecon.fforesight.service.layoutparser.processor.services.factory.DocumentGraphFactory;
@@ -88,6 +89,7 @@ public class LayoutParsingPipeline {
     LayoutGridService layoutGridService;
     ObservationRegistry observationRegistry;
     VisualLayoutParsingAdapter visualLayoutParsingAdapter;
+    ClarifyndClassificationService clarifyndClassificationService;
 
 
     public LayoutParsingFinishedEvent parseLayoutAndSaveFilesToStorage(LayoutParsingRequest layoutParsingRequest) throws IOException {
@@ -291,7 +293,7 @@ public class LayoutParsingPipeline {
             if (layoutParsingType == LayoutParsingType.REDACT_MANAGER) {
                 docstrumBlockificationService.combineBlocks(classificationPage);
             } else if (layoutParsingType == LayoutParsingType.CLARIFYND) {
-                docstrumBlockificationService.mergeIntersectingBlocks(classificationPage.getTextBlocks());
+                docstrumBlockificationService.mergeIntersectingBlocks(classificationPage.getTextBlocks(), 0, 6.5f);
             }
 
             buildPageStatistics(classificationPage);
@@ -306,9 +308,10 @@ public class LayoutParsingPipeline {
         bodyTextFrameService.setBodyTextFrames(classificationDocument, layoutParsingType);
         log.info("Classify TextBlocks for {}", identifier);
         switch (layoutParsingType) {
-            case REDACT_MANAGER, REDACT_MANAGER_PARAGRAPH_DEBUG, REDACT_MANAGER_OLD, CLARIFYND, CLARIFYND_PARAGRAPH_DEBUG ->
+            case REDACT_MANAGER, REDACT_MANAGER_PARAGRAPH_DEBUG, REDACT_MANAGER_OLD, CLARIFYND_PARAGRAPH_DEBUG ->
                     redactManagerClassificationService.classifyDocument(classificationDocument);
             case DOCUMINE -> docuMineClassificationService.classifyDocument(classificationDocument);
+            case CLARIFYND -> clarifyndClassificationService.classifyDocument(classificationDocument);
         }
 
         log.info("Building Sections for {}", identifier);

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

@@ -1,10 +1,10 @@
 package com.knecon.fforesight.service.layoutparser.processor.docstrum.service;
 
-import java.util.ArrayList;
-import java.util.Comparator;
 import java.util.List;
 
 import org.springframework.stereotype.Service;
+import org.tinspin.index.Index;
+import org.tinspin.index.kdtree.KDTree;
 
 import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Character;
 import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Neighbor;
@@ -13,115 +13,23 @@ import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Neigh
 public class NearestNeighbourService {
 
     private static final int NUMBER_OF_NEIGHBOURS = 8;
-    private static final double STEP = 16.0;
 
 
     public void findNearestNeighbors(List<Character> characters) {
+        KDTree<Character> kdTree = KDTree.create(2);
+        characters.forEach(c -> kdTree.insert(new double[]{c.getX(), c.getY()}, c));
 
