Marmelator/layout-parser
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 289 Wiki Activity

TAAS-41: TAAS Document Structure

Browse Source 
* changed TextPageBlock splitting
* changed Header and Footer Classification
* added TAAS Document Structure Prototype
This commit is contained in:
Kilian Schuettler 2023-06-15 14:37:37 +02:00 committed by Timo Bejan
parent dfdeef5812
commit f08c4ced43
51 changed files with 818 additions and 317 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/AtomicPositionBlockData.java → layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/redaction/AtomicPositionBlockData.java
View File

@ -1,4 +1,4 @@
package com.knecon.fforesight.service.layoutparser.internal.api.data;
package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;

2
layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/AtomicTextBlockData.java → layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/redaction/AtomicTextBlockData.java
View File

@ -1,4 +1,4 @@
package com.knecon.fforesight.service.layoutparser.internal.api.data;
package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;

2
layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/DocumentData.java → layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/redaction/DocumentData.java
View File

@ -1,4 +1,4 @@
package com.knecon.fforesight.service.layoutparser.internal.api.data;
package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;

2
layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/DocumentTreeData.java → layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/redaction/DocumentTreeData.java
View File

@ -1,4 +1,4 @@
package com.knecon.fforesight.service.layoutparser.internal.api.data;
package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
import java.util.List;
import java.util.Map;

2
layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/NodeType.java → layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/redaction/NodeType.java
View File

@ -1,4 +1,4 @@
package com.knecon.fforesight.service.layoutparser.internal.api.data;
package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
import java.util.Locale;

2
layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/PageData.java → layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/redaction/PageData.java
View File

@ -1,4 +1,4 @@
package com.knecon.fforesight.service.layoutparser.internal.api.data;
package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;

20
layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/taas/ParagraphData.java Normal file
View File

@ -0,0 +1,20 @@
package com.knecon.fforesight.service.layoutparser.internal.api.data.taas;
import java.util.List;
import lombok.Builder;
import lombok.Data;
@Data
@Builder
public class ParagraphData {
private String text;
List<Range> boldTextBoundaries;
List<Range> italicTextBoundaries;
private String classification;
private String orientation;
private int textDirection;
}

5
layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/taas/Range.java Normal file
View File

@ -0,0 +1,5 @@
package com.knecon.fforesight.service.layoutparser.internal.api.data.taas;
public record Range(int start, int end) {
}

16
layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/taas/ResearchDocumentData.java Normal file
View File

@ -0,0 +1,16 @@
package com.knecon.fforesight.service.layoutparser.internal.api.data.taas;
import java.util.List;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
@Builder
@Data
@AllArgsConstructor
public class ResearchDocumentData {
String originalFile;
List<StructureObject> structureObjects;
}

15
layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/taas/RowData.java Normal file
View File

@ -0,0 +1,15 @@
package com.knecon.fforesight.service.layoutparser.internal.api.data.taas;
import java.util.List;
import lombok.AllArgsConstructor;
import lombok.Data;
@Data
@AllArgsConstructor
public class RowData {
boolean header;
List<String> cellText;
float[] bBox;
}

19
layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/taas/StructureObject.java Normal file
View File

@ -0,0 +1,19 @@
package com.knecon.fforesight.service.layoutparser.internal.api.data.taas;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
@Data
@Builder
@AllArgsConstructor
public class StructureObject {
Integer structureObjectNumber;
int page;
int stringOffset;
float[] boundingBox;
ParagraphData paragraph;
TableData table;
}

15
layoutparser-service/layoutparser-service-internal-api/src/main/java/com/knecon/fforesight/service/layoutparser/internal/api/data/taas/TableData.java Normal file
View File

@ -0,0 +1,15 @@
package com.knecon.fforesight.service.layoutparser.internal.api.data.taas;
import java.util.List;
import lombok.AllArgsConstructor;
import lombok.Data;
@Data
@AllArgsConstructor
public class TableData {
List<RowData> rowData;
Integer numberOfCols;
Integer numberOfRows;
}

25
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/LayoutParsingService.java
View File

@ -19,7 +19,7 @@ import com.knecon.fforesight.service.layoutparser.processor.classification.servi
import com.knecon.fforesight.service.layoutparser.processor.classification.service.SectionsBuilderService;
import com.knecon.fforesight.service.layoutparser.processor.factory.DocumentGraphFactory;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Document;
import com.knecon.fforesight.service.layoutparser.processor.mapper.DocumentDataMapper;
import com.knecon.fforesight.service.layoutparser.processor.mapper.redaction.DocumentDataMapper;
import lombok.RequiredArgsConstructor;
import lombok.extern.slf4j.Slf4j;
@ -84,4 +84,27 @@ public class LayoutParsingService {
return DocumentGraphFactory.buildDocumentGraph(classificationDocument);
}
public Document parseLayoutWithTimer(PDDocument originDocument, ImageServiceResponse imageServiceResponse, TableServiceResponse tableServiceResponse) {
long start = System.currentTimeMillis();
ClassificationDocument classificationDocument = pdfParsingService.parseDocument(originDocument,
cvTableParsingAdapter.buildCvParsedTablesPerPage(tableServiceResponse),
imageServiceResponseAdapter.buildClassifiedImagesPerPage(imageServiceResponse));
System.out.printf("parsed %d ms", System.currentTimeMillis() - start);
start = System.currentTimeMillis();
classificationService.classifyDocument(classificationDocument);
System.out.printf(", classified %d ms", System.currentTimeMillis() - start);
start = System.currentTimeMillis();
sectionsBuilderService.buildSections(classificationDocument);
System.out.printf(", sections built %d ms", System.currentTimeMillis() - start);
start = System.currentTimeMillis();
Document document = DocumentGraphFactory.buildDocumentGraph(classificationDocument);
System.out.printf(", graph constructed %d ms", System.currentTimeMillis() - start);
return document;
}
}

10
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/LayoutParsingStorageService.java
View File

@ -14,11 +14,11 @@ import org.springframework.stereotype.Service;
import com.fasterxml.jackson.databind.ObjectMapper;
import com.iqser.red.storage.commons.service.StorageService;
import com.knecon.fforesight.service.layoutparser.internal.api.data.AtomicPositionBlockData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.AtomicTextBlockData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.DocumentData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.DocumentTreeData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.PageData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.AtomicPositionBlockData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.AtomicTextBlockData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentTreeData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.PageData;
import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingRequest;
import com.knecon.fforesight.service.layoutparser.processor.adapter.model.image.ImageServiceResponse;
import com.knecon.fforesight.service.layoutparser.processor.adapter.model.table.TableServiceResponse;

67
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/classification/model/text/TextPageBlock.java
View File

@ -1,21 +1,28 @@
package com.knecon.fforesight.service.layoutparser.processor.classification.model.text;
import static java.util.stream.Collectors.toSet;
import java.util.ArrayList;
import java.util.Comparator;
import java.util.List;
import com.fasterxml.jackson.annotation.JsonIgnore;
import com.knecon.fforesight.service.layoutparser.processor.classification.model.AbstractPageBlock;
import com.knecon.fforesight.service.layoutparser.processor.classification.model.FloatFrequencyCounter;
import com.knecon.fforesight.service.layoutparser.processor.classification.model.PageBlockType;
import com.knecon.fforesight.service.layoutparser.processor.classification.utils.DoubleComparisons;
import com.knecon.fforesight.service.layoutparser.processor.classification.utils.TextNormalizationUtilities;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
import lombok.EqualsAndHashCode;
import lombok.NoArgsConstructor;
@EqualsAndHashCode(callSuper = true)
@Data
@AllArgsConstructor
@Builder
@Data
@NoArgsConstructor
public class TextPageBlock extends AbstractPageBlock {
@ -67,6 +74,64 @@ public class TextPageBlock extends AbstractPageBlock {
return sequences.get(0).getPageWidth();
}
public static TextPageBlock merge(List<TextPageBlock> textBlocksToMerge) {
List<TextPositionSequence> sequences = textBlocksToMerge.stream().map(TextPageBlock::getSequences).flatMap(java.util.Collection::stream).toList();
sequences = new ArrayList<>(sequences);
return fromTextPositionSequences(sequences);
}
public static TextPageBlock fromTextPositionSequences(List<TextPositionSequence> wordBlockList) {
TextPageBlock textBlock = null;
FloatFrequencyCounter lineHeightFrequencyCounter = new FloatFrequencyCounter();
FloatFrequencyCounter fontSizeFrequencyCounter = new FloatFrequencyCounter();
FloatFrequencyCounter spaceFrequencyCounter = new FloatFrequencyCounter();
StringFrequencyCounter fontFrequencyCounter = new StringFrequencyCounter();
StringFrequencyCounter styleFrequencyCounter = new StringFrequencyCounter();
for (TextPositionSequence wordBlock : wordBlockList) {
lineHeightFrequencyCounter.add(wordBlock.getTextHeight());
fontSizeFrequencyCounter.add(wordBlock.getFontSize());
spaceFrequencyCounter.add(wordBlock.getSpaceWidth());
fontFrequencyCounter.add(wordBlock.getFont());
styleFrequencyCounter.add(wordBlock.getFontStyle());
if (textBlock == null) {
textBlock = new TextPageBlock(wordBlock.getMinXDirAdj(),
wordBlock.getMaxXDirAdj(),
wordBlock.getMinYDirAdj(),
wordBlock.getMaxYDirAdj(),
wordBlockList,
wordBlock.getRotation());
} else {
TextPageBlock spatialEntity = textBlock.union(wordBlock);
textBlock.resize(spatialEntity.getMinX(), spatialEntity.getMinY(), spatialEntity.getWidth(), spatialEntity.getHeight());
}
}
if (textBlock != null) {
textBlock.setMostPopularWordFont(fontFrequencyCounter.getMostPopular());
textBlock.setMostPopularWordStyle(styleFrequencyCounter.getMostPopular());
textBlock.setMostPopularWordFontSize(fontSizeFrequencyCounter.getMostPopular());
textBlock.setMostPopularWordHeight(lineHeightFrequencyCounter.getMostPopular());
textBlock.setMostPopularWordSpaceWidth(spaceFrequencyCounter.getMostPopular());
textBlock.setHighestFontSize(fontSizeFrequencyCounter.getHighest());
}
if (textBlock != null && textBlock.getSequences() != null && textBlock.getSequences()
.stream()
.map(t -> DoubleComparisons.round(t.getMinYDirAdj(), 3))
.collect(toSet())
.size() == 1) {
textBlock.getSequences().sort(Comparator.comparing(TextPositionSequence::getMinXDirAdj));
}
return textBlock;
}
/**
* Returns the minX value in pdf coordinate system.

