Marmelator/redaction-service
1
0
Fork 0
Code Issues Pull Requests 14 Actions Packages Projects Releases 622 Wiki Activity

RED-6009: Document Tree Structure

Browse Source 
* added method getNodeType() to SemanticNodes
* moved DocumentDataMapper to constructors
* moved a few other methods to named constructors
* other minor refactors
This commit is contained in:
Kilian Schuettler 2023-05-17 12:50:13 +02:00
parent 4ff73100af
commit 1dd55d9af6
33 changed files with 416 additions and 306 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/classification/model/AbstractPageBlock.java
View File

@ -33,6 +33,12 @@ public abstract class AbstractPageBlock {
public abstract String getText();
public boolean isHeadline() {
return this instanceof ClassificationTextBlock && this.getClassification() != null && this.getClassification().startsWith("H");
}
public boolean containsBlock(ClassificationTextBlock other) {
return this.minX <= other.getPdfMinX() && this.maxX >= other.getPdfMaxX() && this.minY >= other.getPdfMinY() && this.maxY <= other.getPdfMaxY();

8
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/classification/model/text/SearchableText.java
View File

@ -36,10 +36,14 @@ public class SearchableText {
StringBuilder sb = new StringBuilder();
for (TextPositionSequence word : sequences) {
sb.append(word.toString());
sb.append(word);
sb.append(' ');
}
return TextNormalizationUtilities.removeHyphenLineBreaks(sb.toString()).replaceAll("\n", " ").replaceAll(" {2}", " ");
String text = sb.toString();
text = TextNormalizationUtilities.removeHyphenLineBreaks(text);
text = TextNormalizationUtilities.removeLineBreaks(text);
text = TextNormalizationUtilities.removeRepeatingWhitespaces(text);
return text;
}
}

12
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/data/AtomicPositionBlockData.java
View File

@ -27,21 +27,11 @@ public class AtomicPositionBlockData {
return AtomicPositionBlockData.builder()
.id(atomicTextBlock.getId())
.positions(toPrimitiveFloatMatrix(atomicTextBlock.getPositions()))
.stringIdxToPositionIdx(toPrimitiveIntArray(atomicTextBlock.getStringIdxToPositionIdx()))
.stringIdxToPositionIdx(atomicTextBlock.getStringIdxToPositionIdx().stream().mapToInt(Integer::intValue).toArray())
.build();
}
private static int[] toPrimitiveIntArray(List<Integer> list) {
int[] array = new int[list.size()];
for (int i = 0; i < list.size(); i++) {
array[i] = list.get(i);
}
return array;
}
private static float[][] toPrimitiveFloatMatrix(List<Rectangle2D> positions) {
float[][] positionMatrix = new float[positions.size()][];

14
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/data/AtomicTextBlockData.java
View File

@ -1,7 +1,5 @@
package com.iqser.red.service.redaction.v1.server.layoutparsing.document.data;
import java.util.List;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.textblock.AtomicTextBlock;
import lombok.AccessLevel;
@ -34,18 +32,8 @@ public class AtomicTextBlockData {
.numberOnPage(atomicTextBlock.getNumberOnPage())
.start(atomicTextBlock.getBoundary().start())
.end(atomicTextBlock.getBoundary().end())
.lineBreaks(toPrimitiveIntArray(atomicTextBlock.getLineBreaks()))
.lineBreaks(atomicTextBlock.getLineBreaks().stream().mapToInt(Integer::intValue).toArray())
.build();
}
private static int[] toPrimitiveIntArray(List<Integer> list) {
int[] array = new int[list.size()];
for (int i = 0; i < list.size(); i++) {
array[i] = list.get(i);
}
return array;
}
}

23
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/data/DocumentData.java
View File

@ -1,5 +1,7 @@
package com.iqser.red.service.redaction.v1.server.layoutparsing.document.data;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.nodes.Document;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
@ -17,4 +19,25 @@ public class DocumentData {
AtomicPositionBlockData[] atomicPositionBlocks;
DocumentTreeData documentTreeData;
public static DocumentData fromDocument(Document document) {
var atomicPositionBlocks = document.streamTerminalTextBlocksInOrder()
.flatMap(textBlock -> textBlock.getAtomicTextBlocks().stream())
.distinct()
.map(AtomicPositionBlockData::fromAtomicTextBlock)
.toArray(AtomicPositionBlockData[]::new);
var atomicTextBlocks = document.streamTerminalTextBlocksInOrder()
.flatMap(textBlock -> textBlock.getAtomicTextBlocks().stream())
.distinct()
.map(AtomicTextBlockData::fromAtomicTextBlock)
.toArray(AtomicTextBlockData[]::new);
var pages = document.getPages().stream().map(PageData::fromPage).toArray(PageData[]::new);
var documentTreeData = new DocumentTreeData(DocumentTreeData.EntryData.fromEntry(document.getDocumentTree().getRoot()));
return new DocumentData(pages, atomicTextBlocks, atomicPositionBlocks, documentTreeData);
}
}