-        if (characters.isEmpty() || characters.size() == 1) {
-            return;
-        }
-
-        characters.sort(Comparator.comparingDouble(Character::getX));
-
-        int maxNeighborCount = NUMBER_OF_NEIGHBOURS;
-        if (characters.size() <= NUMBER_OF_NEIGHBOURS) {
-            maxNeighborCount = characters.size() - 1;
-        }
-
-        for (int i = 0; i < characters.size(); i++) {
-
-            Neighbor[] candidates = new Neighbor[maxNeighborCount + 1];
-            int neighborInsertionIndex = 0;
-            int neighborCount = 0;
-            int start = i;
-            int end = i + 1;
-
-            double distance = Double.POSITIVE_INFINITY;
-
-            for (double searchDistance = 0; searchDistance < distance; ) {
-
-                searchDistance += STEP;
-                boolean newCandidatesFound = false;
-
-                while (start > 0 && characters.get(i).getX() - characters.get(start - 1).getX() < searchDistance) {
-                    start--;
-                    candidates[neighborInsertionIndex] = new Neighbor(characters.get(start), characters.get(i));
-                    neighborCount++;
-                    if (neighborCount > maxNeighborCount) {
-                        neighborInsertionIndex = clearMostDistant(candidates);
-                        neighborCount--;
-                    } else {
-                        neighborInsertionIndex++;
-                    }
-                    newCandidatesFound = true;
-                }
-
-                while (end < characters.size() && characters.get(end).getX() - characters.get(i).getX() < searchDistance) {
-                    candidates[neighborInsertionIndex] = new Neighbor(characters.get(end), characters.get(i));
-                    neighborCount++;
-                    if (neighborCount > maxNeighborCount) {
-                        neighborInsertionIndex = clearMostDistant(candidates);
-                        neighborCount--;
-                    } else {
-                        neighborInsertionIndex++;
-                    }
-                    end++;
-                    newCandidatesFound = true;
-                }
-
-                if (newCandidatesFound && neighborCount >= maxNeighborCount) {
-                    distance = maxDistance(candidates);
-                }
+        for(Character c : characters) {
+            Index.PointIteratorKnn<Character> iterator = kdTree.queryKnn(new double[]{c.getX(), c.getY()}, NUMBER_OF_NEIGHBOURS + 1);
+            // skip the first as this is identity
+            if(iterator.hasNext()) {
+                iterator.next();
             }
-            if (neighborCount < maxNeighborCount) {
-                clearMostDistant(candidates);
-            }
-
-            List<Neighbor> candidatesList = new ArrayList<>(maxNeighborCount);
-            for (Neighbor candidate : candidates) {
-                if (candidate != null) {
-                    candidatesList.add(candidate);
-                }
-            }
-            candidatesList.sort(Comparator.comparingDouble(Neighbor::getDistance));
-            assert candidatesList.size() == maxNeighborCount;
-            characters.get(i).setNeighbors(candidatesList);
-        }
-    }
-
-
-    private double maxDistance(Neighbor[] candidates) {
-
-        double maxDistance = 0;
-        for (Neighbor candidate : candidates) {
-            if (candidate == null) {
-                continue;
-            }
-            if (candidate.getDistance() > maxDistance) {
-                maxDistance = candidate.getDistance();
+            while(iterator.hasNext()) {
+                c.getNeighbors().add(new Neighbor(iterator.next().value(), c));
             }
         }
-        return maxDistance;
-    }
 
-
-    private int clearMostDistant(Neighbor[] candidates) {
-
-        double maxDistance = 0;
-        int maxIndex = 0;
-        for (int i = 0; i < candidates.length; i++) {
-            Neighbor candidate = candidates[i];
-            if (candidate == null) {
-                continue;
-            }
-            if (candidate.getDistance() > maxDistance) {
-                maxDistance = candidate.getDistance();
-                maxIndex = i;
-            }
-        }
-        candidates[maxIndex] = null;
-        return maxIndex;
     }
 
 }

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

@@ -80,7 +80,7 @@ public class DocstrumBlockificationService {
 
     public void combineBlocks(ClassificationPage page) {
 
-        mergeIntersectingBlocks(page.getTextBlocks());
+        mergeIntersectingBlocks(page.getTextBlocks(), 0, 0);
 
         TextPageBlock previous = new TextPageBlock();
         ListIterator<AbstractPageBlock> itty = page.getTextBlocks().listIterator();
@@ -123,7 +123,7 @@ public class DocstrumBlockificationService {
             previous = current;
         }
 
-        mergeIntersectingBlocks(page.getTextBlocks());
+        mergeIntersectingBlocks(page.getTextBlocks(), 0, 0);
     }
 
 
@@ -203,7 +203,7 @@ public class DocstrumBlockificationService {
     }
 
 
-    public void mergeIntersectingBlocks(List<AbstractPageBlock> blocks) {
+    public void mergeIntersectingBlocks(List<AbstractPageBlock> blocks, float xThreshold, float yThreshold) {
 
         ListIterator<AbstractPageBlock> itty = blocks.listIterator();
         Set<AbstractPageBlock> toRemove = new HashSet<>();
@@ -237,7 +237,7 @@ public class DocstrumBlockificationService {
                     continue;
                 }
 