20
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/classification/model/text/TextPositionSequence.java
View File

@ -8,8 +8,8 @@ import java.util.stream.Collectors;
import org.apache.pdfbox.text.TextPosition;
import com.dslplatform.json.JsonAttribute;
import com.fasterxml.jackson.annotation.JsonIgnore;
import com.fasterxml.jackson.annotation.JsonIgnoreProperties;
import com.iqser.red.service.persistence.service.v1.api.shared.model.redactionlog.Point;
import com.iqser.red.service.persistence.service.v1.api.shared.model.redactionlog.Rectangle;
@ -25,7 +25,6 @@ import lombok.extern.slf4j.Slf4j;
@Builder
@NoArgsConstructor
@AllArgsConstructor
@JsonIgnoreProperties({"empty"})
public class TextPositionSequence implements CharSequence {
public static final int HEIGHT_PADDING = 2;
@ -36,6 +35,7 @@ public class TextPositionSequence implements CharSequence {
private int rotation;
private float pageHeight;
private float pageWidth;
private boolean isParagraphStart;
public TextPositionSequence(int page) {
@ -44,7 +44,7 @@ public class TextPositionSequence implements CharSequence {
}
public TextPositionSequence(List<TextPosition> textPositions, int page) {
public TextPositionSequence(List<TextPosition> textPositions, int page, boolean isParagraphStart) {
this.textPositions = textPositions.stream().map(RedTextPosition::fromTextPosition).collect(Collectors.toList());
this.page = page;
@ -52,6 +52,7 @@ public class TextPositionSequence implements CharSequence {
this.rotation = textPositions.get(0).getRotation();
this.pageHeight = textPositions.get(0).getPageHeight();
this.pageWidth = textPositions.get(0).getPageWidth();
this.isParagraphStart = isParagraphStart;
}
@ -141,6 +142,7 @@ public class TextPositionSequence implements CharSequence {
* @return the text direction adjusted minX value
*/
@JsonIgnore
@JsonAttribute(ignore = true)
public float getMinXDirAdj() {
return textPositions.get(0).getXDirAdj();
@ -155,6 +157,7 @@ public class TextPositionSequence implements CharSequence {
* @return the text direction adjusted maxX value
*/
@JsonIgnore
@JsonAttribute(ignore = true)
public float getMaxXDirAdj() {
return textPositions.get(textPositions.size() - 1).getXDirAdj() + textPositions.get(textPositions.size() - 1).getWidthDirAdj() + HEIGHT_PADDING;
@ -169,6 +172,7 @@ public class TextPositionSequence implements CharSequence {
* @return the text direction adjusted minY value. The upper border of the bounding box of the word.
*/
@JsonIgnore
@JsonAttribute(ignore = true)
public float getMinYDirAdj() {
return textPositions.get(0).getYDirAdj() - getTextHeight();
@ -183,6 +187,7 @@ public class TextPositionSequence implements CharSequence {
* @return the text direction adjusted maxY value. The lower border of the bounding box of the word.
*/
@JsonIgnore
@JsonAttribute(ignore = true)
public float getMaxYDirAdj() {
return textPositions.get(0).getYDirAdj();
@ -191,6 +196,7 @@ public class TextPositionSequence implements CharSequence {
@JsonIgnore
@JsonAttribute(ignore = true)
public float getTextHeight() {
return textPositions.get(0).getHeightDir() + HEIGHT_PADDING;
@ -198,6 +204,7 @@ public class TextPositionSequence implements CharSequence {
@JsonIgnore
@JsonAttribute(ignore = true)
public float getHeight() {
return getMaxYDirAdj() - getMinYDirAdj();
@ -205,6 +212,7 @@ public class TextPositionSequence implements CharSequence {
@JsonIgnore
@JsonAttribute(ignore = true)
public float getWidth() {
return getMaxXDirAdj() - getMinXDirAdj();
@ -212,6 +220,7 @@ public class TextPositionSequence implements CharSequence {
@JsonIgnore
@JsonAttribute(ignore = true)
public String getFont() {
return textPositions.get(0).getFontName().toLowerCase().replaceAll(",bold", "").replaceAll(",italic", "");
@ -219,6 +228,7 @@ public class TextPositionSequence implements CharSequence {
@JsonIgnore
@JsonAttribute(ignore = true)
public String getFontStyle() {
String lowercaseFontName = textPositions.get(0).getFontName().toLowerCase();
@ -237,6 +247,7 @@ public class TextPositionSequence implements CharSequence {
@JsonIgnore
@JsonAttribute(ignore = true)
public float getFontSize() {
return textPositions.get(0).getFontSizeInPt();
@ -244,6 +255,7 @@ public class TextPositionSequence implements CharSequence {
@JsonIgnore
@JsonAttribute(ignore = true)
public float getSpaceWidth() {
return textPositions.get(0).getWidthOfSpace();
@ -260,6 +272,7 @@ public class TextPositionSequence implements CharSequence {
* @return bounding box of the word in Pdf Coordinate System
*/
@JsonIgnore
@JsonAttribute(ignore = true)
@SneakyThrows
public Rectangle getRectangle() {
@ -299,3 +312,4 @@ public class TextPositionSequence implements CharSequence {
}
}

9
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/classification/parsing/LegacyPDFStreamEngine.java
View File

@ -76,7 +76,7 @@ import org.apache.pdfbox.util.Vector;
* THIS CODE IS DELIBERATELY INCORRECT, USE PDFStreamEngine INSTEAD.
*/
@SuppressWarnings({"PMD", "checkstyle:all"})
class LegacyPDFStreamEngine extends PDFStreamEngine {
public class LegacyPDFStreamEngine extends PDFStreamEngine {
private static final Log LOG = LogFactory.getLog(LegacyPDFStreamEngine.class);
@ -126,7 +126,7 @@ class LegacyPDFStreamEngine extends PDFStreamEngine {
* This will initialize and process the contents of the stream.
*
* @param page the page to process
* @throws IOException if there is an error accessing the stream.
* @throws java.io.IOException if there is an error accessing the stream.
*/
@Override
public void processPage(PDPage page) throws IOException {
@ -149,7 +149,7 @@ class LegacyPDFStreamEngine extends PDFStreamEngine {
* written by Ben Litchfield for PDFStreamEngine.
*/
@Override
protected void showGlyph(Matrix textRenderingMatrix, PDFont font, int code,Vector displacement) throws IOException {
protected void showGlyph(Matrix textRenderingMatrix, PDFont font, int code, Vector displacement) throws IOException {
//
// legacy calculations which were previously in PDFStreamEngine
//
@ -165,7 +165,7 @@ class LegacyPDFStreamEngine extends PDFStreamEngine {
float displacementX = displacement.getX();
// the sorting algorithm is based on the width of the character. As the displacement
// for vertical characters doesn't provide any suitable value for it, we have to
// for vertical characters doesn't provide any suitable value for it, we have to
// calculate our own
if (font.isVertical()) {
displacementX = font.getWidth(code) / 1000;
@ -382,3 +382,4 @@ class LegacyPDFStreamEngine extends PDFStreamEngine {
}
}

18
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/classification/parsing/PDFLinesTextStripper.java
View File

@ -3,7 +3,6 @@ package com.knecon.fforesight.service.layoutparser.processor.classification.pars
import java.awt.geom.Point2D;
import java.io.IOException;
import java.util.ArrayList;
import java.util.Comparator;
import java.util.List;
import org.apache.pdfbox.contentstream.operator.Operator;
@ -208,13 +207,11 @@ public class PDFLinesTextStripper extends PDFTextStripper {
@Override
public void writeString(String text, List<TextPosition> textPositions) throws IOException {
public void writeString(String text, List<TextPosition> textPositions, boolean isParagraphStart) throws IOException {
int startIndex = 0;
RedTextPosition previous = null;
textPositions.sort(Comparator.comparing(TextPosition::getXDirAdj));
for (int i = 0; i <= textPositions.size() - 1; i++) {
if (!textPositionSequences.isEmpty()) {
@ -250,7 +247,7 @@ public class PDFLinesTextStripper extends PDFTextStripper {
if (!(sublist.isEmpty() || sublist.size() == 1 && (sublist.get(0).getUnicode().equals(" ") || sublist.get(0).getUnicode().equals("\u00A0") || sublist.get(0)
.getUnicode()
.equals("\t")))) {
textPositionSequences.add(new TextPositionSequence(sublist, pageNumber));
textPositionSequences.add(new TextPositionSequence(sublist, pageNumber, i == textPositions.size() -1 && isParagraphStart));
}
startIndex = i;
}
@ -260,7 +257,7 @@ public class PDFLinesTextStripper extends PDFTextStripper {
if (!(sublist.isEmpty() || sublist.size() == 1 && (sublist.get(0).getUnicode().equals(" ") || sublist.get(0).getUnicode().equals("\u00A0") || sublist.get(0)
.getUnicode()
.equals("\t")))) {
textPositionSequences.add(new TextPositionSequence(sublist, pageNumber));
textPositionSequences.add(new TextPositionSequence(sublist, pageNumber, i == textPositions.size() -1 && isParagraphStart));
}
startIndex = i;
}
@ -276,11 +273,11 @@ public class PDFLinesTextStripper extends PDFTextStripper {
// Remove false sequence ends (whitespaces)
if (previous != null && sublist.get(0).getYDirAdj() == previous.getYDirAdj() && sublist.get(0)
.getXDirAdj() - (previous.getXDirAdj() + previous.getWidthDirAdj()) < 0.01) {
for (TextPosition textPosition : sublist) {
textPositionSequences.get(textPositionSequences.size() - 1).add(textPosition);
for (TextPosition t : sublist) {
textPositionSequences.get(textPositionSequences.size() - 1).add(t);
}
} else {
textPositionSequences.add(new TextPositionSequence(sublist, pageNumber));
textPositionSequences.add(new TextPositionSequence(sublist, pageNumber, i == textPositions.size() -1 && isParagraphStart));
}
}
startIndex = i + 1;
@ -303,7 +300,7 @@ public class PDFLinesTextStripper extends PDFTextStripper {
textPositionSequences.get(textPositionSequences.size() - 1).add(t);
}
} else {
textPositionSequences.add(new TextPositionSequence(sublist, pageNumber));
textPositionSequences.add(new TextPositionSequence(sublist, pageNumber, isParagraphStart));
}
}
super.writeString(text);
@ -328,3 +325,4 @@ public class PDFLinesTextStripper extends PDFTextStripper {
}
}