43
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/data/DocumentTreeData.java
View File

@ -1,10 +1,19 @@
package com.iqser.red.service.redaction.v1.server.layoutparsing.document.data;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import java.util.stream.Stream;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.data.mapper.PropertiesMapper;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.DocumentTree;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.nodes.Image;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.nodes.NodeType;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.nodes.Page;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.nodes.Section;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.nodes.Table;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.nodes.TableCell;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.textblock.AtomicTextBlock;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
@ -63,12 +72,40 @@ public class DocumentTreeData {
NodeType type;
int[] treeId;
Long[] atomicBlocks;
Long[] pages;
Long[] atomicBlockIds;
Long[] pageNumbers;
Map<String, String> properties;
List<EntryData> subEntries;
public static EntryData fromEntry(DocumentTree.Entry entry) {
Long[] atomicBlockIds = toAtomicTextBlockIds(entry);
Map<String, String> properties = switch (entry.getType()) {
case TABLE -> PropertiesMapper.buildTableProperties((Table) entry.getNode());
case TABLE_CELL -> PropertiesMapper.buildTableCellProperties((TableCell) entry.getNode());
case IMAGE -> PropertiesMapper.buildImageProperties((Image) entry.getNode());
case SECTION -> PropertiesMapper.buildSectionProperties((Section) entry.getNode());
default -> new HashMap<>();
};
var treeId = entry.getTreeId().stream().mapToInt(Integer::intValue).toArray();
var pageNumbers = entry.getNode().getPages().stream().map(Page::getNumber).map(Integer::longValue).toArray(Long[]::new);
var subEntries = entry.getChildren().stream().map(EntryData::fromEntry).toList();
return new EntryData(entry.getType(), treeId, atomicBlockIds, pageNumbers, properties, subEntries);
}
private static Long[] toAtomicTextBlockIds(DocumentTree.Entry entry) {
if (entry.getNode().isLeaf()) {
return entry.getNode().getLeafTextBlock().getAtomicTextBlocks().stream().map(AtomicTextBlock::getId).toArray(Long[]::new);
} else {
return new Long[]{};
}
}
@Override
public String toString() {
@ -83,7 +120,7 @@ public class DocumentTreeData {
sb.append(type);
sb.append(" atbs = ");
sb.append(atomicBlocks.length);
sb.append(atomicBlockIds.length);
return sb.toString();
}

8
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/data/PageData.java
View File

@ -1,5 +1,7 @@
package com.iqser.red.service.redaction.v1.server.layoutparsing.document.data;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.nodes.Page;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
@ -17,4 +19,10 @@ public class PageData {
int width;
int rotation;
public static PageData fromPage(Page page) {
return new PageData(page.getNumber(), page.getHeight(), page.getWidth(), page.getRotation());
}
}

108
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/data/mapper/DocumentDataMapper.java
View File

