Merge branch 'RED-9975-bp' into 'release/0.159.x'

RED-9975: improve SuperSection handling

See merge request fforesight/layout-parser!224
This commit is contained in:
Kilian Schüttler 2024-09-11 13:38:04 +02:00
commit f034c5bfa0
39 changed files with 790 additions and 307 deletions

View File

@ -51,6 +51,10 @@ allprojects {
}
}
pmd {
setConsoleOutput(true)
}
publishing {
publications {
create<MavenPublication>(name) {

View File

@ -2,7 +2,6 @@ package com.knecon.fforesight.service.layoutparser.processor;
import static java.lang.String.format;
import java.awt.geom.Point2D;
import java.awt.geom.Rectangle2D;
import java.io.File;
import java.io.IOException;
@ -25,7 +24,7 @@ import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.No
import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingFinishedEvent;
import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingRequest;
import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingType;
import com.knecon.fforesight.service.layoutparser.processor.markdown.MarkdownMapper;
import com.knecon.fforesight.service.layoutparser.processor.services.mapper.MarkdownMapper;
import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationDocument;
import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
@ -143,7 +142,7 @@ public class LayoutParsingPipeline {
log.info("Creating viewer document for {}", layoutParsingRequest.identifier());
layoutGridService.addLayoutGrid(viewerDocumentFile, documentGraph, viewerDocumentFile, false, layoutParsingRequest.visualLayoutParsingFileId().isPresent());
layoutGridService.addLayoutGrid(viewerDocumentFile, documentGraph, viewerDocumentFile, false);
log.info("Storing resulting files for {}", layoutParsingRequest.identifier());

View File

@ -1,6 +1,5 @@
package com.knecon.fforesight.service.layoutparser.processor.model;
import java.util.ArrayList;
import java.util.Collections;
import java.util.HashMap;
import java.util.List;
@ -13,10 +12,14 @@ import lombok.Getter;
public class FloatFrequencyCounter {
Map<Double, Integer> countPerValue = new HashMap<>();
boolean changed;
Double mostPopularCache;
public void add(double value) {
changed = true;
if (!countPerValue.containsKey(value)) {
countPerValue.put(value, 1);
} else {
@ -27,6 +30,8 @@ public class FloatFrequencyCounter {
public void addAll(Map<Double, Integer> otherCounter) {
changed = true;
for (Map.Entry<Double, Integer> entry : otherCounter.entrySet()) {
if (countPerValue.containsKey(entry.getKey())) {
countPerValue.put(entry.getKey(), countPerValue.get(entry.getKey()) + entry.getValue());
@ -39,27 +44,27 @@ public class FloatFrequencyCounter {
public Double getMostPopular() {
Map.Entry<Double, Integer> mostPopular = null;
for (Map.Entry<Double, Integer> entry : countPerValue.entrySet()) {
if (mostPopular == null || entry.getValue() >= mostPopular.getValue()) {
mostPopular = entry;
if (changed) {
Map.Entry<Double, Integer> mostPopular = null;
for (Map.Entry<Double, Integer> entry : countPerValue.entrySet()) {
if (mostPopular == null || entry.getValue() >= mostPopular.getValue()) {
mostPopular = entry;
}
}
mostPopularCache = mostPopular != null ? mostPopular.getKey() : null;
changed = false;
}
return mostPopular != null ? mostPopular.getKey() : null;
return mostPopularCache;
}
public List<Double> getHigherThanMostPopular() {
public List<Double> getValuesInReverseOrder() {
Double mostPopular = getMostPopular();
List<Double> higher = new ArrayList<>();
for (Double value : countPerValue.keySet()) {
if (value > mostPopular) {
higher.add(value);
}
}
return higher.stream().sorted(Collections.reverseOrder()).collect(Collectors.toList());
return countPerValue.keySet()
.stream()
.sorted(Collections.reverseOrder())
.collect(Collectors.toList());
}

View File

@ -3,6 +3,7 @@ package com.knecon.fforesight.service.layoutparser.processor.model;
import java.util.Collections;
import java.util.LinkedList;
import java.util.List;
import java.util.Locale;
import java.util.regex.Matcher;
import java.util.regex.Pattern;
@ -16,10 +17,12 @@ import lombok.experimental.FieldDefaults;
public class SectionIdentifier {
public static Pattern numericalIdentifierPattern = Pattern.compile("^[\\s]?(\\d+)[\\s.,;]?(\\d+)?[\\s.,;]?(\\d+)?[\\s.,;]?(\\d+)?");
public static Pattern alphanumericIdentifierPattern = Pattern.compile("^[\\s]?[A-Za-z][\\s.,;]?(\\d+)[\\s.,;]?(\\d+)?[\\s.,;]?(\\d+)?[\\s.,;]?(\\d+)?[\\s.,;]?");
public enum Format {
EMPTY,
NUMERICAL,
ALPHANUMERIC,
DOCUMENT
}
@ -41,6 +44,10 @@ public class SectionIdentifier {
if (numericalIdentifierMatcher.find()) {
return buildNumericalSectionIdentifier(headline, numericalIdentifierMatcher);
}
Matcher alphanumericIdentifierMatcher = alphanumericIdentifierPattern.matcher(headline);
if (alphanumericIdentifierMatcher.find()) {
return buildAlphanumericSectionIdentifier(headline, alphanumericIdentifierMatcher);
}
// more formats here
return SectionIdentifier.empty();
}
@ -75,7 +82,36 @@ public class SectionIdentifier {
}
identifiers.add(Integer.parseInt(numericalIdentifier.trim()));
}
return new SectionIdentifier(Format.NUMERICAL, identifierString, identifiers.stream().toList(), false);
return new SectionIdentifier(Format.NUMERICAL,
identifierString,
identifiers.stream()
.toList(),
false);
}
private static SectionIdentifier buildAlphanumericSectionIdentifier(String headline, Matcher alphanumericIdentifierMatcher) {
String identifierString = headline.substring(alphanumericIdentifierMatcher.start(), alphanumericIdentifierMatcher.end());
String alphanumericIdentifier = alphanumericIdentifierMatcher.group(0).substring(0, 1).toUpperCase(Locale.ENGLISH);
int mappedCharacterValue = alphanumericIdentifier.charAt(0) - 'A' + 1;
List<Integer> identifiers = new LinkedList<>();
identifiers.add(mappedCharacterValue);
for (int i = 1; i <= 3; i++) {
String numericalIdentifier = alphanumericIdentifierMatcher.group(i);
if (numericalIdentifier == null || numericalIdentifier.equals("0") || numericalIdentifier.isEmpty() || numericalIdentifier.isBlank()) {
break;
}
identifiers.add(Integer.parseInt(numericalIdentifier.trim()));
}
return new SectionIdentifier(Format.ALPHANUMERIC,
identifierString,
identifiers.stream()
.toList(),
false);
}
@ -123,4 +159,22 @@ public class SectionIdentifier {
return identifierString;
}
public boolean isEmpty() {
return this.format.equals(Format.EMPTY);
}
public int level() {
return identifiers.size();
}
protected List<Integer> getIdentifiers() {
return identifiers;
}
}

View File

@ -84,7 +84,7 @@ public abstract class AbstractNodeVisitor implements NodeVisitor {
}
private void visitChildren(SemanticNode semanticNode) {
protected void visitChildren(SemanticNode semanticNode) {
semanticNode.streamChildren()
.forEach(node -> node.accept(this));

View File

@ -25,11 +25,4 @@ public class DuplicatedParagraph extends Paragraph {
}
@Override
public String toString() {
return super.toString();
}
}

View File

@ -5,7 +5,7 @@ import java.util.List;
import java.util.stream.Collectors;
import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.AbstractSemanticNode;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.GenericSemanticNode;
import com.knecon.fforesight.service.layoutparser.processor.model.image.ClassifiedImage;
import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
@ -24,7 +24,7 @@ public class TableOfContentItem {
private List<AbstractPageBlock> sectionBlocks = new ArrayList<>();
private List<ClassifiedImage> images = new ArrayList<>();
private AbstractSemanticNode section;
private GenericSemanticNode section;
public TableOfContentItem(TextPageBlock headline) {
@ -45,8 +45,7 @@ public class TableOfContentItem {
if (parent != null) {
int index = parent.getChildren().indexOf(this);
if (index > 0) {
return parent.getChildren()
.get(index - 1);
return parent.getChildren().get(index - 1);
}
}
return null;
@ -58,8 +57,7 @@ public class TableOfContentItem {
if (parent != null) {
int index = parent.getChildren().indexOf(this);
if (index >= 0 && index < parent.getChildren().size() - 1) {
return parent.getChildren()
.get(index + 1);
return parent.getChildren().get(index + 1);
}
}
return null;
@ -93,17 +91,19 @@ public class TableOfContentItem {
return false;
}
public List<AbstractPageBlock> getNonEmptySectionBlocks() {
return sectionBlocks.stream().filter(pageBlock -> !pageBlock.isEmpty()).collect(Collectors.toList());
return sectionBlocks.stream()
.filter(pageBlock -> !pageBlock.isEmpty())
.collect(Collectors.toList());
}
@Override
public String toString() {
return "OutlineObjectTreeNode{" + "textPageBlock=" + headline + '}';
}
}