62
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/classification/parsing/PDFTextStripper.java
View File

@ -27,6 +27,7 @@ import java.text.Bidi;
import java.text.Normalizer;
import java.util.ArrayList;
import java.util.Collections;
import java.util.Comparator;
import java.util.HashMap;
import java.util.Iterator;
import java.util.LinkedList;
@ -240,10 +241,10 @@ public class PDFTextStripper extends LegacyPDFStreamEngine {
document = doc;
output = outputStream;
if (getAddMoreFormatting()) {
paragraphEnd = lineSeparator;
paragraphEnd = "\n----ParagraphEnd----\n\n";
pageStart = lineSeparator;
articleStart = lineSeparator;
articleEnd = lineSeparator;
articleStart = "\n----ArticelStart----\n\n";
articleEnd = "\n----ArticelEnd----\n\n";
}
startDocument(document);
processPages(document.getPages());
@ -594,9 +595,14 @@ public class PDFTextStripper extends LegacyPDFStreamEngine {
// but this caused a lot of regression test failures. So, I'm leaving it be for
// now
if (!overlap(positionY, positionHeight, maxYForLine, maxHeightForLine)) {
writeLine(normalize(line));
line.clear();
var normalized = normalize(line);
// normalized.stream().filter(l -> System.out.println(l.getText().contains("Plenarprotokoll 20/24")).findFirst().isPresent()
lastLineStartPosition = handleLineSeparation(current, lastPosition, lastLineStartPosition, maxHeightForLine);
writeLine(normalized, current.isParagraphStart);
line.clear();
expectedStartOfNextWordX = EXPECTED_START_OF_NEXT_WORD_X_RESET_VALUE;
maxYForLine = MAX_Y_FOR_LINE_RESET_VALUE;
maxHeightForLine = MAX_HEIGHT_FOR_LINE_RESET_VALUE;
@ -630,7 +636,24 @@ public class PDFTextStripper extends LegacyPDFStreamEngine {
if (startOfPage && lastPosition == null) {
writeParagraphStart();// not sure this is correct for RTL?
}
line.add(new LineItem(position));
// Collections.sort(line, new Comparator<LineItem>() {
//
// @Override
// public int compare(LineItem str1, LineItem str2) {
// if(null == str1.getTextPosition()) {
// return 0;
// }
// else if(null == str2.getTextPosition()) {
// return 0;
// }
// return Float.compare(str1.getTextPosition().getX(), str2.getTextPosition().getX());
// }
// });
// line.sort(Comparator.comparing(a -> a.getTextPosition() != null && a.getTextPosition().getX()));
}
maxHeightForLine = Math.max(maxHeightForLine, positionHeight);
minYTopForLine = Math.min(minYTopForLine, positionY - positionHeight);
@ -646,7 +669,7 @@ public class PDFTextStripper extends LegacyPDFStreamEngine {
}
// print the final line
if (line.size() > 0) {
writeLine(normalize(line));
writeLine(normalize(line), false);
writeParagraphEnd();
}
endArticle();
@ -703,7 +726,7 @@ public class PDFTextStripper extends LegacyPDFStreamEngine {
* @param textPositions The TextPositions belonging to the text.
* @throws IOException If there is an error when writing the text.
*/
protected void writeString(String text, List<TextPosition> textPositions) throws IOException {
protected void writeString(String text, List<TextPosition> textPositions, boolean isParagraphEnd) throws IOException {
writeString(text);
}
@ -998,7 +1021,7 @@ public class PDFTextStripper extends LegacyPDFStreamEngine {
/**
* By default the text stripper will attempt to remove text that overlapps each other. Word paints the same
* By default, the text stripper will attempt to remove text that overlapps each other. Word paints the same
* character several times in order to make it look bold. By setting this to false all text will be extracted, which
* means that certain sections will be duplicated, but better performance will be noticed.
*
@ -1385,6 +1408,7 @@ public class PDFTextStripper extends LegacyPDFStreamEngine {
} else {
writeLineSeparator();
writeParagraphSeparator();
lastLineStartPosition.setEndParagraphWritten();
}
} else {
writeLineSeparator();
@ -1428,6 +1452,10 @@ public class PDFTextStripper extends LegacyPDFStreamEngine {
float newXVal = multiplyFloat(getIndentThreshold(), position.getTextPosition().getWidthOfSpace());
float positionWidth = multiplyFloat(0.25f, position.getTextPosition().getWidth());
// if(xGap < 0){
// result = true;
// }
// else
if (yGap > newYVal) {
result = true;
} else if (xGap > newXVal) {
@ -1636,12 +1664,13 @@ public class PDFTextStripper extends LegacyPDFStreamEngine {
* @param line a list with the words of the given line
* @throws IOException if something went wrong
*/
private void writeLine(List<WordWithTextPositions> line) throws IOException {
private void writeLine(List<WordWithTextPositions> line, boolean isParagraphEnd) throws IOException {
int numberOfStrings = line.size();
for (int i = 0; i < numberOfStrings; i++) {
WordWithTextPositions word = line.get(i);
writeString(word.getText(), word.getTextPositions());
word.getTextPositions().sort(Comparator.comparing(TextPosition::getX));
writeString(word.getText(), word.getTextPositions(), isParagraphEnd && i == numberOfStrings - 1);
if (i < numberOfStrings - 1) {
writeWordSeparator();
}
@ -1963,6 +1992,8 @@ public class PDFTextStripper extends LegacyPDFStreamEngine {
private boolean isHangingIndent = false;
private boolean isArticleStart = false;
private boolean endParagraphWritten = false;
private TextPosition position = null;
@ -2024,6 +2055,16 @@ public class PDFTextStripper extends LegacyPDFStreamEngine {
}
public boolean isEndParagraphWritten() {
return endParagraphWritten;
}
public void setEndParagraphWritten(){
endParagraphWritten = true;
}
/**
* Sets the isArticleStart() flag to true.
*/
@ -2065,3 +2106,4 @@ public class PDFTextStripper extends LegacyPDFStreamEngine {
}
}

266
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/classification/service/BlockificationService.java
View File