@ -1,108 +0,0 @@
package com.iqser.red.service.redaction.v1.server.layoutparsing.document.data.mapper;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.data.AtomicPositionBlockData;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.data.AtomicTextBlockData;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.data.DocumentData;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.data.DocumentTreeData;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.data.PageData;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.DocumentTree;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.nodes.Document;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.nodes.Image;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.nodes.Page;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.nodes.Section;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.nodes.Table;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.nodes.TableCell;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.textblock.AtomicTextBlock;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.textblock.TextBlock;
import lombok.experimental.UtilityClass;
@UtilityClass
public class DocumentDataMapper {
public DocumentData toDocumentData(Document document) {
List<AtomicTextBlockData> atomicTextBlockData = document.streamTerminalTextBlocksInOrder()
.flatMap(textBlock -> textBlock.getAtomicTextBlocks().stream())
.distinct()
.map(AtomicTextBlockData::fromAtomicTextBlock)
.toList();
List<AtomicPositionBlockData> atomicPositionBlockData = document.streamTerminalTextBlocksInOrder()
.flatMap(textBlock -> textBlock.getAtomicTextBlocks().stream())
.distinct()
.map(AtomicPositionBlockData::fromAtomicTextBlock)
.toList();
List<PageData> pageData = document.getPages().stream().map(DocumentDataMapper::toPageData).toList();
DocumentTreeData documentTreeData = toTableOfContentsData(document.getDocumentTree());
return DocumentData.builder()
.atomicTextBlocks(atomicTextBlockData.toArray(new AtomicTextBlockData[0]))
.atomicPositionBlocks(atomicPositionBlockData.toArray(new AtomicPositionBlockData[0]))
.pages(pageData.toArray(new PageData[0]))
.documentTreeData(documentTreeData)
.build();
}
private DocumentTreeData toTableOfContentsData(DocumentTree documentTree) {
return new DocumentTreeData(toEntryData(documentTree.getRoot()));
}
private DocumentTreeData.EntryData toEntryData(DocumentTree.Entry entry) {
Long[] atomicTextBlocks;
if (entry.getNode().isLeaf()) {
atomicTextBlocks = toAtomicTextBlockIds(entry.getNode().getLeafTextBlock());
} else {
atomicTextBlocks = new Long[]{};
}
Map<String, String> properties = switch (entry.getType()) {
case TABLE -> PropertiesMapper.buildTableProperties((Table) entry.getNode());
case TABLE_CELL -> PropertiesMapper.buildTableCellProperties((TableCell) entry.getNode());
case IMAGE -> PropertiesMapper.buildImageProperties((Image) entry.getNode());
case SECTION -> PropertiesMapper.buildSectionProperties((Section) entry.getNode());
default -> new HashMap<>();
};
return DocumentTreeData.EntryData.builder()
.treeId(toPrimitiveIntArray(entry.getTreeId()))
.subEntries(entry.getChildren().stream().map(DocumentDataMapper::toEntryData).toList())
.type(entry.getType())
.atomicBlocks(atomicTextBlocks)
.pages(entry.getNode().getPages().stream().map(Page::getNumber).map(Integer::longValue).toArray(Long[]::new))
.properties(properties)
.build();
}
private Long[] toAtomicTextBlockIds(TextBlock textBlock) {
return textBlock.getAtomicTextBlocks().stream().map(AtomicTextBlock::getId).toArray(Long[]::new);
}
private PageData toPageData(Page p) {
return PageData.builder().rotation(p.getRotation()).height(p.getHeight()).width(p.getWidth()).number(p.getNumber()).build();
}
private static int[] toPrimitiveIntArray(List<Integer> list) {
int[] array = new int[list.size()];
for (int i = 0; i < list.size(); i++) {
array[i] = list.get(i);
}
return array;
}
}

8
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/data/mapper/DocumentGraphMapper.java
View File

@ -65,7 +65,7 @@ public class DocumentGraphMapper {
for (DocumentTreeData.EntryData entryData : entries) {
boolean terminal = isTerminal(entryData);
List<Page> pages = Arrays.stream(entryData.getPages()).map(pageNumber -> getPage(pageNumber, context)).toList();
List<Page> pages = Arrays.stream(entryData.getPageNumbers()).map(pageNumber -> getPage(pageNumber, context)).toList();
SemanticNode node = switch (entryData.getType()) {
case SECTION -> buildSection(context, entryData.getProperties());
@ -75,12 +75,12 @@ public class DocumentGraphMapper {
case FOOTER -> buildFooter(context, terminal);
case TABLE -> buildTable(context, entryData.getProperties());
case TABLE_CELL -> buildTableCell(context, entryData.getProperties(), terminal);
case IMAGE -> buildImage(context, entryData.getProperties(), entryData.getPages());
case IMAGE -> buildImage(context, entryData.getProperties(), entryData.getPageNumbers());
default -> throw new UnsupportedOperationException("Not yet implemented for type " + entryData.getType());
};
if (node.isLeaf()) {
TextBlock textBlock = toTextBlock(entryData.getAtomicBlocks(), context, node);
TextBlock textBlock = toTextBlock(entryData.getAtomicBlockIds(), context, node);
node.setLeafTextBlock(textBlock);
}
List<Integer> treeId = Arrays.stream(entryData.getTreeId()).boxed().toList();
@ -107,7 +107,7 @@ public class DocumentGraphMapper {
private static boolean isTerminal(DocumentTreeData.EntryData entryData) {
return entryData.getAtomicBlocks().length > 0;
return entryData.getAtomicBlockIds().length > 0;
}