View File

@ -2,6 +2,7 @@ package com.knecon.fforesight.service.layoutparser.processor.model.text;
import java.awt.geom.Rectangle2D;
import java.util.ArrayList;
import java.util.Collections;
import java.util.List;
import com.fasterxml.jackson.annotation.JsonIgnore;
@ -45,10 +46,13 @@ public class TextPageBlock extends AbstractPageBlock {
private boolean toDuplicate;
private String text;
private boolean changed;
public TextPageBlock(List<TextPositionSequence> sequences) {
this.sequences = sequences;
this.sequences = new ArrayList<>(sequences);
if (!sequences.isEmpty()) {
calculateFrequencyCounters();
}
@ -56,6 +60,12 @@ public class TextPageBlock extends AbstractPageBlock {
}
public List<TextPositionSequence> getSequences() {
return Collections.unmodifiableList(sequences);
}
public TextDirection getDir() {
return sequences.get(0).getDir();
@ -136,7 +146,7 @@ public class TextPageBlock extends AbstractPageBlock {
public TextPageBlock union(TextPositionSequence r) {
TextPageBlock union = this.copy();
union.getSequences().add(r);
union.add(r);
calculateFrequencyCounters();
calculateBBox();
return union;
@ -146,24 +156,35 @@ public class TextPageBlock extends AbstractPageBlock {
public TextPageBlock union(TextPageBlock r) {
TextPageBlock union = this.copy();
union.getSequences().addAll(r.getSequences());
union.addAll(r.getSequences());
calculateFrequencyCounters();
calculateBBox();
return union;
}
public void add(TextPageBlock r) {
public void add(TextPageBlock textPageBlock) {
sequences.addAll(r.getSequences());
changed = true;
sequences.addAll(textPageBlock.getSequences());
calculateFrequencyCounters();
calculateBBox();
}
public void add(TextPositionSequence r) {
public void add(TextPositionSequence textPositionSequence) {
sequences.add(r);
changed = true;
sequences.add(textPositionSequence);
calculateFrequencyCounters();
calculateBBox();
}
public void addAll(List<TextPositionSequence> textPositionSequences) {
changed = true;
sequences.addAll(textPositionSequences);
calculateFrequencyCounters();
calculateBBox();
}
@ -198,22 +219,28 @@ public class TextPageBlock extends AbstractPageBlock {
@JsonIgnore
public String getText() {
StringBuilder sb = new StringBuilder();
if (text == null || changed) {
TextPositionSequence previous = null;
for (TextPositionSequence word : sequences) {
if (previous != null) {
if (Math.abs(previous.getMaxYDirAdj() - word.getMaxYDirAdj()) > word.getTextHeight()) {
sb.append('\n');
} else {
sb.append(' ');
StringBuilder sb = new StringBuilder();
TextPositionSequence previous = null;
for (TextPositionSequence word : sequences) {
if (previous != null) {
if (Math.abs(previous.getMaxYDirAdj() - word.getMaxYDirAdj()) > word.getTextHeight()) {
sb.append('\n');
} else {
sb.append(' ');
}
}
sb.append(word.toString());
previous = word;
}
sb.append(word.toString());
previous = word;
text = TextNormalizationUtilities.removeHyphenLinebreaks(sb.toString());
changed = false;
}
return TextNormalizationUtilities.removeHyphenLinebreaks(sb.toString());
return text;
}

View File

@ -5,6 +5,7 @@ import static com.knecon.fforesight.service.layoutparser.processor.model.text.Re
import java.util.ArrayList;
import java.util.List;
import java.util.Locale;
import java.util.Objects;
import java.util.stream.Collectors;
import org.apache.pdfbox.text.TextPosition;
@ -14,7 +15,6 @@ import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.TextB
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
import lombok.EqualsAndHashCode;
import lombok.NoArgsConstructor;
import lombok.extern.slf4j.Slf4j;
@ -23,7 +23,6 @@ import lombok.extern.slf4j.Slf4j;
@Builder
@NoArgsConstructor
@AllArgsConstructor
@EqualsAndHashCode(onlyExplicitlyIncluded = true, callSuper = true) // needs the bbox to be unique
public class TextPositionSequence extends TextBoundingBox implements CharSequence {
public static final String STANDARD = "standard";
@ -31,10 +30,8 @@ public class TextPositionSequence extends TextBoundingBox implements CharSequenc
public static final String BOLD = "bold";
public static final String ITALIC = "italic";
@EqualsAndHashCode.Include
private int page;
@EqualsAndHashCode.Include
@Builder.Default
private List<RedTextPosition> textPositions = new ArrayList<>();
@ -42,6 +39,8 @@ public class TextPositionSequence extends TextBoundingBox implements CharSequenc
private boolean strikethrough;
private boolean underline;
private Integer hashcodeCache;
public TextPositionSequence(List<TextPosition> textPositions, int pageNumber, boolean isParagraphStart) {
@ -50,13 +49,14 @@ public class TextPositionSequence extends TextBoundingBox implements CharSequenc
.collect(Collectors.toList());
this.page = pageNumber;
this.isParagraphStart = isParagraphStart;
calculateBBox();
calculateBBoxAndHashcode();
}
private void calculateBBox() {
private void calculateBBoxAndHashcode() {
setToBBoxOfComponents(getTextPositions());
hashcodeCache = null;
}
@ -64,7 +64,7 @@ public class TextPositionSequence extends TextBoundingBox implements CharSequenc
this.textPositions = textPositions;
this.page = page;
calculateBBox();
calculateBBoxAndHashcode();
}
@ -125,16 +125,17 @@ public class TextPositionSequence extends TextBoundingBox implements CharSequenc
this.textPositions.add(textPosition);
this.page = textPositionSequence.getPage();
calculateBBox();
calculateBBoxAndHashcode();
}
public void add(TextPosition textPosition) {
this.textPositions.add(RedTextPosition.fromTextPosition(textPosition));
calculateBBox();
calculateBBoxAndHashcode();
}
public double getTextHeightNoPadding() {
return textPositions.get(0).getHeightDirAdj();
@ -186,5 +187,55 @@ public class TextPositionSequence extends TextBoundingBox implements CharSequenc
return textPositions.get(0).getWidthOfSpace();
}
public boolean equals(final Object o) {
// auto-generated with lombok
if (o == this) {
return true;
}
if (!(o instanceof TextPositionSequence other)) {
return false;
}
if (!other.canEqual((Object) this)) {
return false;
}
if (!super.equals(o)) {
return false;
}
if (this.getPage() != other.getPage()) {
return false;
}
final Object this$textPositions = this.getTextPositions();
final Object other$textPositions = other.getTextPositions();
if (!Objects.equals(this$textPositions, other$textPositions)) {
return false;
}
return Objects.equals(this.getHashcodeCache(), other.getHashcodeCache());
}
protected boolean canEqual(final Object other) {return other instanceof TextPositionSequence;}
public int hashCode() {
if (hashcodeCache == null) {
hashcodeCache = hashcodeCalculation();
}
return hashcodeCache;
}
private int hashcodeCalculation() {
final int PRIME = 59;
int result = super.hashCode();
result = result * PRIME + this.getPage();
final Object $textPositions = this.getTextPositions();
result = result * PRIME + ($textPositions == null ? 43 : $textPositions.hashCode());
return result;
}
}

View File

@ -357,7 +357,7 @@ public class BlockificationPostprocessingService {
if (firstBlock != null && !firstBlock.getSequences().isEmpty()) {
if (textPageBlock.getDir() == firstBlock.getDir()) {
firstBlock.getSequences().addAll(textPageBlock.getSequences());
firstBlock.addAll(textPageBlock.getSequences());
mergedBlocks.add(textPageBlock);
}
}

View File

@ -182,7 +182,7 @@ public class DocstrumBlockificationService {
private TextPageBlock combineBlocksAndResetIterator(TextPageBlock previous, TextPageBlock current, ListIterator<AbstractPageBlock> itty, boolean toDuplicate) {
previous.getSequences().addAll(current.getSequences());
previous.addAll(current.getSequences());
previous = buildTextBlock(previous.getSequences(), 0);
previous.setToDuplicate(toDuplicate);
if (current.getClassification() != null && previous.getClassification() == null) {
@ -283,7 +283,7 @@ public class DocstrumBlockificationService {
if (current.getDir() == inner.getDir() && current.intersects(inner, yThreshold, xThreshold)) {
boolean toDuplicate = current.isToDuplicate() || inner.isToDuplicate();
current.getSequences().addAll(inner.getSequences());
current.addAll(inner.getSequences());
current = buildTextBlock(current.getSequences(), 0);
current.setToDuplicate(toDuplicate);