@ -1,20 +1,18 @@
package com.knecon.fforesight.service.layoutparser.processor.classification.service;
import static java.util.stream.Collectors.toSet;
import java.util.ArrayList;
import java.util.Comparator;
import java.util.Iterator;
import java.util.LinkedList;
import java.util.List;
import java.util.regex.Matcher;
import java.util.regex.Pattern;
import org.springframework.stereotype.Service;
import com.knecon.fforesight.service.layoutparser.processor.classification.model.AbstractPageBlock;
import com.knecon.fforesight.service.layoutparser.processor.classification.model.ClassificationPage;
import com.knecon.fforesight.service.layoutparser.processor.classification.model.FloatFrequencyCounter;
import com.knecon.fforesight.service.layoutparser.processor.classification.model.Orientation;
import com.knecon.fforesight.service.layoutparser.processor.classification.model.table.Ruling;
import com.knecon.fforesight.service.layoutparser.processor.classification.model.text.StringFrequencyCounter;
import com.knecon.fforesight.service.layoutparser.processor.classification.model.text.TextPageBlock;
import com.knecon.fforesight.service.layoutparser.processor.classification.model.text.TextPositionSequence;
import com.knecon.fforesight.service.layoutparser.processor.classification.utils.RulingTextDirAdjustUtil;
@ -23,95 +21,64 @@ import com.knecon.fforesight.service.layoutparser.processor.classification.utils
@SuppressWarnings("all")
public class BlockificationService {
static final float THRESHOLD = 1f;
private static final float THRESHOLD = 1f;
private static final float Y_GAP_SPLIT_HEIGHT_MODIFIER = 1.25f;
private static final int X_GAP_SPLIT_CONSTANT = 50;
/**
* This method is building blocks by expanding the minX/maxX and minY/maxY value on each word that is not split by the conditions.
* This method must use text direction adjusted postions (DirAdj). Where {0,0} is on the upper left. Never try to change this!
* Rulings (TablePageBlock lines) must be adjusted to the text directions as well, when checking if a block is split by a ruling.
* Rulings (Table lines) must be adjusted to the text directions as well, when checking if a block is split by a ruling.
*
* @param textPositions The words of a page.
* @param horizontalRulingLines Horizontal table lines.
* @param verticalRulingLines Vertical table lines.
* @return Page object that contains the Textblock and text statistics.
* @return ClassificationPage object that contains the Textblock and text statistics.
*/
public ClassificationPage blockify(List<TextPositionSequence> textPositions, List<Ruling> horizontalRulingLines, List<Ruling> verticalRulingLines) {
int indexOnPage = 0;
List<TextPositionSequence> chunkWords = new ArrayList<>();
List<AbstractPageBlock> chunkBlockList = new ArrayList<>();
List<TextPageBlock> classificationTextBlocks = constructFineGranularTextPageBlocks(textPositions, horizontalRulingLines, verticalRulingLines);
float minX = 1000, maxX = 0, minY = 1000, maxY = 0;
TextPositionSequence prev = null;
classificationTextBlocks = mergeFineGranularTextPageBlocks(classificationTextBlocks);
boolean wasSplitted = false;
Float splitX1 = null;
for (TextPositionSequence word : textPositions) {
return new ClassificationPage(new ArrayList<>(classificationTextBlocks.stream().map(classificationTextBlock -> (AbstractPageBlock) classificationTextBlock).toList()));
}
boolean lineSeparation = word.getMinYDirAdj() - maxY > word.getHeight() * 1.25;
boolean startFromTop = prev != null && word.getMinYDirAdj() < prev.getMinYDirAdj() - prev.getTextHeight();
boolean splitByX = prev != null && maxX + 50 < word.getMinXDirAdj() && prev.getMinYDirAdj() == word.getMinYDirAdj();
boolean xIsBeforeFirstX = prev != null && word.getMinXDirAdj() < minX;
boolean newLineAfterSplit = prev != null && word.getMinYDirAdj() != prev.getMinYDirAdj() && wasSplitted && splitX1 != word.getMinXDirAdj();
boolean isSplitByRuling = isSplitByRuling(minX, minY, maxX, maxY, word, horizontalRulingLines, verticalRulingLines);
boolean splitByDir = prev != null && !prev.getDir().equals(word.getDir());
if (prev != null && (lineSeparation || startFromTop || splitByX || splitByDir || isSplitByRuling)) {
private List<TextPageBlock> mergeFineGranularTextPageBlocks(List<TextPageBlock> classificationTextBlocks) {
Orientation prevOrientation = null;
if (!chunkBlockList.isEmpty()) {
prevOrientation = chunkBlockList.get(chunkBlockList.size() - 1).getOrientation();
}
TextPageBlock cb1 = buildTextBlock(chunkWords, indexOnPage);
indexOnPage++;
chunkBlockList.add(cb1);
chunkWords = new ArrayList<>();
if (splitByX && !isSplitByRuling) {
wasSplitted = true;
cb1.setOrientation(Orientation.LEFT);
splitX1 = word.getMinXDirAdj();
} else if (newLineAfterSplit && !isSplitByRuling) {
wasSplitted = false;
cb1.setOrientation(Orientation.RIGHT);
splitX1 = null;
} else if (prevOrientation != null && prevOrientation.equals(Orientation.RIGHT) && (lineSeparation || !startFromTop || !splitByX || !newLineAfterSplit || !isSplitByRuling)) {
cb1.setOrientation(Orientation.LEFT);
}
minX = 1000;
maxX = 0;
minY = 1000;
maxY = 0;
prev = null;
}
chunkWords.add(word);
prev = word;
if (word.getMinXDirAdj() < minX) {
minX = word.getMinXDirAdj();
}
if (word.getMaxXDirAdj() > maxX) {
maxX = word.getMaxXDirAdj();
}
if (word.getMinYDirAdj() < minY) {
minY = word.getMinYDirAdj();
}
if (word.getMaxYDirAdj() > maxY) {
maxY = word.getMaxYDirAdj();
}
if (classificationTextBlocks.isEmpty()) {
return new ArrayList<>();
}
TextPageBlock cb1 = buildTextBlock(chunkWords, indexOnPage);
if (cb1 != null) {
chunkBlockList.add(cb1);
List<List<TextPageBlock>> textBlocksToMerge = new LinkedList<>();
List<TextPageBlock> currentTextBlocksToMerge = new LinkedList<>();
textBlocksToMerge.add(currentTextBlocksToMerge);
TextPageBlock previousTextBlock = null;
for (TextPageBlock currentTextBlock : classificationTextBlocks) {
if (previousTextBlock == null) {
currentTextBlocksToMerge.add(currentTextBlock);
previousTextBlock = currentTextBlock;
continue;
}
boolean alignsXRight = Math.abs(currentTextBlock.getPdfMaxX() - previousTextBlock.getPdfMaxX()) < 1;
boolean smallYGap = Math.abs(currentTextBlock.getPdfMaxY() - previousTextBlock.getPdfMinY()) < 5;
if (alignsXRight && smallYGap) {
currentTextBlocksToMerge.add(currentTextBlock);
} else {
currentTextBlocksToMerge = new LinkedList<>();
currentTextBlocksToMerge.add(currentTextBlock);
textBlocksToMerge.add(currentTextBlocksToMerge);
}
previousTextBlock = currentTextBlock;
}
return textBlocksToMerge.stream().map(TextPageBlock::merge).toList();
}
Iterator<AbstractPageBlock> itty = chunkBlockList.iterator();
private void assignOrientations(List<TextPageBlock> classificationTextBlocks) {
Iterator<TextPageBlock> itty = classificationTextBlocks.iterator();
TextPageBlock previousLeft = null;
TextPageBlock previousRight = null;
@ -141,12 +108,13 @@ public class BlockificationService {
}
}
itty = chunkBlockList.iterator();
itty = classificationTextBlocks.iterator();
TextPageBlock previous = null;
while (itty.hasNext()) {
TextPageBlock block = (TextPageBlock) itty.next();
if (previous != null && previous.getOrientation().equals(Orientation.LEFT) && block.getOrientation().equals(Orientation.LEFT) && equalsWithThreshold(block.getMaxY(),
if (previous != null && previous.getOrientation().equals(Orientation.LEFT) && block.getOrientation().equals(Orientation.LEFT) && equalsWithThreshold(
block.getMaxY(),
previous.getMaxY()) || previous != null && previous.getOrientation().equals(Orientation.LEFT) && block.getOrientation()
.equals(Orientation.RIGHT) && equalsWithThreshold(block.getMaxY(), previous.getMaxY())) {
previous.add(block);
@ -156,8 +124,95 @@ public class BlockificationService {
previous = block;
}
}
return new ClassificationPage(chunkBlockList);
private List<TextPageBlock> constructFineGranularTextPageBlocks(List<TextPositionSequence> textPositions,
List<Ruling> horizontalRulingLines,
List<Ruling> verticalRulingLines) {
int indexOnPage = 0;
List<TextPositionSequence> wordClusterToCombine = new ArrayList<>();
List<TextPageBlock> classificationTextBlocks = new ArrayList<>();
float minX = 1000, maxX = 0, minY = 1000, maxY = 0;
TextPositionSequence prev = null;
var listIdentitifier = Pattern.compile("\\b(?:[1-9]|1\\d|20|[ivxlc]|[a-z])\\s*(?:[.)])", Pattern.CASE_INSENSITIVE);
boolean wasSplitted = false;
Float splitX1 = null;
for (TextPositionSequence word : textPositions) {
Matcher listIdentifierPattern = listIdentitifier.matcher(word.toString());
boolean yGap = word.getMinYDirAdj() - maxY > word.getHeight() * Y_GAP_SPLIT_HEIGHT_MODIFIER;
boolean sameLine = prev != null && equalsWithThreshold(prev.getMinYDirAdj(), word.getMinYDirAdj());
boolean positiveXGapInline = prev != null && maxX + X_GAP_SPLIT_CONSTANT < word.getMinXDirAdj() && sameLine;
boolean negativeXGap = prev != null && word.getMinXDirAdj() - minX < -5;
boolean startFromTop = prev != null && word.getMinYDirAdj() < prev.getMinYDirAdj() - prev.getTextHeight();
boolean newLineAfterSplit = prev != null && word.getMinYDirAdj() != prev.getMinYDirAdj() && wasSplitted && splitX1 != word.getMinXDirAdj();
boolean splitByRuling = isSplitByRuling(minX, minY, maxX, maxY, word, horizontalRulingLines, verticalRulingLines);
boolean splitByDir = prev != null && !prev.getDir().equals(word.getDir());
boolean fontChange = prev != null && (!word.getFont().equals(prev.getFont()) || !word.getFontStyle()
.equals(prev.getFontStyle()) || word.getFontSize() != prev.getFontSize());
boolean newline = prev != null && Math.abs(word.getMinYDirAdj() - prev.getMinYDirAdj()) > word.getHeight();
boolean isListIdentifier = listIdentifierPattern.matches();
if (prev != null && (prev.isParagraphStart() || negativeXGap || positiveXGapInline || yGap || startFromTop || splitByRuling || (newline && (fontChange || isListIdentifier)))) {
// if (prev != null && (lineSeparation || startFromTop || splitByX || splitByDir || isSplitByRuling)) {
Orientation prevOrientation = null;
if (!classificationTextBlocks.isEmpty()) {
prevOrientation = classificationTextBlocks.get(classificationTextBlocks.size() - 1).getOrientation();
}
TextPageBlock classificationTextBlock = TextPageBlock.fromTextPositionSequences(wordClusterToCombine);
classificationTextBlocks.add(classificationTextBlock);
wordClusterToCombine = new ArrayList<>();
if (positiveXGapInline && !splitByRuling) {
wasSplitted = true;
classificationTextBlock.setOrientation(Orientation.LEFT);
splitX1 = word.getMinXDirAdj();
} else if (newLineAfterSplit && !splitByRuling) {
wasSplitted = false;
classificationTextBlock.setOrientation(Orientation.RIGHT);
splitX1 = null;
} else if (prevOrientation != null && prevOrientation.equals(Orientation.RIGHT) && (yGap || !startFromTop || !positiveXGapInline || !newLineAfterSplit || !splitByRuling)) {
classificationTextBlock.setOrientation(Orientation.LEFT);
}
minX = 1000;
maxX = 0;
minY = 1000;
maxY = 0;
prev = null;
}
wordClusterToCombine.add(word);
prev = word;
if (word.getMinXDirAdj() < minX) {
minX = word.getMinXDirAdj();
}
if (word.getMaxXDirAdj() > maxX) {
maxX = word.getMaxXDirAdj();
}
if (word.getMinYDirAdj() < minY) {
minY = word.getMinYDirAdj();
}
if (word.getMaxYDirAdj() > maxY) {
maxY = word.getMaxYDirAdj();
}
}
TextPageBlock classificationTextBlock = TextPageBlock.fromTextPositionSequences(wordClusterToCombine);
if (classificationTextBlock != null) {
classificationTextBlocks.add(classificationTextBlock);
}
return classificationTextBlocks;
}
@ -167,53 +222,6 @@ public class BlockificationService {
}
private TextPageBlock buildTextBlock(List<TextPositionSequence> wordBlockList, int indexOnPage) {
TextPageBlock textBlock = null;
FloatFrequencyCounter lineHeightFrequencyCounter = new FloatFrequencyCounter();
FloatFrequencyCounter fontSizeFrequencyCounter = new FloatFrequencyCounter();
FloatFrequencyCounter spaceFrequencyCounter = new FloatFrequencyCounter();
StringFrequencyCounter fontFrequencyCounter = new StringFrequencyCounter();
StringFrequencyCounter styleFrequencyCounter = new StringFrequencyCounter();
for (TextPositionSequence wordBlock : wordBlockList) {
lineHeightFrequencyCounter.add(wordBlock.getTextHeight());
fontSizeFrequencyCounter.add(wordBlock.getFontSize());
spaceFrequencyCounter.add(wordBlock.getSpaceWidth());
fontFrequencyCounter.add(wordBlock.getFont());
styleFrequencyCounter.add(wordBlock.getFontStyle());
if (textBlock == null) {
textBlock = new TextPageBlock(wordBlock.getMinXDirAdj(),
wordBlock.getMaxXDirAdj(),
wordBlock.getMinYDirAdj(),
wordBlock.getMaxYDirAdj(),
wordBlockList,
wordBlock.getRotation());
} else {
TextPageBlock spatialEntity = textBlock.union(wordBlock);
textBlock.resize(spatialEntity.getMinX(), spatialEntity.getMinY(), spatialEntity.getWidth(), spatialEntity.getHeight());
}
}
if (textBlock != null) {
textBlock.setMostPopularWordFont(fontFrequencyCounter.getMostPopular());
textBlock.setMostPopularWordStyle(styleFrequencyCounter.getMostPopular());
textBlock.setMostPopularWordFontSize(fontSizeFrequencyCounter.getMostPopular());
textBlock.setMostPopularWordHeight(lineHeightFrequencyCounter.getMostPopular());
textBlock.setMostPopularWordSpaceWidth(spaceFrequencyCounter.getMostPopular());
textBlock.setHighestFontSize(fontSizeFrequencyCounter.getHighest());
}
if (textBlock != null && textBlock.getSequences() != null && textBlock.getSequences().stream().map(t -> round(t.getMinYDirAdj(), 3)).collect(toSet()).size() == 1) {
textBlock.getSequences().sort(Comparator.comparing(TextPositionSequence::getMinXDirAdj));
}
return textBlock;
}
private boolean isSplitByRuling(float minX,
float minY,
float maxX,
@ -253,7 +261,7 @@ public class BlockificationService {
verticalRulingLines,
word.getDir().getDegrees(),
word.getPageWidth(),
word.getPageHeight());
word.getPageHeight()); //
}
@ -268,11 +276,5 @@ public class BlockificationService {
return false;
}
private double round(float value, int decimalPoints) {
var d = Math.pow(10, decimalPoints);
return Math.round(value * d) / d;
}
}