79
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/factory/DocumentGraphFactory.java
View File

@ -11,7 +11,6 @@ import java.util.List;
import java.util.Map;
import java.util.NoSuchElementException;
import java.util.Set;
import java.util.concurrent.atomic.AtomicInteger;
import com.iqser.red.service.redaction.v1.server.layoutparsing.classification.model.AbstractPageBlock;
import com.iqser.red.service.redaction.v1.server.layoutparsing.classification.model.ClassificationDocument;
@ -36,6 +35,7 @@ import com.iqser.red.service.redaction.v1.server.layoutparsing.document.utils.Te
import com.iqser.red.service.redaction.v1.server.redaction.utils.IdBuilder;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Getter;
import lombok.experimental.FieldDefaults;
@ -46,11 +46,10 @@ public class DocumentGraphFactory {
public Document buildDocumentGraph(ClassificationDocument document) {
TextBlockFactory textBlockFactory = new TextBlockFactory();
Document documentGraph = new Document();
Context context = new Context(new DocumentTree(documentGraph), new HashMap<>(), new LinkedList<>(), new LinkedList<>(), textBlockFactory);
Context context = new Context(documentGraph);
document.getPages().stream().map(DocumentGraphFactory::buildPage).forEach(page -> context.getPages().put(page, new AtomicInteger(1)));
document.getPages().forEach(context::buildAndAddPageWithCounter);
document.getSections().stream().flatMap(section -> section.getImages().stream()).forEach(image -> context.getImages().add(image));
addSections(document, context);
addHeaderAndFooterToEachPage(document, context);
@ -69,12 +68,6 @@ public class DocumentGraphFactory {
}
public static boolean pageBlockIsHeadline(AbstractPageBlock abstractPageBlock) {
return abstractPageBlock instanceof ClassificationTextBlock && abstractPageBlock.getClassification() != null && abstractPageBlock.getClassification().startsWith("H");
}
public void addParagraphOrHeadline(GenericSemanticNode parentNode,
ClassificationTextBlock originalTextBlock,
Context context,
@ -83,7 +76,7 @@ public class DocumentGraphFactory {
Page page = context.getPage(originalTextBlock.getPage());
GenericSemanticNode node;
if (pageBlockIsHeadline(originalTextBlock)) {
if (originalTextBlock.isHeadline()) {
node = Headline.builder().documentTree(context.getDocumentTree()).build();
} else {
node = Paragraph.builder().documentTree(context.getDocumentTree()).build();
@ -94,17 +87,9 @@ public class DocumentGraphFactory {
List<ClassificationTextBlock> textBlocks = new LinkedList<>(textBlocksToMerge);
textBlocks.add(originalTextBlock);
AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock(TextPositionOperations.mergeAndSortTextPositionSequenceByYThenX(textBlocks), node, context, page);
if (node instanceof Headline headline) {
List<Integer> tocId = context.documentTree.createNewChildEntryAndReturnId(parentNode, NodeType.HEADLINE, node);
headline.setLeafTextBlock(textBlock);
headline.setTreeId(tocId);
}
if (node instanceof Paragraph paragraph) {
List<Integer> tocId = context.documentTree.createNewChildEntryAndReturnId(parentNode, NodeType.PARAGRAPH, node);
paragraph.setLeafTextBlock(textBlock);
paragraph.setTreeId(tocId);
}
List<Integer> treeId = context.documentTree.createNewChildEntryAndReturnId(parentNode, node.getType(), node);
node.setLeafTextBlock(textBlock);
node.setTreeId(treeId);
}
@ -178,11 +163,7 @@ public class DocumentGraphFactory {
Page page = context.getPage(textBlocks.get(0).getPage());
Header header = Header.builder().documentTree(context.getDocumentTree()).build();
AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock(TextPositionOperations.mergeAndSortTextPositionSequenceByYThenX(textBlocks),
header,
context,
0,
page);
AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock(TextPositionOperations.mergeAndSortTextPositionSequenceByYThenX(textBlocks), header, 0, page);
List<Integer> tocId = context.getDocumentTree().createNewMainEntryAndReturnId(NodeType.HEADER, header);
header.setTreeId(tocId);
header.setLeafTextBlock(textBlock);
@ -214,30 +195,46 @@ public class DocumentGraphFactory {
}
private Page buildPage(ClassificationPage p) {
return Page.builder()
.height((int) p.getPageHeight())
.width((int) p.getPageWidth())
.number(p.getPageNumber())
.rotation(p.getRotation())
.mainBody(new LinkedList<>())
.build();
}
@Getter
@Builder
@AllArgsConstructor
@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
static final class Context {
public static final class Context {
DocumentTree documentTree;
Map<Page, AtomicInteger> pages;
Map<Page, Integer> pages;
List<Section> sections;
List<ClassifiedImage> images;
TextBlockFactory textBlockFactory;
public Context(Document document) {
documentTree = new DocumentTree(document);
pages = new HashMap<>();
sections = new LinkedList<>();
images = new LinkedList<>();
textBlockFactory = new TextBlockFactory();
}
public void buildAndAddPageWithCounter(ClassificationPage classificationPage) {
Page page = Page.fromClassificationPage(classificationPage);
//this counter counts the TextBlocks per page
//initial value is set to 1, because 0 is reserved for Header
pages.put(page, 1);
}
public int getAndIncrementTextBlockNumberOnPage(Page page) {
Integer textBlockNumberOnPage = pages.get(page);
pages.merge(page, 1, Integer::sum);
return textBlockNumberOnPage;
}
public Page getPage(int pageIndex) {
return pages.keySet()