View File

@ -62,7 +62,6 @@ public class DocuMineBlockificationService {
boolean splitByDir = prev != null && !prev.getDir().equals(word.getDir());
boolean splitByOtherFontAndOtherY = prev != null && Math.abs(prev.getMaxYDirAdj() - word.getMaxYDirAdj()) > word.getTextHeight() * 0.2 //
&& (word.getFontStyle().contains("bold") && !prev.getFontStyle().contains("bold")
//
|| prev.getFontStyle().contains("bold") && !word.getFontStyle().contains("bold")
|| Math.abs(prev.getFontSize() - word.getFontSize()) >= 1
|| Math.abs(word.getTextHeight() - prev.getTextHeight()) > 0.8);
@ -170,7 +169,7 @@ public class DocuMineBlockificationService {
.equals(inner.getClassification()))) {
boolean toDuplicate = current.isToDuplicate() || inner.isToDuplicate();
current.getSequences().addAll(inner.getSequences());
current.addAll(inner.getSequences());
current = buildTextBlock(current.getSequences(), 0);
current.setClassification(inner.getClassification());
current.setToDuplicate(toDuplicate);

View File

@ -23,7 +23,7 @@ public class ClarifyndClassificationService {
public void classifyDocument(ClassificationDocument document) {
List<Double> headlineFontSizes = document.getFontSizeCounter().getHigherThanMostPopular();
List<Double> headlineFontSizes = document.getFontSizeCounter().getValuesInReverseOrder();
log.debug("Document FontSize counters are: {}", document.getFontSizeCounter().getCountPerValue());
@ -35,7 +35,10 @@ public class ClarifyndClassificationService {
}
private void classifyPage(HeadlineClassificationService headlineClassificationService, ClassificationPage page, ClassificationDocument document, List<Double> headlineFontSizes) {
private void classifyPage(HeadlineClassificationService headlineClassificationService,
ClassificationPage page,
ClassificationDocument document,
List<Double> headlineFontSizes) {
for (AbstractPageBlock textBlock : page.getTextBlocks()) {
if (textBlock instanceof TextPageBlock) {
@ -45,7 +48,11 @@ public class ClarifyndClassificationService {
}
private void classifyBlock(HeadlineClassificationService headlineClassificationService, TextPageBlock textBlock, ClassificationPage page, ClassificationDocument document, List<Double> headlineFontSizes) {
private void classifyBlock(HeadlineClassificationService headlineClassificationService,
TextPageBlock textBlock,
ClassificationPage page,
ClassificationDocument document,
List<Double> headlineFontSizes) {
var bodyTextFrame = page.getBodyTextFrame();
@ -57,59 +64,58 @@ public class ClarifyndClassificationService {
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()
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()
} 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)) {
} 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()) {
} 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)) {
PageBlockType headlineType = PageBlockType.getHeadlineType(i);
headlineClassificationService.classifyHeadline(textBlock, headlineType);
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()) {
PageBlockType headlineType = PageBlockType.getHeadlineType(headlineFontSizes.size() + 1);
PageBlockType headlineType = HeadlineClassificationService.headlineClassByFontSize(textBlock, headlineFontSizes);
headlineClassificationService.classifyHeadline(textBlock, headlineType);
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 (!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()) {
PageBlockType headlineType = HeadlineClassificationService.headlineClassByFontSize(textBlock, headlineFontSizes);
headlineClassificationService.classifyHeadline(textBlock, headlineType);
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()) {
} 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);

View File

@ -31,7 +31,7 @@ public class DocuMineClassificationService {
public void classifyDocument(ClassificationDocument document) {
List<Double> headlineFontSizes = document.getFontSizeCounter().getHigherThanMostPopular();
List<Double> headlineFontSizes = document.getFontSizeCounter().getValuesInReverseOrder();
log.debug("Document FontSize counters are: {}", document.getFontSizeCounter().getCountPerValue());
@ -118,15 +118,16 @@ public class DocuMineClassificationService {
|| textBlock.toString().startsWith("TABLE"))
&& !textBlock.toString().endsWith(":")
&& atLeast3Matcher.reset().find()) {
PageBlockType headlineType = PageBlockType.getHeadlineType(1);
PageBlockType headlineType = HeadlineClassificationService.headlineClassByFontSize(textBlock, headlineFontSizes);
headlineClassificationService.classifyHeadline(textBlock, headlineType);
document.setHeadlines(true);
} else if (headlineWithIdentifierMatcher.reset().find()
&& PositionUtils.getApproxLineCount(textBlock) < 2.9
&& atLeast3Matcher.reset().find()
&& !headlineWithSlashesMatcher.reset().matches()) {
PageBlockType headlineType = PageBlockType.getHeadlineType(2);
PageBlockType headlineType = HeadlineClassificationService.headlineClassByFontSize(textBlock, headlineFontSizes);
headlineClassificationService.classifyHeadline(textBlock, headlineType);
document.setHeadlines(true);
} else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock)

View File

@ -2,7 +2,10 @@ package com.knecon.fforesight.service.layoutparser.processor.services.classifica
import static com.knecon.fforesight.service.layoutparser.processor.model.PageBlockType.getHeadlineNumber;
import java.util.List;
import com.knecon.fforesight.service.layoutparser.processor.model.PageBlockType;
import com.knecon.fforesight.service.layoutparser.processor.model.SectionIdentifier;
import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
import lombok.Getter;
@ -16,6 +19,7 @@ public class HeadlineClassificationService {
PageBlockType originalClassifiedBlockType;
TextPageBlock lastHeadlineFromOutline;
public void setLastHeadlineFromOutline(TextPageBlock lastHeadlineFromOutline) {
this.lastHeadlineFromOutline = lastHeadlineFromOutline;
@ -25,28 +29,57 @@ public class HeadlineClassificationService {
public void classifyHeadline(TextPageBlock textBlock, PageBlockType initialHeadlineType) {
TextPageBlock lastHeadline = getLastHeadline();
TextPageBlock lastHeadlineFromOutline = getLastHeadlineFromOutline();
PageBlockType originalClassifiedBlockType = getOriginalClassifiedBlockType();
PageBlockType finalHeadlineType = initialHeadlineType;
if (lastHeadline != null) {
if (lastHeadline.equals(lastHeadlineFromOutline)) {
finalHeadlineType = PageBlockType.getHeadlineType(getHeadlineNumber(lastHeadline.getClassification()) + 1);
} else if (originalClassifiedBlockType != null && lastHeadline.getClassification() != originalClassifiedBlockType) {
PageBlockType lastHeadlineType = lastHeadline.getClassification();
int difference = getHeadlineNumber(originalClassifiedBlockType) - getHeadlineNumber(lastHeadlineType);
finalHeadlineType = PageBlockType.getHeadlineType(getHeadlineNumber(initialHeadlineType) - difference);
}
finalHeadlineType = decideOnClassification(textBlock, initialHeadlineType);
}
setOriginalClassifiedBlockType(initialHeadlineType);
lastHeadline = textBlock;
originalClassifiedBlockType = initialHeadlineType;
textBlock.setClassification(finalHeadlineType);
setLastHeadline(textBlock);
}
private PageBlockType decideOnClassification(TextPageBlock textBlock, PageBlockType initialHeadlineType) {
SectionIdentifier identifier = SectionIdentifier.fromSearchText(textBlock.getText());
TextPageBlock lastHeadlineFromOutline = getLastHeadlineFromOutline();
PageBlockType originalClassifiedBlockType = getOriginalClassifiedBlockType();
if (!identifier.isEmpty()) {
return PageBlockType.getHeadlineType(identifier.level());
}
if (lastHeadline.equals(lastHeadlineFromOutline) && lastHeadline.getMostPopularWordFontSize() >= textBlock.getMostPopularWordFontSize()) {
return PageBlockType.getHeadlineType(getHeadlineNumber(lastHeadline.getClassification()) + 1);
} else if (originalClassifiedBlockType != null && lastHeadline.getClassification() != originalClassifiedBlockType) {
return adjustInitialLevelToLastHeadlineLevel(initialHeadlineType);
}
return initialHeadlineType;
}
private PageBlockType adjustInitialLevelToLastHeadlineLevel(PageBlockType initialHeadlineType) {
int difference = getHeadlineNumber(originalClassifiedBlockType) - getHeadlineNumber(lastHeadline.getClassification());
return PageBlockType.getHeadlineType(Math.max(1, getHeadlineNumber(initialHeadlineType) - difference));
}
public static PageBlockType headlineClassByFontSize(TextPageBlock textBlock, List<Double> fontSizeGroups) {
PageBlockType headlineType = PageBlockType.H1;
for (int i = 1; i <= fontSizeGroups.size(); i++) {
if (textBlock.getMostPopularWordFontSize() == fontSizeGroups.get(i - 1)) {
headlineType = PageBlockType.getHeadlineType(i);
}
}
return headlineType;
}
}