2
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/classification/service/BodyTextFrameService.java
View File

@ -17,7 +17,7 @@ import com.knecon.fforesight.service.layoutparser.processor.classification.utils
@Service
public class BodyTextFrameService {
private static final float APPROXIMATE_HEADER_LINE_COUNT = 2.9f;
private static final float APPROXIMATE_HEADER_LINE_COUNT = 2.0f;
/**

7
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/classification/service/ClassificationService.java
View File

@ -57,12 +57,9 @@ public class ClassificationService {
textBlock.setClassification(PageBlockType.OTHER);
return;
}
if (PositionUtils.isOverBodyTextFrame(bodyTextFrame, textBlock, page.getRotation()) && (document.getFontSizeCounter()
.getMostPopular() == null || textBlock.getHighestFontSize() <= document.getFontSizeCounter().getMostPopular())) {
if (PositionUtils.isOverBodyTextFrame(bodyTextFrame, textBlock, page.getRotation())) {
textBlock.setClassification(PageBlockType.HEADER);
} else if (PositionUtils.isUnderBodyTextFrame(bodyTextFrame, textBlock, page.getRotation()) && (document.getFontSizeCounter()
.getMostPopular() == null || textBlock.getHighestFontSize() <= document.getFontSizeCounter().getMostPopular())) {
} else if (PositionUtils.isUnderBodyTextFrame(bodyTextFrame, textBlock, page.getRotation())) {
textBlock.setClassification(PageBlockType.FOOTER);
} else if (page.getPageNumber() == 1 && (PositionUtils.getHeightDifferenceBetweenChunkWordAndDocumentWord(textBlock,
document.getTextHeightCounter().getMostPopular()) > 2.5 && textBlock.getHighestFontSize() > document.getFontSizeCounter().getMostPopular() || page.getTextBlocks()

8
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/classification/service/TableExtractionService.java
View File

@ -136,6 +136,14 @@ public class TableExtractionService {
public List<Cell> findCells(List<Ruling> horizontalRulingLines, List<Ruling> verticalRulingLines) {
for (Ruling r : horizontalRulingLines) {
if (r.getX2() < r.getX1()) {
double a = r.getX2();
r.x2 = (float) r.getX1();
r.x1 = (float) a;
}
}
List<Cell> cellsFound = new ArrayList<>();
Map<Point2D, Ruling[]> intersectionPoints = Ruling.findIntersections(horizontalRulingLines, verticalRulingLines);
List<Point2D> intersectionPointsList = new ArrayList<>(intersectionPoints.keySet());

56
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/classification/utils/FileUtils.java
View File

@ -1,56 +0,0 @@
package com.knecon.fforesight.service.layoutparser.processor.classification.utils;
import java.io.File;
import java.io.IOException;
import java.nio.file.Files;
import lombok.experimental.UtilityClass;
import lombok.extern.slf4j.Slf4j;
@Slf4j
@UtilityClass
public class FileUtils {
public File createTempFile(String filenamePrefix, String filenameSuffix) throws IOException {
File tempFile = Files.createTempFile(filenamePrefix, filenameSuffix).toFile();
setRWPermissionsOnlyForOwner(tempFile);
return tempFile;
}
/**
* Deletes a file; logs a message with the reason if the deletion fails.
* This method is null-safe.
*
* @param file The file to delete. Can be null.
*/
public void deleteFile(File file) {
if (file != null) {
try {
Files.deleteIfExists(file.toPath());
} catch (IOException ex) {
log.warn("Could not delete file!", ex);
}
}
}
// We don't need to check the results of the permission setters below,
// since we're manipulating a file we created ourselves.
@SuppressWarnings({"ResultOfMethodCallIgnored", "squid:S899"})
private void setRWPermissionsOnlyForOwner(File tempFile) {
try {
tempFile.setReadable(true, true);
tempFile.setWritable(true, true);
tempFile.setExecutable(false);
} catch (SecurityException ex) {
// This should never happen since we're creating a temp file ourselves.
log.warn("Caught an exception during temp file creation. This should not happend. Check the code.", ex);
}
}
}

8
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/factory/SearchTextWithTextPositionDto.java
View File

@ -4,6 +4,8 @@ import java.awt.geom.Rectangle2D;
import java.util.Collections;
import java.util.List;
import com.knecon.fforesight.service.layoutparser.processor.graph.Boundary;
import lombok.AccessLevel;
import lombok.Builder;
import lombok.Getter;
@ -16,7 +18,9 @@ public class SearchTextWithTextPositionDto {
String searchText;
List<Integer> lineBreaks;
List<Integer> stringCoordsToPositionCoords;
List<Integer> stringIdxToPositionIdx;
List<Boundary> boldTextBoundaries;
List<Boundary> italicTextBoundaries;
List<Rectangle2D> positions;
@ -26,7 +30,7 @@ public class SearchTextWithTextPositionDto {
.searchText("")
.lineBreaks(Collections.emptyList())
.positions(Collections.emptyList())
.stringCoordsToPositionCoords(Collections.emptyList())
.stringIdxToPositionIdx(Collections.emptyList())
.build();
}

39
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/factory/SearchTextWithTextPositionFactory.java
View File