-                if (current.getDir() == inner.getDir() && current.almostIntersects(inner, 0, 0)) {
+                if (current.getDir() == inner.getDir() && current.almostIntersects(inner, yThreshold, xThreshold)) {
 
                     current.getSequences().addAll(inner.getSequences());
                     QuickSort.sort(current.getSequences(), new TextPositionSequenceComparator());

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

@@ -0,0 +1,114 @@
+package com.knecon.fforesight.service.layoutparser.processor.services.classification;
+
+import java.util.List;
+import java.util.regex.Pattern;
+
+import org.springframework.stereotype.Service;
+
+import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
+import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationDocument;
+import com.knecon.fforesight.service.layoutparser.processor.model.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;
+import lombok.extern.slf4j.Slf4j;
+
+@Slf4j
+@Service
+@RequiredArgsConstructor
+public class ClarifyndClassificationService {
+
+    public void classifyDocument(ClassificationDocument document) {
+
+        List<Float> headlineFontSizes = document.getFontSizeCounter().getHighterThanMostPopular();
+
+        log.debug("Document FontSize counters are: {}", document.getFontSizeCounter().getCountPerValue());
+
+        for (ClassificationPage page : document.getPages()) {
+            classifyPage(page, document, headlineFontSizes);
+        }
+    }
+
+
+    private void classifyPage(ClassificationPage page, ClassificationDocument document, List<Float> headlineFontSizes) {
+
+        for (AbstractPageBlock textBlock : page.getTextBlocks()) {
+            if (textBlock instanceof TextPageBlock) {
+                classifyBlock((TextPageBlock) textBlock, page, document, headlineFontSizes);
+            }
+        }
+    }
+
+
+    private void classifyBlock(TextPageBlock textBlock, ClassificationPage page, ClassificationDocument document, List<Float> headlineFontSizes) {
+
+        var bodyTextFrame = page.getBodyTextFrame();
+
+        if (document.getFontSizeCounter().getMostPopular() == null) {
+            textBlock.setClassification(PageBlockType.PARAGRAPH);
+            return;
+        }
+        if (MarkedContentUtils.intersects(textBlock, page.getMarkedContentBboxPerType(), MarkedContentUtils.HEADER) || PositionUtils.isOverBodyTextFrame(bodyTextFrame,
+                textBlock,
+                page.getRotation()) && (document.getFontSizeCounter().getMostPopular() == null || textBlock.getHighestFontSize() <= document.getFontSizeCounter()
+                .getMostPopular())) {
+            textBlock.setClassification(PageBlockType.PARAGRAPH);
+
+        } else if (MarkedContentUtils.intersects(textBlock, page.getMarkedContentBboxPerType(), MarkedContentUtils.FOOTER) || PositionUtils.isUnderBodyTextFrame(bodyTextFrame,
+                textBlock,
+                page.getRotation()) && (document.getFontSizeCounter().getMostPopular() == null || textBlock.getHighestFontSize() <= document.getFontSizeCounter()
+                .getMostPopular())) {
+            textBlock.setClassification(PageBlockType.PARAGRAPH);
+        } else if (page.getPageNumber() == 1 && (PositionUtils.getHeightDifferenceBetweenChunkWordAndDocumentWord(textBlock,
+                document.getTextHeightCounter().getMostPopular()) > 2.5 && textBlock.getHighestFontSize() > document.getFontSizeCounter().getMostPopular() || page.getTextBlocks()
+                .size() == 1)) {
+            if (!Pattern.matches("[0-9]+", textBlock.toString())) {
+                textBlock.setClassification(PageBlockType.TITLE);
+            }
+        } else if (textBlock.getMostPopularWordFontSize() > document.getFontSizeCounter()
+                .getMostPopular() && PositionUtils.getApproxLineCount(textBlock) < 4.9 && (textBlock.getMostPopularWordStyle().equals("bold") || !document.getFontStyleCounter()
+                .getCountPerValue()
+                .containsKey("bold") && textBlock.getMostPopularWordFontSize() > document.getFontSizeCounter().getMostPopular() + 1) && textBlock.getSequences()
+                .get(0)
+                .getTextPositions()
+                .get(0)
+                .getFontSizeInPt() >= textBlock.getMostPopularWordFontSize()) {
+
+            for (int i = 1; i <= headlineFontSizes.size(); i++) {
+                if (textBlock.getMostPopularWordFontSize() == headlineFontSizes.get(i - 1)) {
+                    textBlock.setClassification(PageBlockType.getHeadlineType(i));
+                    document.setHeadlines(true);
+                }
+            }
+        } else if (!textBlock.getText().startsWith("Figure ") && PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock) && textBlock.getMostPopularWordStyle()
+                .equals("bold") && !document.getFontStyleCounter().getMostPopular().equals("bold") && PositionUtils.getApproxLineCount(textBlock) < 2.9 && textBlock.getSequences()
+                .get(0)
+                .getTextPositions()
+                .get(0)
+                .getFontSizeInPt() >= textBlock.getMostPopularWordFontSize()) {
+            textBlock.setClassification(PageBlockType.getHeadlineType(headlineFontSizes.size() + 1));
+            document.setHeadlines(true);
+        } 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()) {
+            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) {
+            textBlock.setClassification(PageBlockType.PARAGRAPH_ITALIC);
+        } else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock)) {
+            textBlock.setClassification(PageBlockType.PARAGRAPH_UNKNOWN);
+        } else {
+            textBlock.setClassification(PageBlockType.PARAGRAPH);
+        }
+    }
+
+}
+