View File

@ -22,10 +22,9 @@ import lombok.extern.slf4j.Slf4j;
@RequiredArgsConstructor
public class RedactManagerClassificationService {
public void classifyDocument(ClassificationDocument document) {
List<Double> headlineFontSizes = document.getFontSizeCounter().getHigherThanMostPopular();
List<Double> headlineFontSizes = document.getFontSizeCounter().getValuesInReverseOrder();
log.debug("Document FontSize counters are: {}", document.getFontSizeCounter().getCountPerValue());
@ -37,7 +36,10 @@ public class RedactManagerClassificationService {
}
private void classifyPage(HeadlineClassificationService headlineClassificationService, ClassificationPage page, ClassificationDocument document, List<Double> headlineFontSizes) {
private void classifyPage(HeadlineClassificationService headlineClassificationService,
ClassificationPage page,
ClassificationDocument document,
List<Double> headlineFontSizes) {
for (AbstractPageBlock textBlock : page.getTextBlocks()) {
if (textBlock instanceof TextPageBlock) {
@ -47,7 +49,11 @@ public class RedactManagerClassificationService {
}
private void classifyBlock(HeadlineClassificationService headlineClassificationService, TextPageBlock textBlock, ClassificationPage page, ClassificationDocument document, List<Double> headlineFontSizes) {
private void classifyBlock(HeadlineClassificationService headlineClassificationService,
TextPageBlock textBlock,
ClassificationPage page,
ClassificationDocument document,
List<Double> headlineFontSizes) {
var bodyTextFrame = page.getBodyTextFrame();
@ -71,15 +77,18 @@ public class RedactManagerClassificationService {
|| 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())) {
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)) {
if (!Pattern.matches("[0-9]+", textBlock.toString())) {
textBlock.setClassification(PageBlockType.TITLE);
}
@ -88,45 +97,42 @@ public class RedactManagerClassificationService {
&& (textBlock.getMostPopularWordStyle().equals("bold")
|| !document.getFontStyleCounter().getCountPerValue().containsKey("bold")
&& textBlock.getMostPopularWordFontSize() > document.getFontSizeCounter().getMostPopular() + 1)
&& textBlock.getSequences()
.get(0).getTextPositions()
.get(0).getFontSizeInPt() >= textBlock.getMostPopularWordFontSize()) {
&& textBlock.getSequences().get(0).getTextPositions().get(0).getFontSizeInPt() >= textBlock.getMostPopularWordFontSize()) {
for (int i = 1; i <= headlineFontSizes.size(); i++) {
if (textBlock.getMostPopularWordFontSize() == headlineFontSizes.get(i - 1)) {
PageBlockType headlineType = PageBlockType.getHeadlineType(i);
headlineClassificationService.classifyHeadline(textBlock, headlineType);
document.setHeadlines(true);
}
}
PageBlockType headlineType = HeadlineClassificationService.headlineClassByFontSize(textBlock, headlineFontSizes);
headlineClassificationService.classifyHeadline(textBlock, headlineType);
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()) {
PageBlockType headlineType = PageBlockType.getHeadlineType(headlineFontSizes.size() + 1);
&& textBlock.getSequences().get(0).getTextPositions().get(0).getFontSizeInPt() >= textBlock.getMostPopularWordFontSize()) {
PageBlockType headlineType = HeadlineClassificationService.headlineClassByFontSize(textBlock, headlineFontSizes);
headlineClassificationService.classifyHeadline(textBlock, headlineType);
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);

View File

@ -95,8 +95,8 @@ public class DocumentGraphFactory {
private void addSections(LayoutParsingType layoutParsingType, ClassificationDocument classificationDocument, Context context, Document document) {
for (TableOfContentItem tocItem : classificationDocument.getTableOfContents()) {
var parent = tocItem.getParent() == null ? null : tocItem.getParent().getSection();
Optional<AbstractSemanticNode> section = SectionNodeFactory.addSection(layoutParsingType,
GenericSemanticNode parent = tocItem.getParent() == null ? null : tocItem.getParent().getSection();
Optional<GenericSemanticNode> section = SectionNodeFactory.addSection(layoutParsingType,
parent,
tocItem.getChildren().isEmpty(),
tocItem.getNonEmptySectionBlocks(),
@ -129,10 +129,10 @@ public class DocumentGraphFactory {
textBlocks.add(originalTextBlock);
textBlocks.addAll(textBlocksToMerge);
AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock2(TextPositionOperations.mergeAndSort(textBlocks), node, context, page);
AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock(TextPositionOperations.mergeAndSort(textBlocks), node, context, page);
if (node instanceof DuplicatedParagraph duplicatedParagraph) {
AtomicTextBlock unsortedTextBlock = context.textBlockFactory.buildAtomicTextBlock2(textBlocks.stream()
AtomicTextBlock unsortedTextBlock = context.textBlockFactory.buildAtomicTextBlock(textBlocks.stream()
.flatMap(tb -> tb.getSequences()
.stream())
.collect(Collectors.toList()), node, context, page);
@ -207,7 +207,7 @@ public class DocumentGraphFactory {
Page page = context.getPage(textBlocks.get(0).getPage());
Footer footer = Footer.builder().documentTree(context.getDocumentTree()).build();
AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock2(TextPositionOperations.merge(textBlocks), footer, context, page);
AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock(TextPositionOperations.merge(textBlocks), footer, context, page);
List<Integer> tocId = context.getDocumentTree().createNewMainEntryAndReturnId(footer);
footer.setTreeId(tocId);
footer.setLeafTextBlock(textBlock);

View File

@ -2,13 +2,11 @@ package com.knecon.fforesight.service.layoutparser.processor.services.factory;
import static java.lang.String.format;
import static java.util.Collections.emptyList;
import static java.util.stream.Collectors.groupingBy;
import java.util.ArrayList;
import java.util.HashSet;
import java.util.LinkedList;
import java.util.List;
import java.util.Map;
import java.util.Optional;
import java.util.Set;
@ -17,7 +15,6 @@ import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBl
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.AbstractSemanticNode;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.GenericSemanticNode;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Page;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Section;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.SuperSection;
import com.knecon.fforesight.service.layoutparser.processor.model.image.ClassifiedImage;
@ -30,13 +27,13 @@ import lombok.experimental.UtilityClass;
@UtilityClass
public class SectionNodeFactory {
public Optional<AbstractSemanticNode> addSection(LayoutParsingType layoutParsingType,
GenericSemanticNode parentNode,
boolean isLeaf,
List<AbstractPageBlock> pageBlocks,
List<ClassifiedImage> images,
DocumentGraphFactory.Context context,
Document document) {
public Optional<GenericSemanticNode> addSection(LayoutParsingType layoutParsingType,
GenericSemanticNode parentNode,
boolean isLeaf,
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.
@ -52,8 +49,7 @@ public class SectionNodeFactory {
}
AbstractSemanticNode section;
boolean containsTablesAndTextBlocks = containsTablesAndTextBlocks(pageBlocks);
if (isLeaf && !containsTablesAndTextBlocks) {
if (isLeaf) {
section = Section.builder().documentTree(context.getDocumentTree()).build();
} else {
section = SuperSection.builder().documentTree(context.getDocumentTree()).build();
@ -64,6 +60,7 @@ public class SectionNodeFactory {
section.setTreeId(getTreeId(parentNode, context, section));
addFirstHeadlineDirectlyToSection(layoutParsingType, pageBlocks, context, section, document);
boolean containsTablesAndTextBlocks = containsTablesAndTextBlocks(pageBlocks);
if (containsTablesAndTextBlocks) {
splitPageBlocksIntoSubSections(pageBlocks).forEach(subSectionPageBlocks -> addSection(layoutParsingType,
section,
@ -153,7 +150,8 @@ public class SectionNodeFactory {
private boolean containsTablesAndTextBlocks(List<AbstractPageBlock> pageBlocks) {
return pageBlocks.stream()
.anyMatch(pageBlock -> pageBlock instanceof TablePageBlock) && pageBlocks.stream()
.anyMatch(pageBlock -> pageBlock instanceof TablePageBlock) //
&& pageBlocks.stream()
.anyMatch(pageBlock -> pageBlock instanceof TextPageBlock);
}
@ -236,6 +234,4 @@ public class SectionNodeFactory {
.toList();
}
}

View File

@ -4,7 +4,6 @@ import static java.util.Collections.emptyList;
import java.util.Collection;
import java.util.List;
import java.util.Set;
import java.util.stream.Collectors;
import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingType;
@ -12,7 +11,6 @@ import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBl
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.GenericSemanticNode;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Page;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.SemanticNode;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Table;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.TableCell;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
@ -117,7 +115,7 @@ public class TableNodeFactory {
if (cell.getTextBlocks().isEmpty()) {
tableCell.setLeafTextBlock(context.getTextBlockFactory().emptyTextBlock(tableNode, context, page));
} else if (cell.getTextBlocks().size() == 1) {
textBlock = context.getTextBlockFactory().buildAtomicTextBlock2(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(layoutParsingType,
@ -132,7 +130,7 @@ public class TableNodeFactory {
document);
} else if (cellAreaIsSmallerThanPageAreaTimesThreshold(cell, page)) {
List<TextPositionSequence> sequences = TextPositionOperations.mergeAndSort(cell.getTextBlocks());
textBlock = context.getTextBlockFactory().buildAtomicTextBlock2(sequences, tableCell, context, page);
textBlock = context.getTextBlockFactory().buildAtomicTextBlock(sequences, tableCell, context, page);
tableCell.setLeafTextBlock(textBlock);
} else {
cell.getTextBlocks()