@ -2,6 +2,7 @@ package com.knecon.fforesight.service.layoutparser.processor.factory;
import java.awt.geom.AffineTransform;
import java.awt.geom.Rectangle2D;
import java.util.Collections;
import java.util.LinkedList;
import java.util.List;
import java.util.Objects;
@ -9,6 +10,7 @@ import java.util.Objects;
import com.knecon.fforesight.service.layoutparser.processor.classification.model.text.RedTextPosition;
import com.knecon.fforesight.service.layoutparser.processor.classification.model.text.TextDirection;
import com.knecon.fforesight.service.layoutparser.processor.classification.model.text.TextPositionSequence;
import com.knecon.fforesight.service.layoutparser.processor.graph.Boundary;
import lombok.experimental.UtilityClass;
@ -24,7 +26,7 @@ public class SearchTextWithTextPositionFactory {
public final int MAX_HYPHEN_LINEBREAK_DISTANCE = 3;
public SearchTextWithTextPositionDto buildSearchTextToTextPositionModel(List<TextPositionSequence> sequences) {
public SearchTextWithTextPositionDto buildSearchTextToTextPositionDto(List<TextPositionSequence> sequences) {
if (sequences.isEmpty() || sequences.stream().allMatch(sequence -> sequence.getTextPositions().isEmpty())) {
return SearchTextWithTextPositionDto.empty();
@ -69,8 +71,10 @@ public class SearchTextWithTextPositionFactory {
return SearchTextWithTextPositionDto.builder()
.searchText(context.stringBuilder.toString())
.lineBreaks(context.lineBreaksStringIdx)
.stringCoordsToPositionCoords(context.stringIdxToPositionIdx)
.stringIdxToPositionIdx(context.stringIdxToPositionIdx)
.positions(positions)
.boldTextBoundaries(mergeToBoundaries(context.boldTextsStringIdx))
.italicTextBoundaries(mergeToBoundaries(context.italicTextStringIdx))
.build();
}
@ -82,6 +86,8 @@ public class SearchTextWithTextPositionFactory {
// unicode characters with more than 16-bit encoding have a length > 1 in java strings
for (int j = 0; j < currentTextPosition.getUnicode().length(); j++) {
context.stringIdxToPositionIdx.add(context.positionIdx);
addTextPositionWithFontType(currentTextPosition, "bold", context.boldTextsStringIdx, context.stringIdx);
addTextPositionWithFontType(currentTextPosition, "italic", context.italicTextStringIdx, context.stringIdx);
}
context.stringIdx += currentTextPosition.getUnicode().length();
}
@ -103,6 +109,33 @@ public class SearchTextWithTextPositionFactory {
return context.stringIdx - context.lastHyphenIdx < MAX_HYPHEN_LINEBREAK_DISTANCE;
}
private static List<Boundary> mergeToBoundaries(List<Integer> integers) {
if (integers.isEmpty()) {
return Collections.emptyList();
}
List<Boundary> boundaries = new LinkedList<>();
int start = integers.get(0);
int end = integers.get(0) + 1;
for (int current : integers) {
if (current > end + 1) {
boundaries.add(new Boundary(start, end));
start = current;
}
end = current + 1;
}
if (boundaries.isEmpty())
boundaries.add(new Boundary(start, end));
return boundaries;
}
private static void addTextPositionWithFontType(RedTextPosition currentTextPosition, String fontType, List<Integer> fontTypePositions, int stringIdx) {
if (currentTextPosition.getFontName().toLowerCase().contains(fontType)) {
fontTypePositions.add(stringIdx);
}
}
private boolean isLineBreak(RedTextPosition currentTextPosition, RedTextPosition previousTextPosition) {
@ -173,6 +206,8 @@ public class SearchTextWithTextPositionFactory {
List<Integer> stringIdxToPositionIdx = new LinkedList<>();
List<Integer> lineBreaksStringIdx = new LinkedList<>();
List<Integer> boldTextsStringIdx = new LinkedList<>();
List<Integer> italicTextStringIdx = new LinkedList<>();
StringBuilder stringBuilder = new StringBuilder();
int stringIdx;

25
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/factory/TextBlockFactory.java
View File

@ -26,12 +26,33 @@ public class TextBlockFactory {
public AtomicTextBlock buildAtomicTextBlock(List<TextPositionSequence> sequences, SemanticNode parent, Integer numberOnPage, Page page) {
SearchTextWithTextPositionDto searchTextWithTextPositionDto = SearchTextWithTextPositionFactory.buildSearchTextToTextPositionModel(sequences);
SearchTextWithTextPositionDto searchTextWithTextPositionDto = SearchTextWithTextPositionFactory.buildSearchTextToTextPositionDto(sequences);
int offset = stringOffset;
stringOffset += searchTextWithTextPositionDto.getSearchText().length();
long idx = textBlockIdx;
textBlockIdx++;
return AtomicTextBlock.fromSearchTextWithTextPositionDto(searchTextWithTextPositionDto, parent, offset, idx, numberOnPage, page);
String orientation;
int textDirection;
if (sequences.isEmpty()) {
orientation = null;
textDirection = 0;
} else {
orientation = sequences.get(0).getDir().toString();
textDirection = sequences.get(0).getRotation();
}
return AtomicTextBlock.fromSearchTextWithTextPosition(searchTextWithTextPositionDto.getSearchText(),
searchTextWithTextPositionDto.getLineBreaks(),
searchTextWithTextPositionDto.getBoldTextBoundaries(),
searchTextWithTextPositionDto.getItalicTextBoundaries(),
searchTextWithTextPositionDto.getPositions(),
searchTextWithTextPositionDto.getStringIdxToPositionIdx(),
idx,
parent,
numberOnPage,
page,
offset,
orientation,
textDirection);
}

24
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/graph/Boundary.java
View File

@ -6,8 +6,6 @@ import java.util.Collection;
import java.util.LinkedList;
import java.util.List;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlock;
import lombok.EqualsAndHashCode;
import lombok.Setter;
@ -138,26 +136,4 @@ public class Boundary implements Comparable<Boundary> {
return 0;
}
/**
* shrinks the boundary, such that textBlock.subSequence(boundary) returns a string without whitespaces.
*
* @param textBlock TextBlock to check whitespaces against
* @return boundary
*/
public Boundary trim(TextBlock textBlock) {
int trimmedStart = this.start;
while (Character.isWhitespace(textBlock.charAt(trimmedStart))) {
trimmedStart++;
}
int trimmedEnd = this.end;
while (Character.isWhitespace(textBlock.charAt(trimmedEnd - 1))) {
trimmedEnd--;
}
return new Boundary(trimmedStart, Math.max(trimmedEnd, trimmedStart));
}
}

2
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/graph/DocumentTree.java
View File

@ -7,7 +7,7 @@ import java.util.LinkedList;
import java.util.List;
import java.util.stream.Stream;
import com.knecon.fforesight.service.layoutparser.internal.api.data.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Document;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.GenericSemanticNode;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.SemanticNode;

2
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/graph/nodes/Document.java
View File

@ -11,7 +11,7 @@ import java.util.stream.Collectors;
import java.util.stream.Stream;
import com.amazonaws.services.kms.model.NotFoundException;
import com.knecon.fforesight.service.layoutparser.internal.api.data.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.graph.DocumentTree;
import com.knecon.fforesight.service.layoutparser.processor.graph.entity.RedactionEntity;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlock;

2
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/graph/nodes/Footer.java
View File

@ -4,7 +4,7 @@ import java.util.HashSet;
import java.util.List;
import java.util.Set;
import com.knecon.fforesight.service.layoutparser.internal.api.data.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.graph.DocumentTree;
import com.knecon.fforesight.service.layoutparser.processor.graph.entity.RedactionEntity;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlock;

2
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/graph/nodes/Header.java
View File

@ -4,7 +4,7 @@ import java.util.HashSet;
import java.util.List;
import java.util.Set;
import com.knecon.fforesight.service.layoutparser.internal.api.data.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.graph.DocumentTree;
import com.knecon.fforesight.service.layoutparser.processor.graph.entity.RedactionEntity;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlock;

2
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/graph/nodes/Headline.java
View File

@ -4,7 +4,7 @@ import java.util.HashSet;
import java.util.List;
import java.util.Set;
import com.knecon.fforesight.service.layoutparser.internal.api.data.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.graph.DocumentTree;
import com.knecon.fforesight.service.layoutparser.processor.graph.entity.RedactionEntity;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlock;

2
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/graph/nodes/Image.java
View File

@ -8,7 +8,7 @@ import java.util.List;
import java.util.Map;
import java.util.Set;
import com.knecon.fforesight.service.layoutparser.internal.api.data.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.graph.DocumentTree;
import com.knecon.fforesight.service.layoutparser.processor.graph.entity.RedactionEntity;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlock;

2
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/graph/nodes/Paragraph.java
View File

@ -4,7 +4,7 @@ import java.util.HashSet;
import java.util.List;
import java.util.Set;
import com.knecon.fforesight.service.layoutparser.internal.api.data.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.graph.DocumentTree;
import com.knecon.fforesight.service.layoutparser.processor.graph.entity.RedactionEntity;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlock;

2
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/graph/nodes/Section.java
View File

@ -4,7 +4,7 @@ import java.util.HashSet;
import java.util.List;
import java.util.Set;
import com.knecon.fforesight.service.layoutparser.internal.api.data.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.graph.DocumentTree;
import com.knecon.fforesight.service.layoutparser.processor.graph.entity.RedactionEntity;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlock;

10
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/graph/nodes/SemanticNode.java
View File

@ -3,6 +3,7 @@ package com.knecon.fforesight.service.layoutparser.processor.graph.nodes;
import static java.lang.String.format;
import java.awt.geom.Rectangle2D;
import java.util.Comparator;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
@ -10,7 +11,7 @@ import java.util.Set;
import java.util.stream.Collectors;
import java.util.stream.Stream;
import com.knecon.fforesight.service.layoutparser.internal.api.data.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.graph.Boundary;
import com.knecon.fforesight.service.layoutparser.processor.graph.DocumentTree;
import com.knecon.fforesight.service.layoutparser.processor.graph.entity.EntityType;
@ -59,6 +60,12 @@ public interface SemanticNode {
}
default Page getFirstPage() {
return getTextBlock().getPages().stream().min(Comparator.comparingInt(Page::getNumber)).orElseThrow(() -> new IllegalStateException("SemanticNode has no Page!"));
}
/**
* Each AtomicTextBlock is assigned a page, so to get the pages for this boundary, it collects the PageNodes from each AtomicTextBlock belonging to this node's TextBlock.
*
@ -306,7 +313,6 @@ public interface SemanticNode {
}
/**
* This function is used during insertion of EntityNodes into the graph, it checks if the boundary of the RedactionEntity intersects or even contains the RedactionEntity.
* It sets the fields accordingly and recursively calls this function on all its children.