12
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/factory/SearchTextWithTextPositionDto.java
View File

@ -1,6 +1,7 @@
package com.iqser.red.service.redaction.v1.server.layoutparsing.document.factory;
import java.awt.geom.Rectangle2D;
import java.util.Collections;
import java.util.List;
import lombok.AccessLevel;
@ -18,4 +19,15 @@ public class SearchTextWithTextPositionDto {
List<Integer> stringCoordsToPositionCoords;
List<Rectangle2D> positions;
public static SearchTextWithTextPositionDto empty() {
return SearchTextWithTextPositionDto.builder()
.searchText("")
.lineBreaks(Collections.emptyList())
.positions(Collections.emptyList())
.stringCoordsToPositionCoords(Collections.emptyList())
.build();
}
}

105
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/factory/SearchTextWithTextPositionFactory.java
View File

@ -2,7 +2,6 @@ package com.iqser.red.service.redaction.v1.server.layoutparsing.document.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;
@ -19,21 +18,10 @@ public class SearchTextWithTextPositionFactory {
public static SearchTextWithTextPositionDto buildSearchTextToTextPositionModel(List<TextPositionSequence> sequences) {
if (sequences.isEmpty() || sequences.stream().allMatch(sequence -> sequence.getTextPositions().isEmpty())) {
return SearchTextWithTextPositionDto.builder()
.searchText("")
.lineBreaks(Collections.emptyList())
.positions(Collections.emptyList())
.stringCoordsToPositionCoords(Collections.emptyList())
.build();
return SearchTextWithTextPositionDto.empty();
}
List<Integer> stringIdxToPositionIdx = new LinkedList<>();
List<Integer> lineBreaksStringIdx = new LinkedList<>();
StringBuilder sb = new StringBuilder();
int stringIdx = 0;
int positionIdx = 0;
int lastHyphenIdx = -3;
Context context = new Context();
RedTextPosition currentTextPosition = sequences.get(0).getTextPositions().get(0);
RedTextPosition previousTextPosition = RedTextPosition.builder().unicode(" ").position(currentTextPosition.getPosition()).build();
@ -43,55 +31,72 @@ public class SearchTextWithTextPositionFactory {
currentTextPosition = word.getTextPositions().get(i);
if (isLineBreak(currentTextPosition, previousTextPosition)) {
if (stringIdx - lastHyphenIdx < 3) {
sb.delete(lastHyphenIdx, sb.length());
stringIdxToPositionIdx = stringIdxToPositionIdx.subList(0, lastHyphenIdx);
stringIdx = lastHyphenIdx;
lastHyphenIdx = -3;
}
lineBreaksStringIdx.add(stringIdx);
}
removeHyphenLinebreaks(currentTextPosition, previousTextPosition, context);
if (!isRepeatedWhitespace(currentTextPosition.getUnicode(), previousTextPosition.getUnicode())) {
if (isHyphen(currentTextPosition.getUnicode())) {
lastHyphenIdx = stringIdx;
context.lastHyphenIdx = context.stringIdx;
}
sb.append(currentTextPosition.getUnicode());
// unicode characters with more than 16 bit encoding have a length > 1 in java strings
for (int j = 0; j < currentTextPosition.getUnicode().length(); j++) {
stringIdxToPositionIdx.add(positionIdx);
}
stringIdx += currentTextPosition.getUnicode().length();
appendCurrentTextPosition(context, currentTextPosition);
}
previousTextPosition = currentTextPosition;
++positionIdx;
++context.positionIdx;
}
previousTextPosition = RedTextPosition.builder().unicode(" ").position(previousTextPosition.getPosition()).build();
sb.append(" ");
stringIdxToPositionIdx.add(positionIdx);
++stringIdx;
context.sb.append(" ");
context.stringIdxToPositionIdx.add(context.positionIdx);
++context.stringIdx;
}