View File

@ -17,7 +17,7 @@ public class TextBlockFactory {
long textBlockIdx;
public AtomicTextBlock buildAtomicTextBlock2(List<TextPositionSequence> sequences, SemanticNode parent, DocumentGraphFactory.Context context, Page page) {
public AtomicTextBlock buildAtomicTextBlock(List<TextPositionSequence> sequences, SemanticNode parent, DocumentGraphFactory.Context context, Page page) {
Integer numberOnPage = context.getAndIncrementTextBlockNumberOnPage(page);
return buildAtomicTextBlock(sequences, parent, numberOnPage, page);

View File

@ -1,4 +1,4 @@
package com.knecon.fforesight.service.layoutparser.processor.markdown;
package com.knecon.fforesight.service.layoutparser.processor.services.mapper;
import java.util.ArrayList;
import java.util.HashSet;

View File

@ -0,0 +1,84 @@
package com.knecon.fforesight.service.layoutparser.processor.services.mapper;
import java.awt.geom.AffineTransform;
import java.awt.geom.Point2D;
import java.awt.geom.Rectangle2D;
import java.util.LinkedList;
import java.util.Map;
import java.util.Optional;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Headline;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Page;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Section;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.SemanticNode;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.SuperSection;
import com.knecon.fforesight.service.layoutparser.processor.utils.CoordinateTransforms;
import com.knecon.fforesight.service.layoutparser.processor.utils.PageInformation;
import com.knecon.fforesight.service.viewerdoc.model.Outline;
import lombok.SneakyThrows;
import lombok.experimental.UtilityClass;
@UtilityClass
public class OutlineMapper {
public Outline createOutline(Document document) {
Outline outline = new Outline();
addChildren(document, null, outline);
return outline;
}
public void addChildren(SemanticNode parentNode, Outline.Entry parentEntry, Outline outline) {
parentNode.streamChildren()
.filter(child -> child instanceof Section || child instanceof SuperSection)
.forEach(child -> {
Optional<Headline> headline = findHeadline(child);
if (headline.isPresent()) {
Outline.Entry entry = buildEntry(child.getHeadline());
if (parentEntry != null) {
parentEntry.children().add(entry);
} else {
outline.getEntries().add(entry);
}
addChildren(child, entry, outline);
} else {
addChildren(child, parentEntry, outline);
}
});
}
private static Optional<Headline> findHeadline(SemanticNode child) {
return child.streamChildren()
.filter(node -> node instanceof Headline)
.map(node -> (Headline) node)
.findFirst();
}
@SneakyThrows
private Outline.Entry buildEntry(Headline headline) {
Map<Page, Rectangle2D> bbox = headline.getBBox();
Rectangle2D r = bbox.get(headline.getFirstPage());
Point2D.Double position = new Point2D.Double(r.getMinX(), r.getMaxY());
PageInformation pageInformation = PageInformation.fromPage(headline.getFirstPage());
AffineTransform pdfToPage = CoordinateTransforms.calculateInitialUserSpaceCoordsToPageCoords(pageInformation);
pdfToPage.transform(position, position);
AffineTransform mirror = new AffineTransform(1, 0, 0, -1, 0, pageInformation.heightRot());
mirror.transform(position, position);
AffineTransform.getTranslateInstance(0, 5).transform(position, position);
Outline.JumpAction action = new Outline.JumpAction(headline.getFirstPage().getNumber(), position);
return new Outline.Entry(headline.getTextBlock().getSearchText(), action, new LinkedList<>());
}
}

View File

@ -10,7 +10,9 @@ import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.He
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Image;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Paragraph;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Table;
import com.knecon.fforesight.service.layoutparser.processor.services.mapper.OutlineMapper;
import com.knecon.fforesight.service.layoutparser.processor.visualization.LayoutGrid;
import com.knecon.fforesight.service.viewerdoc.model.Outline;
import com.knecon.fforesight.service.viewerdoc.service.PDFTronViewerDocumentService;
import io.micrometer.observation.annotation.Observed;
@ -29,16 +31,15 @@ public class LayoutGridService {
@SneakyThrows
@Observed(name = "ViewerDocumentService", contextualName = "create-viewer-document")
public void addLayoutGrid(File originFile, Document document, File destinationFile, boolean layerVisibilityDefaultValue, boolean writeVisualLayoutParsingGrid) {
public void addLayoutGrid(File originFile, Document document, File destinationFile, boolean layerVisibilityDefaultValue) {
LayoutGrid layoutGrid = createLayoutGrid(document);
Outline outline = OutlineMapper.createOutline(document);
layoutGrid.setVisibleByDefault(layerVisibilityDefaultValue);
// Visualizations visualLayoutGrid = this.addLayoutGrid(document, layerVisibilityDefaultValue, true);
if (document.getLayoutDebugLayer().isActive()) {
viewerDocumentService.addLayerGroups(originFile, destinationFile, List.of(layoutGrid, document.getLayoutDebugLayer()));
viewerDocumentService.addLayerGroups(originFile, destinationFile, List.of(layoutGrid, document.getLayoutDebugLayer()), outline);
} else {
viewerDocumentService.addLayerGroups(originFile, destinationFile, List.of(layoutGrid));
viewerDocumentService.addLayerGroups(originFile, destinationFile, List.of(layoutGrid), outline);
}
}

View File

@ -5,14 +5,22 @@ import java.awt.geom.Rectangle2D;
import org.apache.pdfbox.pdmodel.PDPage;
import org.apache.pdfbox.pdmodel.common.PDRectangle;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Page;
public record PageInformation(Rectangle2D mediabox, int number, int rotationDegrees) {
public static PageInformation fromPDPage(int pageNum, PDPage page) {
PDRectangle mediaBox = page.getMediaBox();
return new PageInformation(new Rectangle2D.Double(mediaBox.getLowerLeftX(), mediaBox.getLowerLeftY(), mediaBox.getWidth(), mediaBox.getHeight()),
pageNum,
page.getRotation());
pageNum,
page.getRotation());
}
public static PageInformation fromPage(Page page) {
return new PageInformation(new Rectangle2D.Double(0, 0, page.getWidth(), page.getHeight()), page.getNumber(), page.getRotation());
}

View File

@ -1,6 +1,5 @@
package com.knecon.fforesight.service.layoutparser.processor.utils;
import java.util.regex.Matcher;
import java.util.regex.Pattern;
import lombok.experimental.UtilityClass;

View File

@ -14,6 +14,8 @@ import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.TextB
import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.UnionFind;
import com.knecon.fforesight.service.layoutparser.processor.docstrum.utils.DoubleUtils;
import com.knecon.fforesight.service.layoutparser.processor.docstrum.utils.FastAtan2;
import com.knecon.fforesight.service.layoutparser.processor.model.text.RedTextPosition;
import com.knecon.fforesight.service.layoutparser.processor.model.text.TextDirection;
import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
@ -81,12 +83,10 @@ public class TextPositionOperations {
double maxLineDistance = sequences.stream()
.map(TextPositionSequence::getBBoxDirAdj)
.mapToDouble(RectangularShape::getHeight).average()
.orElse(10) * MAX_LINE_HEIGHT_FACTOR;
.mapToDouble(RectangularShape::getHeight).average().orElse(10) * MAX_LINE_HEIGHT_FACTOR;
double maxXGap = sequences.stream()
.map(TextPositionSequence::getBBoxDirAdj)
.mapToDouble(RectangularShape::getWidth).average()
.orElse(75) * MAX_WORD_DISTANCE_FACTOR;
.mapToDouble(RectangularShape::getWidth).average().orElse(75) * MAX_WORD_DISTANCE_FACTOR;
UnionFind<TextPositionSequence> unionFind = new UnionFind<>(sequences);

View File

@ -133,9 +133,6 @@ public class LayoutGrid extends LayoutGridLayerConfig {
.collect(Collectors.toList());
pagesInOrder.remove(0);
handleFirstPageOfSection(section, firstPage, bBoxMap.get(firstPage), treeIdString, maxChildDepth, ownDepth);
if (section instanceof SuperSection) {
return;
}
for (Page middlePage : pagesInOrder.subList(0, pagesInOrder.size() - 1)) {
handleForMiddlePageOfSection(section, middlePage, bBoxMap.get(middlePage), treeIdString, maxChildDepth, ownDepth);
}
@ -203,9 +200,9 @@ public class LayoutGrid extends LayoutGridLayerConfig {
List<PlacedText> placedTexts = getOrCreateVisualizationsOnPage(page.getNumber(), visualizations).getPlacedTexts();
PlacedText newText = PlacedText.textFacingUp(s, upperLeftCorner, FONT_SIZE, color, FONT);
float threshold = 1.5f * FONT_SIZE;
Optional<PlacedText> conflictingText = placedTexts.stream()
.filter(pt -> Math.abs(pt.lineStart().getY() - newText.lineStart().getY()) <= FONT_SIZE)
.filter(pt -> Math.abs(pt.lineStart().getY() - newText.lineStart().getY()) <= threshold || Math.abs(pt.lineStart().getX() - newText.lineStart().getX()) <= threshold)
.findFirst();
if (conflictingText.isPresent()) {