2
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/graph/nodes/Table.java
View File

@ -9,7 +9,7 @@ import java.util.Set;
import java.util.stream.IntStream;
import java.util.stream.Stream;
import com.knecon.fforesight.service.layoutparser.internal.api.data.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.graph.DocumentTree;
import com.knecon.fforesight.service.layoutparser.processor.graph.entity.RedactionEntity;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlock;

2
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/graph/nodes/TableCell.java
View File

@ -7,7 +7,7 @@ import java.util.List;
import java.util.Map;
import java.util.Set;
import com.knecon.fforesight.service.layoutparser.internal.api.data.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.graph.DocumentTree;
import com.knecon.fforesight.service.layoutparser.processor.graph.entity.RedactionEntity;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlock;

59
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/graph/textblock/AtomicTextBlock.java
View File

@ -3,6 +3,7 @@ package com.knecon.fforesight.service.layoutparser.processor.graph.textblock;
import static java.lang.String.format;
import java.awt.geom.Rectangle2D;
import java.util.ArrayList;
import java.util.Arrays;
import java.util.Collection;
import java.util.Collections;
@ -10,9 +11,8 @@ import java.util.HashMap;
import java.util.List;
import java.util.Map;
import com.knecon.fforesight.service.layoutparser.internal.api.data.AtomicPositionBlockData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.AtomicTextBlockData;
import com.knecon.fforesight.service.layoutparser.processor.factory.SearchTextWithTextPositionDto;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.AtomicPositionBlockData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.AtomicTextBlockData;
import com.knecon.fforesight.service.layoutparser.processor.graph.Boundary;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Page;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.SemanticNode;
@ -38,11 +38,20 @@ public class AtomicTextBlock implements TextBlock {
//string coordinates
Boundary boundary;
String searchText;
List<Integer> lineBreaks;
@Builder.Default
List<Integer> lineBreaks = new ArrayList<>();
@Builder.Default
List<Boundary> boldTextBoundaries = new ArrayList<>();
@Builder.Default
List<Boundary> italicTextBoundaries = new ArrayList<>();
String orientation;
int textDirection;
//position coordinates
List<Integer> stringIdxToPositionIdx;
List<Rectangle2D> positions;
@Builder.Default
List<Integer> stringIdxToPositionIdx = new ArrayList<>();
@Builder.Default
List<Rectangle2D> positions = new ArrayList<>();
@EqualsAndHashCode.Exclude
SemanticNode parent;
@ -55,23 +64,34 @@ public class AtomicTextBlock implements TextBlock {
}
public static AtomicTextBlock fromSearchTextWithTextPositionDto(SearchTextWithTextPositionDto searchTextWithTextPositionDto,
SemanticNode parent,
int stringOffset,
Long textBlockIdx,
Integer numberOnPage,
Page page) {
public static AtomicTextBlock fromSearchTextWithTextPosition(String searchText,
List<Integer> lineBreaks,
List<Boundary> boldTextBoundaries,
List<Boundary> italicTextBoundaries,
List<Rectangle2D> positions,
List<Integer> stringIdxToPositionIdx,
long idx,
SemanticNode parent,
int numberOnPage,
Page page,
int offset,
String orientation,
int textDirection) {
return AtomicTextBlock.builder()
.id(textBlockIdx)
.id(idx)
.parent(parent)
.searchText(searchTextWithTextPositionDto.getSearchText())
.searchText(searchText)
.numberOnPage(numberOnPage)
.page(page)
.lineBreaks(searchTextWithTextPositionDto.getLineBreaks())
.positions(searchTextWithTextPositionDto.getPositions())
.stringIdxToPositionIdx(searchTextWithTextPositionDto.getStringCoordsToPositionCoords())
.boundary(new Boundary(stringOffset, stringOffset + searchTextWithTextPositionDto.getSearchText().length()))
.lineBreaks(lineBreaks)
.boldTextBoundaries(boldTextBoundaries)
.italicTextBoundaries(italicTextBoundaries)
.positions(positions)
.stringIdxToPositionIdx(stringIdxToPositionIdx)
.boundary(new Boundary(offset, offset + searchText.length()))
.textDirection(textDirection)
.orientation(orientation)
.build();
}
@ -82,11 +102,8 @@ public class AtomicTextBlock implements TextBlock {
.id(textBlockIdx)
.boundary(new Boundary(stringOffset, stringOffset))
.searchText("")
.lineBreaks(Collections.emptyList())
.page(page)
.numberOnPage(numberOnPage)
.stringIdxToPositionIdx(Collections.emptyList())
.positions(Collections.emptyList())
.parent(parent)
.build();
}

35
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/graph/textblock/ConcatenatedTextBlock.java
View File

@ -3,6 +3,7 @@ package com.knecon.fforesight.service.layoutparser.processor.graph.textblock;
import static java.lang.String.format;
import java.awt.geom.Rectangle2D;
import java.util.Collection;
import java.util.Collections;
import java.util.HashMap;
import java.util.LinkedList;
@ -182,4 +183,38 @@ public class ConcatenatedTextBlock implements TextBlock {
return getSearchText();
}
@Override
public List<Boundary> getBoldTextBoundaries() {
return getAtomicTextBlocks().stream().map(AtomicTextBlock::getBoldTextBoundaries).flatMap(Collection::stream).toList();
}
@Override
public List<Boundary> getItalicTextBoundaries() {
return getAtomicTextBlocks().stream().map(AtomicTextBlock::getItalicTextBoundaries).flatMap(Collection::stream).toList();
}
@Override
public String getOrientation() {
if (atomicTextBlocks.isEmpty()) {
return "";
}
return atomicTextBlocks.get(0).getOrientation();
}
@Override
public int getTextDirection() {
if (atomicTextBlocks.isEmpty()) {
return 0;
}
return atomicTextBlocks.get(0).getTextDirection();
}
}

12
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/graph/textblock/TextBlock.java
View File

@ -21,6 +21,18 @@ public interface TextBlock extends CharSequence {
List<AtomicTextBlock> getAtomicTextBlocks();
List<Boundary> getBoldTextBoundaries();
List<Boundary> getItalicTextBoundaries();
String getOrientation();
int getTextDirection();
Boundary getBoundary();

12
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/mapper/DocumentDataMapper.java → layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/mapper/redaction/DocumentDataMapper.java
View File

@ -1,15 +1,15 @@
package com.knecon.fforesight.service.layoutparser.processor.mapper;
package com.knecon.fforesight.service.layoutparser.processor.mapper.redaction;
import java.awt.geom.Rectangle2D;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import com.knecon.fforesight.service.layoutparser.internal.api.data.AtomicPositionBlockData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.AtomicTextBlockData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.DocumentData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.PageData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.DocumentTreeData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.AtomicPositionBlockData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.AtomicTextBlockData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.PageData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentTreeData;
import com.knecon.fforesight.service.layoutparser.processor.graph.DocumentTree;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Document;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Image;

12
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/mapper/DocumentGraphMapper.java → layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/mapper/redaction/DocumentGraphMapper.java
View File

@ -1,4 +1,4 @@
package com.knecon.fforesight.service.layoutparser.processor.mapper;
package com.knecon.fforesight.service.layoutparser.processor.mapper.redaction;
import java.util.Arrays;
import java.util.HashSet;
@ -7,11 +7,11 @@ import java.util.List;
import java.util.Map;
import java.util.NoSuchElementException;
import com.knecon.fforesight.service.layoutparser.internal.api.data.AtomicPositionBlockData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.AtomicTextBlockData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.DocumentData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.DocumentTreeData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.PageData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.AtomicPositionBlockData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.AtomicTextBlockData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentTreeData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.PageData;
import com.knecon.fforesight.service.layoutparser.processor.graph.DocumentTree;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Document;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Footer;

2
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/mapper/PropertiesMapper.java → layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/mapper/redaction/PropertiesMapper.java
View File

@ -1,4 +1,4 @@
package com.knecon.fforesight.service.layoutparser.processor.mapper;
package com.knecon.fforesight.service.layoutparser.processor.mapper.redaction;
import java.awt.geom.Rectangle2D;
import java.util.Arrays;

108
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/mapper/taas/TaasDocumentDataMapper.java Normal file
View File