layoutparser-service/layoutparser-service-server/src/main/java/com/knecon/fforesight/service/layoutparser/server/queue/MessageHandler.java

@@ -1,6 +1,9 @@
 package com.knecon.fforesight.service.layoutparser.server.queue;
 
 import java.util.Arrays;
+import java.util.Locale;
+import java.util.Map;
+import java.util.stream.Collectors;
 
 import org.springframework.amqp.AmqpRejectAndDontRequeueException;
 import org.springframework.amqp.core.Message;
@@ -28,6 +31,7 @@ public class MessageHandler {
     private final LayoutParsingPipeline layoutParsingPipeline;
     private final ObjectMapper objectMapper;
     private final RabbitTemplate rabbitTemplate;
+    private final static String X_PIPELINE_PREFIX = "X-PIPE-";
 
 
     @RabbitHandler
@@ -46,14 +50,24 @@ public class MessageHandler {
                     layoutParsingRequest.identifier()));
         }
         LayoutParsingFinishedEvent layoutParsingFinishedEvent = layoutParsingPipeline.parseLayoutAndSaveFilesToStorage(layoutParsingRequest);
-        sendLayoutParsingFinishedEvent(layoutParsingFinishedEvent);
+        sendLayoutParsingFinishedEvent(layoutParsingFinishedEvent, message);
     }
 
 
-    public void sendLayoutParsingFinishedEvent(LayoutParsingFinishedEvent layoutParsingFinishedEvent) {
+    public void sendLayoutParsingFinishedEvent(LayoutParsingFinishedEvent layoutParsingFinishedEvent, Message message) {
 
         Arrays.stream(layoutParsingFinishedEvent.message().split("\n")).forEach(log::info);
-        rabbitTemplate.convertAndSend(LayoutParsingQueueNames.LAYOUT_PARSING_FINISHED_EVENT_QUEUE, layoutParsingFinishedEvent);
+        rabbitTemplate.convertAndSend(LayoutParsingQueueNames.LAYOUT_PARSING_FINISHED_EVENT_QUEUE, layoutParsingFinishedEvent, m -> {
+            var forwardHeaders = message.getMessageProperties()
+                    .getHeaders()
+                    .entrySet()
+                    .stream()
+                    .filter(e -> e.getKey().toUpperCase(Locale.ROOT).startsWith(X_PIPELINE_PREFIX))
+                    .collect(Collectors.toMap(Map.Entry::getKey,
+                            Map.Entry::getValue));
+            m.getMessageProperties().getHeaders().putAll(forwardHeaders);
+            return m;
+        });
     }
 
 }