View File

@ -14,4 +14,6 @@
<appender-ref ref="${logType}"/>
</root>
<logger name="org.apache.fontbox.ttf" level="ERROR"/>
</configuration>

View File

@ -0,0 +1,86 @@
package com.knecon.fforesight.service.layoutparser.processor.model;
import static org.junit.jupiter.api.Assertions.*;
import java.util.List;
import org.junit.jupiter.api.Test;
class SectionIdentifierTest {
@Test
void testSectionIdentifier() {
SectionIdentifier identifier = SectionIdentifier.fromSearchText("1.1.2: Headline");
assertEquals(SectionIdentifier.Format.NUMERICAL, identifier.getFormat());
assertEquals(3, identifier.level());
assertEquals(List.of(1, 1, 2), identifier.getIdentifiers());
SectionIdentifier child = SectionIdentifier.asChildOf(identifier);
assertTrue(child.isChildOf(identifier));
SectionIdentifier parent = SectionIdentifier.fromSearchText("1.1: Headline");
assertTrue(parent.isParentOf(identifier));
}
@Test
void testSectionIdentifier2() {
SectionIdentifier identifier = SectionIdentifier.fromSearchText("A.1.2: Headline");
assertEquals(SectionIdentifier.Format.ALPHANUMERIC, identifier.getFormat());
assertEquals(3, identifier.level());
assertEquals(List.of(1, 1, 2), identifier.getIdentifiers());
}
@Test
void testSectionIdentifier3() {
SectionIdentifier identifier = SectionIdentifier.fromSearchText("D.1.2: Headline");
assertEquals(SectionIdentifier.Format.ALPHANUMERIC, identifier.getFormat());
assertEquals(3, identifier.level());
assertEquals(List.of(4, 1, 2), identifier.getIdentifiers());
}
@Test
void testSectionIdentifier4() {
SectionIdentifier identifier = SectionIdentifier.fromSearchText("4.1.2.4: Headline");
assertEquals(SectionIdentifier.Format.NUMERICAL, identifier.getFormat());
assertEquals(4, identifier.level());
assertEquals(List.of(4, 1, 2, 4), identifier.getIdentifiers());
}
@Test
void testSectionIdentifier5() {
SectionIdentifier identifier = SectionIdentifier.fromSearchText("D.1.2.4.5: Headline");
assertEquals(SectionIdentifier.Format.ALPHANUMERIC, identifier.getFormat());
assertEquals(4, identifier.level());
assertEquals(List.of(4, 1, 2, 4), identifier.getIdentifiers());
}
@Test
void testSectionIdentifier6() {
SectionIdentifier identifier = SectionIdentifier.fromSearchText("d.1.2.4.5: Headline");
assertEquals(SectionIdentifier.Format.ALPHANUMERIC, identifier.getFormat());
assertEquals(4, identifier.level());
assertEquals(List.of(4, 1, 2, 4), identifier.getIdentifiers());
}
@Test
void testSectionIdentifier7() {
SectionIdentifier identifier = SectionIdentifier.fromSearchText("4.1.2.4.5: Headline");
assertEquals(SectionIdentifier.Format.NUMERICAL, identifier.getFormat());
assertEquals(4, identifier.level());
assertEquals(List.of(4, 1, 2, 4), identifier.getIdentifiers());
}
}

View File

@ -37,7 +37,7 @@ public class LayoutparserEnd2EndTest extends AbstractTest {
@Disabled
public void testLayoutParserEndToEnd() {
String filePath = "/home/kschuettler/Dokumente/Ticket Related/RED-9964/17a25133-e098-4610-b553-d1bf11a56d96/560e6ab1ab4754b9a62fd2e6d4d71327/560e6ab1ab4754b9a62fd2e6d4d71327.ORIGIN.pdf";
String filePath = "/home/kschuettler/Dokumente/TestFiles/NER Dataset/Syngenta prod/77c680315c31d403d2e023be023b2087.PREVIEW.pdf";
runForFile(filePath);
}
@ -48,7 +48,7 @@ public class LayoutparserEnd2EndTest extends AbstractTest {
@SneakyThrows
public void testLayoutParserEndToEndWithFolder() {
String folder = "/home/kschuettler/Dokumente/TestFiles/ReadingOrder";
String folder = "/home/kschuettler/Dokumente/Ticket Related/RED-9975";
List<Path> pdfFiles = Files.walk(Path.of(folder))
.filter(path -> path.getFileName().toString().endsWith(".pdf"))
.sorted(Comparator.comparing(Path::getFileName))

View File

@ -36,7 +36,6 @@ import com.knecon.fforesight.service.layoutparser.processor.services.visualizati
import com.knecon.fforesight.service.layoutparser.server.utils.AbstractTest;
import com.knecon.fforesight.service.viewerdoc.service.PDFTronViewerDocumentService;
import jakarta.annotation.PostConstruct;
import lombok.SneakyThrows;
public class OutlineDetectionTest extends AbstractTest {
@ -81,7 +80,8 @@ public class OutlineDetectionTest extends AbstractTest {
long start = System.currentTimeMillis();
ClassificationDocument classificationDocument = parseLayout(fileName, LayoutParsingType.REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH);
Document document = buildGraph(fileName, classificationDocument);
layoutGridService.addLayoutGrid(documentFile, document, new File(tmpFileName), true);
OutlineObjectTree outlineObjectTree = classificationDocument.getOutlineObjectTree();
assertEquals(outlineObjectTree.getRootNodes().size(), 8);
assertEquals(outlineObjectTree.getOutlineObjectsPerPage().get(2).size(), 1);
@ -102,7 +102,7 @@ public class OutlineDetectionTest extends AbstractTest {
TableOfContents tableOfContents = classificationDocument.getTableOfContents();
assertEquals(tableOfContents.getMainSections().size(), 9);
assertEquals(tableOfContents.getMainSections().size(), 10);
assertEquals(tableOfContents.getMainSections().subList(1, 9)
.stream()
.map(tableOfContentItem -> sanitizeString(tableOfContentItem.getHeadline().toString()))
@ -111,17 +111,15 @@ public class OutlineDetectionTest extends AbstractTest {
.stream()
.map(outlineObjectTreeNode -> sanitizeString(outlineObjectTreeNode.getOutlineObject().getTitle()))
.toList());
assertEquals(tableOfContents.getMainSections().get(5).getChildren().size(), 6);
assertEquals(tableOfContents.getMainSections().get(7).getChildren().size(), 3);
assertEquals(tableOfContents.getMainSections().get(8).getChildren().size(), 3);
assertEquals(tableOfContents.getMainSections().get(8).getChildren().get(2).getChildren().size(), 1);
assertEquals(tableOfContents.getMainSections().get(8).getChildren().get(2).getChildren().get(0).getChildren().size(), 3);
assertEquals(tableOfContents.getMainSections().get(0).getImages().size(), 1);
assertEquals(tableOfContents.getMainSections().get(6).getImages().size(), 1);
assertEquals(tableOfContents.getMainSections().get(8).getChildren().get(2).getChildren().get(0).getChildren().get(2).getImages().size(), 1);
Document document = buildGraph(fileName, classificationDocument);
// assertEquals(tableOfContents.getMainSections().get(5).getChildren().size(), 6);
// assertEquals(tableOfContents.getMainSections().get(7).getChildren().size(), 3);
// assertEquals(tableOfContents.getMainSections().get(8).getChildren().size(), 3);
// assertEquals(tableOfContents.getMainSections().get(8).getChildren().get(2).getChildren().size(), 1);
// assertEquals(tableOfContents.getMainSections().get(8).getChildren().get(2).getChildren().get(0).getChildren().size(), 3);
//
// assertEquals(tableOfContents.getMainSections().get(0).getImages().size(), 1);
// assertEquals(tableOfContents.getMainSections().get(6).getImages().size(), 1);
// assertEquals(tableOfContents.getMainSections().get(8).getChildren().get(2).getChildren().get(0).getChildren().get(2).getImages().size(), 1);
assertTrue(tableOfContents.getAllTableOfContentItems()
.stream()
@ -137,7 +135,7 @@ public class OutlineDetectionTest extends AbstractTest {
List<SemanticNode> childrenOfTypeSectionOrSuperSection = document.getChildrenOfTypeSectionOrSuperSection();
assertEquals(childrenOfTypeSectionOrSuperSection.size(), 9);
assertEquals(childrenOfTypeSectionOrSuperSection.size(), 10);
assertEquals(childrenOfTypeSectionOrSuperSection.subList(1, 9)
.stream()
.map(section -> sanitizeString(section.getHeadline().getLeafTextBlock().toString()))
@ -146,38 +144,37 @@ public class OutlineDetectionTest extends AbstractTest {
.stream()
.map(outlineObjectTreeNode -> sanitizeString(outlineObjectTreeNode.getOutlineObject().getTitle()))
.toList());
Predicate<SemanticNode> isSectionOrSuperSection = semanticNode -> semanticNode instanceof Section || semanticNode instanceof SuperSection;
assertEquals(childrenOfTypeSectionOrSuperSection.get(5).streamChildren()
.filter(isSectionOrSuperSection)
.count(), 6 + 1); // 1 additional for main text of parent section
assertEquals(childrenOfTypeSectionOrSuperSection.get(7).streamChildren()
.filter(isSectionOrSuperSection)
.count(), 3 + 1);
assertEquals(childrenOfTypeSectionOrSuperSection.get(8).streamChildren()
.filter(isSectionOrSuperSection)
.count(), 3 + 1);
assertEquals(childrenOfTypeSectionOrSuperSection.get(8).streamChildren()
.filter(isSectionOrSuperSection)
.toList().get(3).streamChildren()
.filter(isSectionOrSuperSection)
.count(), 1 + 1);
assertEquals(childrenOfTypeSectionOrSuperSection.get(8).streamChildren()
.filter(isSectionOrSuperSection)
.toList().get(3).streamChildren()
.filter(isSectionOrSuperSection)
.toList().get(1).streamChildren()
.filter(isSectionOrSuperSection)
.count(), 3 + 1);
// Predicate<SemanticNode> isSectionOrSuperSection = semanticNode -> semanticNode instanceof Section || semanticNode instanceof SuperSection;
// assertEquals(childrenOfTypeSectionOrSuperSection.get(5).streamChildren()
// .filter(isSectionOrSuperSection)
// .count(), 6 + 1); // 1 additional for main text of parent section
// assertEquals(childrenOfTypeSectionOrSuperSection.get(7).streamChildren()
// .filter(isSectionOrSuperSection)
// .count(), 3 + 1);
// assertEquals(childrenOfTypeSectionOrSuperSection.get(8).streamChildren()
// .filter(isSectionOrSuperSection)
// .count(), 3 + 1);
// assertEquals(childrenOfTypeSectionOrSuperSection.get(8).streamChildren()
// .filter(isSectionOrSuperSection)
// .toList().get(3).streamChildren()
// .filter(isSectionOrSuperSection)
// .count(), 1 + 1);
// assertEquals(childrenOfTypeSectionOrSuperSection.get(8).streamChildren()
// .filter(isSectionOrSuperSection)
// .toList().get(3).streamChildren()
// .filter(isSectionOrSuperSection)
// .toList().get(1).streamChildren()
// .filter(isSectionOrSuperSection)
// .count(), 3 + 1);
List<List<Integer>> imageTreeIdList = document.streamAllImages()
.map(image -> image.getParent().getTreeId())
.toList();
// List<List<Integer>> imageTreeIdList = document.streamAllImages()
// .map(image -> image.getParent().getTreeId())
// .toList();
//
// assertEquals(imageTreeIdList.get(0), List.of(0));
// assertEquals(imageTreeIdList.get(1), List.of(6));
// assertEquals(imageTreeIdList.get(2), List.of(8, 4, 2, 4));
assertEquals(imageTreeIdList.get(0), List.of(0));
assertEquals(imageTreeIdList.get(1), List.of(6));
assertEquals(imageTreeIdList.get(2), List.of(8, 4, 2, 4));
layoutGridService.addLayoutGrid(documentFile, document, new File(tmpFileName), true, false);
System.out.printf("Total time: %.2fs%n", ((float) (System.currentTimeMillis() - start)) / 1000);
}