@ -0,0 +1,108 @@
package com.knecon.fforesight.service.layoutparser.processor.mapper.taas;
import java.awt.geom.Rectangle2D;
import java.util.List;
import java.util.Locale;
import java.util.concurrent.atomic.AtomicInteger;
import java.util.stream.IntStream;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.data.taas.ParagraphData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.taas.Range;
import com.knecon.fforesight.service.layoutparser.internal.api.data.taas.ResearchDocumentData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.taas.RowData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.taas.StructureObject;
import com.knecon.fforesight.service.layoutparser.internal.api.data.taas.TableData;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Document;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Page;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.SemanticNode;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Table;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.TableCell;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlock;
public class TaasDocumentDataMapper {
public static ResearchDocumentData fromDocument(Document document) {
AtomicInteger structureObjectNumber = new AtomicInteger();
List<StructureObject> structureObjects = document.streamAllSubNodes()
.filter(node -> !node.getType().equals(NodeType.TABLE_CELL))
.filter(node -> !node.getType().equals(NodeType.SECTION))
.map(node -> {
if (node.getType().equals(NodeType.TABLE)) {
return TaasDocumentDataMapper.fromTableWithTableData((Table) node, structureObjectNumber.getAndIncrement());
} else {
return TaasDocumentDataMapper.fromSemanticNodeWithParagraphData(node, structureObjectNumber.getAndIncrement());
}
})
.toList();
return ResearchDocumentData.builder().structureObjects(structureObjects).build();
}
public static ParagraphData fromTextBlock(String classification, TextBlock textBlock) {
return ParagraphData.builder()
.boldTextBoundaries(textBlock.getBoldTextBoundaries().stream().map(b -> new Range(b.start(), b.end())).toList())
.italicTextBoundaries(textBlock.getItalicTextBoundaries().stream().map(b -> new Range(b.start(), b.end())).toList())
.text(textBlock.getSearchText())
.classification(classification)
.orientation(textBlock.getOrientation())
.textDirection(textBlock.getTextDirection())
.build();
}
public static TableData fromTable(Table table) {
List<RowData> rowData = IntStream.range(0, table.getNumberOfRows()).boxed().map(rowIdx -> table.streamRow(rowIdx).toList()).map(TaasDocumentDataMapper::fromTableCells).toList();
return new TableData(rowData, table.getNumberOfCols(), table.getNumberOfRows());
}
public static RowData fromTableCells(List<TableCell> tableCells) {
if (tableCells.isEmpty()) {
throw new IllegalArgumentException("no table cells provided");
}
boolean header = tableCells.stream().allMatch(TableCell::isHeader);
Page firstPage = tableCells.get(0).getFirstPage();
Rectangle2D bBox = tableCells.stream().map(TableCell::getBBox).reduce((map1, map2) -> {
map1.forEach((page, rectangle) -> map2.merge(page, rectangle, (rect1, rect2) -> rect1.createUnion(rect2).getBounds2D()));
return map2;
}).orElseThrow().get(firstPage);
List<String> cellText = tableCells.stream().map(TableCell::getTextBlock).map(TextBlock::getSearchText).toList();
return new RowData(header, cellText, toFloatArray(bBox));
}
public static StructureObject fromSemanticNodeWithParagraphData(SemanticNode semanticNode, Integer structureObjectNumber) {
Page page = semanticNode.getFirstPage();
Rectangle2D bBox = semanticNode.getBBox().get(page);
return StructureObject.builder()
.structureObjectNumber(structureObjectNumber)
.boundingBox(toFloatArray(bBox))
.stringOffset(semanticNode.getBoundary().start())
.page(page.getNumber())
.paragraph(TaasDocumentDataMapper.fromTextBlock(semanticNode.getType().toString().toLowerCase(Locale.ROOT), semanticNode.getTextBlock()))
.table(null)
.build();
}
public static StructureObject fromTableWithTableData(Table table, int structureObjectNumber) {
Page page = table.getFirstPage();
Rectangle2D bBox = table.getBBox().get(page);
return StructureObject.builder()
.structureObjectNumber(structureObjectNumber)
.boundingBox(toFloatArray(bBox))
.stringOffset(table.getBoundary().start())
.page(page.getNumber())
.paragraph(null)
.table(TaasDocumentDataMapper.fromTable(table))
.build();
}
private static float[] toFloatArray(Rectangle2D bBox) {
return new float[]{(float) bBox.getX(), (float) bBox.getY(), (float) bBox.getWidth(), (float) bBox.getHeight()};
}
}

2
layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/utils/PdfVisualisationUtility.java
View File

@ -13,7 +13,7 @@ import org.apache.pdfbox.pdmodel.PDPageContentStream;
import org.apache.pdfbox.pdmodel.font.PDType1Font;
import org.apache.pdfbox.pdmodel.font.Standard14Fonts;
import com.knecon.fforesight.service.layoutparser.internal.api.data.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.graph.DocumentTree;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Document;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Page;

103
layoutparser-service/layoutparser-service-server/src/test/java/com/knecon/fforesight/service/layoutparser/server/BdrJsonBuildTest.java Normal file
View File

@ -0,0 +1,103 @@
package com.knecon.fforesight.service.layoutparser.server;
import java.awt.Color;
import java.io.File;
import java.io.FileInputStream;
import java.io.FileOutputStream;
import java.io.IOException;
import java.io.InputStream;
import java.nio.file.Files;
import java.nio.file.Path;
import java.nio.file.Paths;
import java.time.Duration;
import java.util.List;
import org.apache.pdfbox.Loader;
import org.apache.pdfbox.pdmodel.PDDocument;
import org.junit.jupiter.api.Test;
import org.springframework.beans.factory.annotation.Autowired;
import com.fasterxml.jackson.databind.ObjectMapper;
import com.knecon.fforesight.service.layoutparser.internal.api.data.taas.ResearchDocumentData;
import com.knecon.fforesight.service.layoutparser.processor.LayoutParsingService;
import com.knecon.fforesight.service.layoutparser.processor.adapter.model.image.ImageServiceResponse;
import com.knecon.fforesight.service.layoutparser.processor.adapter.model.table.TableServiceResponse;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Document;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlock;
import com.knecon.fforesight.service.layoutparser.processor.mapper.taas.TaasDocumentDataMapper;
import com.knecon.fforesight.service.layoutparser.server.utils.BaseTest;
import com.knecon.fforesight.service.layoutparser.server.utils.visualizations.PdfDraw;
import lombok.SneakyThrows;
public class BdrJsonBuildTest extends BaseTest {
@Autowired
private ObjectMapper objectMapper;
@Autowired
private LayoutParsingService layoutParsingService;
@SneakyThrows
protected Document buildGraph(File filename) {
try (InputStream inputStream = new FileInputStream(filename)) {
PDDocument pdDocument = Loader.loadPDF(inputStream);
return layoutParsingService.parseLayoutWithTimer(pdDocument, new ImageServiceResponse(), new TableServiceResponse());
}
}
@Test
public void writeBDRDocumentData() throws IOException {
String sourcePath = "/tmp/bdr_files";
String targetPath = "/tmp/result";
Paths.get(targetPath).toFile().mkdirs();
List<File> files = Files.walk(Paths.get(sourcePath)).filter(currentPath -> currentPath.toString().endsWith(".pdf")).map(Path::toFile).toList();
System.out.printf("Found %d files \n", files.size());
for (int i = 0; i < files.size(); i++) {
System.out.println(i + ": " + files.get(i));
}
System.out.println();
for (var file : files) {
long start = System.currentTimeMillis();
System.out.println("Starting Structure Analysis for: " + file);
Document document = buildGraph(file);
long start2 = System.currentTimeMillis();
ResearchDocumentData researchDocumentData = TaasDocumentDataMapper.fromDocument(document);
researchDocumentData.setOriginalFile(file.toString());
System.out.printf(", mapped to research data %d ms \n", System.currentTimeMillis() - start2);
File jsonFile = Paths.get(targetPath, file.getName().replace(".pdf", ".json")).toFile();
try (FileOutputStream fileOutputStream = new FileOutputStream(jsonFile)) {
System.out.println("json written to: " + jsonFile);
fileOutputStream.write(objectMapper.writeValueAsBytes(researchDocumentData));
}
File visualizationFile = Paths.get(targetPath, file.getName().replace(".pdf", "_BBOX.pdf")).toFile();
visualizeSemanticNodes(file, visualizationFile, document, document.getTextBlock());
System.out.println("visualization pdf written to: " + visualizationFile);
System.out.printf("Full analysis and file creation took %s\n\n", Duration.ofMillis(System.currentTimeMillis() - start));
}
}
private static void visualizeSemanticNodes(File file, File resultingFileName, Document document, TextBlock textBlock) throws IOException {
try (var fileStream = new FileInputStream(file); var outputStream = new FileOutputStream(resultingFileName)) {
PDDocument pdDocument = Loader.loadPDF(fileStream);
PdfDraw.drawDocumentGraph(pdDocument, document);
PdfDraw.drawTextBlock(pdDocument, textBlock, PdfDraw.Options.builder().stroke(true).strokeWidth(0.1f).strokeColor(Color.YELLOW).build());
pdDocument.save(outputStream);
pdDocument.close();
}
}
}

4
layoutparser-service/layoutparser-service-server/src/test/java/com/knecon/fforesight/service/layoutparser/server/graph/DocumentGraphJsonWritingTest.java
View File

@ -8,9 +8,9 @@ import org.junit.jupiter.api.Test;
import com.fasterxml.jackson.databind.ObjectMapper;
import com.iqser.red.commons.jackson.ObjectMapperFactory;
import com.knecon.fforesight.service.layoutparser.internal.api.data.DocumentData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentData;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Document;
import com.knecon.fforesight.service.layoutparser.processor.mapper.DocumentDataMapper;
import com.knecon.fforesight.service.layoutparser.processor.mapper.redaction.DocumentDataMapper;
import lombok.SneakyThrows;

6
layoutparser-service/layoutparser-service-server/src/test/java/com/knecon/fforesight/service/layoutparser/server/graph/DocumentGraphMappingTest.java
View File

@ -2,11 +2,11 @@ package com.knecon.fforesight.service.layoutparser.server.graph;
import org.junit.jupiter.api.Test;
import com.knecon.fforesight.service.layoutparser.internal.api.data.DocumentData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentData;
import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingRequest;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Document;
import com.knecon.fforesight.service.layoutparser.processor.mapper.DocumentDataMapper;
import com.knecon.fforesight.service.layoutparser.processor.mapper.DocumentGraphMapper;
import com.knecon.fforesight.service.layoutparser.processor.mapper.redaction.DocumentDataMapper;
import com.knecon.fforesight.service.layoutparser.processor.mapper.redaction.DocumentGraphMapper;
import lombok.SneakyThrows;

2
layoutparser-service/layoutparser-service-server/src/test/java/com/knecon/fforesight/service/layoutparser/server/utils/visualizations/PdfDraw.java
View File

@ -14,7 +14,7 @@ import org.apache.pdfbox.pdmodel.font.PDType1Font;
import org.apache.pdfbox.pdmodel.font.Standard14Fonts;
import org.apache.pdfbox.util.Matrix;
import com.knecon.fforesight.service.layoutparser.internal.api.data.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.graph.DocumentTree;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Document;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Page;