View File

@ -13,6 +13,7 @@ import java.util.List;
import org.apache.commons.text.similarity.LevenshteinDistance;
import org.junit.jupiter.api.AfterEach;
import org.junit.jupiter.api.BeforeEach;
import org.junit.jupiter.api.Disabled;
import org.junit.jupiter.api.Test;
import org.springframework.beans.factory.annotation.Autowired;
@ -50,7 +51,7 @@ import lombok.extern.slf4j.Slf4j;
@Slf4j
public class DocumentReadingOrderTest extends BuildDocumentTest {
private static final boolean DRAW_DIR_ADJ_COORDS = false;
private static final boolean DRAW_DIR_ADJ_COORDS = true;
public static final List<LayoutParsingType> LAYOUT_PARSING_TYPES = List.of(LayoutParsingType.DOCUMINE,
LayoutParsingType.DOCUMINE_OLD,
LayoutParsingType.REDACT_MANAGER,
@ -77,6 +78,18 @@ public class DocumentReadingOrderTest extends BuildDocumentTest {
}
@Test
@Disabled
public void drawDirAdjForFile() {
String pdfFile = "/home/kschuettler/Dokumente/Ticket Related/RED-9974/026dc94b019bc2348a4c54f0c6c4516f.ORIGIN.pdf";
ClassificationDocument classificationDocument = parseLayout(pdfFile, LayoutParsingType.DOCUMINE_OLD);
drawDirAdjCoords(pdfFile, classificationDocument, LayoutParsingType.DOCUMINE_OLD);
}
@Test
public void readingOrderTestSeite14() {

View File

@ -4,18 +4,13 @@ import java.io.File;
import java.nio.file.Path;
import java.util.Map;
import org.junit.jupiter.api.AfterAll;
import org.junit.jupiter.api.BeforeEach;
import org.junit.jupiter.api.Disabled;
import org.junit.jupiter.api.Test;
import org.mockito.MockitoAnnotations;
import org.springframework.amqp.rabbit.core.RabbitTemplate;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.boot.test.mock.mockito.MockBean;
import org.springframework.core.io.ClassPathResource;
import com.iqser.red.commons.jackson.ObjectMapperFactory;
import com.iqser.red.storage.commons.service.StorageService;
import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingType;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
import com.knecon.fforesight.service.layoutparser.processor.python_api.model.image.ImageServiceResponse;
@ -26,10 +21,7 @@ import com.knecon.fforesight.service.layoutparser.processor.services.visualizati
import com.knecon.fforesight.service.layoutparser.server.PDFNetInitializer;
import com.knecon.fforesight.service.layoutparser.server.utils.BuildDocumentTest;
import com.knecon.fforesight.service.viewerdoc.service.PDFTronViewerDocumentService;
import com.knecon.fforesight.tenantcommons.TenantsClient;
import com.pdftron.pdf.PDFNet;
import jakarta.annotation.PostConstruct;
import lombok.SneakyThrows;
public class ViewerDocumentTest extends BuildDocumentTest {
@ -59,7 +51,7 @@ public class ViewerDocumentTest extends BuildDocumentTest {
long start = System.currentTimeMillis();
Document document = buildGraph(fileName, LayoutParsingType.REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH);
layoutGridService.addLayoutGrid(documentFile, document, new File(tmpFileName), true, false);
layoutGridService.addLayoutGrid(documentFile, document, new File(tmpFileName), true);
System.out.printf("Total time: %.2fs%n", ((float) (System.currentTimeMillis() - start)) / 1000);
}
@ -87,7 +79,7 @@ public class ViewerDocumentTest extends BuildDocumentTest {
LayoutGridService layoutGridService = new LayoutGridService(viewerDocumentService);
Document document = DocumentGraphFactory.buildDocumentGraph(LayoutParsingType.DOCUMINE_OLD, classificationDocument);
layoutGridService.addLayoutGrid(documentFile, document, new File(tmpFileName), true, false);
layoutGridService.addLayoutGrid(documentFile, document, new File(tmpFileName), true);
}
}

View File

@ -141,16 +141,25 @@ public abstract class AbstractTest {
@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);
if (file.startsWith("/")) {
try (InputStream fileInputStream = new FileInputStream(file)) {
return prepareStorage(Path.of(file).getFileName().toString(),
fileInputStream,
cvServiceResponseFileResource.getInputStream(),
imageInfoFileResource.getInputStream(),
visualLayoutParsingResponseResource.getInputStream());
}
} else {
return prepareStorage(Path.of(file).getFileName().toString(),
new ClassPathResource(file).getInputStream(),
cvServiceResponseFileResource.getInputStream(),
imageInfoFileResource.getInputStream(),
visualLayoutParsingResponseResource.getInputStream());
}
return prepareStorage(Path.of(file).getFileName().toString(),
pdfFileResource.getInputStream(),
cvServiceResponseFileResource.getInputStream(),
imageInfoFileResource.getInputStream(),
visualLayoutParsingResponseResource.getInputStream());
}

View File

@ -45,7 +45,12 @@ public abstract class BuildDocumentTest extends AbstractTest {
@SneakyThrows
protected ClassificationDocument parseLayout(String filename, LayoutParsingType layoutParsingType) {
File fileResource = new ClassPathResource(filename).getFile();
File fileResource;
if (filename.startsWith("/")) {
fileResource = new File(filename);
} else {
fileResource = new ClassPathResource(filename).getFile();
}
prepareStorage(filename);
return layoutParsingPipeline.parseLayout(layoutParsingType,
fileResource,
@ -89,6 +94,5 @@ public abstract class BuildDocumentTest extends AbstractTest {
}
}

View File

@ -14,4 +14,6 @@
<appender-ref ref="${logType}"/>
</root>
<logger name="org.apache.fontbox.ttf" level="ERROR"/>
</configuration>

View File

@ -0,0 +1,25 @@
package com.knecon.fforesight.service.viewerdoc.model;
import java.awt.geom.Point2D;
import java.util.LinkedList;
import java.util.List;
import lombok.AccessLevel;
import lombok.Getter;
import lombok.experimental.FieldDefaults;
@Getter
@FieldDefaults(level = AccessLevel.PRIVATE, makeFinal = true)
public class Outline {
List<Entry> entries = new LinkedList<>();
public record Entry(String name, JumpAction action, List<Entry> children) {
}
public record JumpAction(int pageNumber, Point2D position) {
}
}

View File

@ -0,0 +1,78 @@
package com.knecon.fforesight.service.viewerdoc.service;
import com.knecon.fforesight.service.viewerdoc.model.Outline;
import com.pdftron.common.PDFNetException;
import com.pdftron.pdf.Action;
import com.pdftron.pdf.Bookmark;
import com.pdftron.pdf.Destination;
import com.pdftron.pdf.PDFDoc;
import lombok.SneakyThrows;
import lombok.experimental.UtilityClass;
@UtilityClass
public class OutlineUtility {
@SneakyThrows
public void addOutline(PDFDoc doc, Outline outline) {
if (outline.getEntries().isEmpty()) {
return;
}
deleteExistingOutline(doc);
for (Outline.Entry entry : outline.getEntries()) {
Destination destination = createXyzAction(doc, entry);
Action action = Action.createGoto(destination);
Bookmark bookmark = createBookmark(doc, entry, action);
doc.addRootBookmark(bookmark);
addChildren(doc, entry, bookmark);
}
}
@SneakyThrows
private static void addChildren(PDFDoc doc, Outline.Entry parent, Bookmark parentBookmark) {
if (parent.children().isEmpty()) {
return;
}
for (Outline.Entry entry : parent.children()) {
Destination destination = createXyzAction(doc, entry);
Action action = Action.createGoto(destination);
Bookmark bookmark = createBookmark(doc, entry, action);
parentBookmark.addChild(bookmark);
addChildren(doc, entry, bookmark);
}
}
private static Bookmark createBookmark(PDFDoc doc, Outline.Entry entry, Action action) throws PDFNetException {
Bookmark bookmark = Bookmark.create(doc, entry.name());
bookmark.setAction(action);
return bookmark;
}
private static Destination createXyzAction(PDFDoc doc, Outline.Entry entry) throws PDFNetException {
return Destination.createXYZ(doc.getPage(entry.action().pageNumber()), entry.action().position().getX(), entry.action().position().getY(), 1);
}
@SneakyThrows
private static void deleteExistingOutline(PDFDoc doc) {
Bookmark firstBookmark = doc.getFirstBookmark();
while (firstBookmark != null && firstBookmark.isValid()) {
firstBookmark.delete();
firstBookmark = doc.getFirstBookmark();
}
}
}

View File

@ -19,6 +19,7 @@ import com.knecon.fforesight.service.viewerdoc.layers.LayoutDebugLayerConfig;
import com.knecon.fforesight.service.viewerdoc.layers.LayoutGridLayerConfig;
import com.knecon.fforesight.service.viewerdoc.layers.OcrDebugLayerConfig;
import com.knecon.fforesight.service.viewerdoc.model.EmbeddableFont;
import com.knecon.fforesight.service.viewerdoc.model.Outline;
import com.knecon.fforesight.service.viewerdoc.model.PlacedText;
import com.knecon.fforesight.service.viewerdoc.model.Visualizations;
import com.knecon.fforesight.service.viewerdoc.model.VisualizationsOnPage;
@ -27,6 +28,7 @@ import com.pdftron.pdf.ElementReader;
import com.pdftron.pdf.ElementWriter;
import com.pdftron.pdf.Font;
import com.pdftron.pdf.PDFDoc;
import com.pdftron.pdf.PDFNet;
import com.pdftron.pdf.Page;
import com.pdftron.pdf.PageIterator;
import com.pdftron.pdf.ocg.Group;
@ -52,71 +54,83 @@ public class PDFTronViewerDocumentService {
@SneakyThrows
@Observed(name = "PDFTronViewerDocumentService", contextualName = "add-visualizations")
public synchronized void addLayerGroups(File originFile, File destinationFile, List<LayerGroup> layerGroups) {
public void addLayerGroups(File originFile, File destinationFile, List<LayerGroup> layerGroups, Outline outline) {
// originFile and destinationFile might be the same, so we use a temp file.
// Otherwise, saving the document might corrupt the file
Path tmpFile = Files.createTempFile("tmpViewerDocument", ".pdf");
Files.copy(originFile.toPath(), tmpFile, StandardCopyOption.REPLACE_EXISTING);
synchronized (PDFNet.class) { // synchronized with class, to ensure multiple instances are also synchronized
try (PDFDoc pdfDoc = loadPdfDoc(tmpFile);//
ElementWriter pageWriter = new ElementWriter();//
ElementReader reader = new ElementReader();//
ElementBuilder builder = new ElementBuilder()//
) {
enrichObservation(registry,
pdfDoc.getPageCount(),
layerGroups.stream()
.map(LayerGroup::getVisualizations)
.flatMap(Collection::stream)
.map(Visualizations::getLayer)
.toList());
// originFile and destinationFile might be the same, so we use a temp file.
// Otherwise, saving the document might corrupt the file
Path tmpFile = Files.createTempFile("tmpViewerDocument", ".pdf");
Files.copy(originFile.toPath(), tmpFile, StandardCopyOption.REPLACE_EXISTING);
Map<LayerIdentifier, Group> groupMap = PdftronLayerUtility.addLayersToDocument(layerGroups, pdfDoc);
try (PDFDoc pdfDoc = loadPdfDoc(tmpFile);//
ElementWriter pageWriter = new ElementWriter();//
ElementReader reader = new ElementReader();//
ElementBuilder builder = new ElementBuilder()//
) {
enrichObservation(registry,
pdfDoc.getPageCount(),
layerGroups.stream()
.map(LayerGroup::getVisualizations)
.flatMap(Collection::stream)
.map(Visualizations::getLayer)
.toList());
Map<EmbeddableFont, Font> fontMap = buildFontMap(layerGroups, pdfDoc);
Map<LayerIdentifier, Group> groupMap = PdftronLayerUtility.addLayersToDocument(layerGroups, pdfDoc);
Set<String> markedContentToDraw = mapMarkedContentNames(layerGroups);
Map<EmbeddableFont, Font> fontMap = buildFontMap(layerGroups, pdfDoc);
PageContentCleaner pageContentCleaner = PageContentCleaner.builder()
.writer(pageWriter)
.reader(reader)
.elementBuilder(builder)
.markedContentToRemove(markedContentToDraw)
.build();
Set<String> markedContentToDraw = mapMarkedContentNames(layerGroups);
VisualizationWriter visualizationWriter = VisualizationWriter.builder()
.writer(pageWriter)
.builder(builder)
.groupMap(groupMap)
.layerGroups(layerGroups)
.fontMap(fontMap)
.build();
PageContentCleaner pageContentCleaner = PageContentCleaner.builder()
.writer(pageWriter)
.reader(reader)
.elementBuilder(builder)
.markedContentToRemove(markedContentToDraw)
.build();
boolean isCurrentVersion = ViewerDocVersioningUtility.docIsCurrentVersion(pdfDoc);
VisualizationWriter visualizationWriter = VisualizationWriter.builder()
.writer(pageWriter)
.builder(builder)
.groupMap(groupMap)
.layerGroups(layerGroups)
.fontMap(fontMap)
.build();
int pageNumber = 1;
try (PageIterator iterator = pdfDoc.getPageIterator()) {
while (iterator.hasNext()) {
boolean isCurrentVersion = ViewerDocVersioningUtility.docIsCurrentVersion(pdfDoc);
Page page = iterator.next();
int pageNumber = 1;
try (PageIterator iterator = pdfDoc.getPageIterator()) {
while (iterator.hasNext()) {
if (isCurrentVersion) {
pageContentCleaner.removeMarkedContent(page);
Page page = iterator.next();
if (isCurrentVersion) {
pageContentCleaner.removeMarkedContent(page);
}
visualizationWriter.drawVisualizationsOnPage(pageNumber, page);
pageNumber++;
}
visualizationWriter.drawVisualizationsOnPage(pageNumber, page);
pageNumber++;
}
OutlineUtility.addOutline(pdfDoc, outline);
ViewerDocVersioningUtility.setVersionInDocument(pdfDoc);
saveDocument(pdfDoc, destinationFile);
} finally {
assert !tmpFile.toFile().exists() || tmpFile.toFile().delete();
}
ViewerDocVersioningUtility.setVersionInDocument(pdfDoc);
saveDocument(pdfDoc, destinationFile);
} finally {
assert !tmpFile.toFile().exists() || tmpFile.toFile().delete();
}
}
@SneakyThrows
@Observed(name = "PDFTronViewerDocumentService", contextualName = "add-visualizations")
public void addLayerGroups(File originFile, File destinationFile, List<LayerGroup> layerGroups) {
addLayerGroups(originFile, destinationFile, layerGroups, new Outline());
}