assert sb.length() == stringIdxToPositionIdx.size();
assert context.sb.length() == context.stringIdxToPositionIdx.size();
List<Rectangle2D> positions = sequences.stream()
.flatMap(sequence -> sequence.getTextPositions().stream().map(textPosition -> mapRedTextPositionToInitialUserSpace(textPosition, sequence)))
.toList();
return SearchTextWithTextPositionDto.builder()
.searchText(sb.toString())
.lineBreaks(lineBreaksStringIdx)
.stringCoordsToPositionCoords(stringIdxToPositionIdx)
.searchText(context.sb.toString())
.lineBreaks(context.lineBreaksStringIdx)
.stringCoordsToPositionCoords(context.stringIdxToPositionIdx)
.positions(positions)
.build();
}
private static void appendCurrentTextPosition(Context context, RedTextPosition currentTextPosition) {
context.sb.append(currentTextPosition.getUnicode());
// 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);
}
context.stringIdx += currentTextPosition.getUnicode().length();
}
private static void removeHyphenLinebreaks(RedTextPosition currentTextPosition, RedTextPosition previousTextPosition, Context context) {
if (isLineBreak(currentTextPosition, previousTextPosition)) {
if (lastHyphenDirectlyBeforeLineBreak(context)) {
context.sb.delete(context.lastHyphenIdx, context.sb.length());
context.stringIdxToPositionIdx = context.stringIdxToPositionIdx.subList(0, context.lastHyphenIdx);
context.stringIdx = context.lastHyphenIdx;
context.lastHyphenIdx = -3;
}
context.lineBreaksStringIdx.add(context.stringIdx);
}
}
private static boolean lastHyphenDirectlyBeforeLineBreak(Context context) {
return context.stringIdx - context.lastHyphenIdx < 3;
}
private static boolean isLineBreak(RedTextPosition currentTextPosition, RedTextPosition previousTextPosition) {
return Objects.equals(currentTextPosition.getUnicode(), "\n") || isDeltaYLargerThanTextHeight(currentTextPosition, previousTextPosition);
@ -156,4 +161,22 @@ public class SearchTextWithTextPositionFactory {
return transform.createTransformedShape(rectangle2D).getBounds2D();
}
private static class Context {
List<Integer> stringIdxToPositionIdx = new LinkedList<>();
List<Integer> lineBreaksStringIdx = new LinkedList<>();
StringBuilder sb = new StringBuilder();
int stringIdx;
int positionIdx;
// when checking for a hyphen linebreak, we need to check after a linebreak if the last hyphen was less than three symbols away.
// We detect a linebreak as either a "\n" character or if two adjacent symbol's position differ in y-coordinates by at least one character height.
// If there is a hyphen linebreak, the hyphen will be 1 position in front of a "\n" or 2 positions in front of the character which has a lower y-coordinate
// This is why, we need to initialize this to <= -3, otherwise, if the very first symbol is a \n we would detect a hyphen linebreak that isn't there.
// Also, Integer.MIN_VALUE is a bad idea due to potential overflow during arithmetic operations. This is why the default should be -3.
int lastHyphenIdx = -3;
}
}

47
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/factory/SectionNodeFactory.java
View File

@ -1,6 +1,5 @@
package com.iqser.red.service.redaction.v1.server.layoutparsing.document.factory;
import static com.iqser.red.service.redaction.v1.server.layoutparsing.document.factory.DocumentGraphFactory.pageBlockIsHeadline;
import static java.util.Collections.emptyList;
import static java.util.stream.Collectors.groupingBy;
@ -36,19 +35,13 @@ public class SectionNodeFactory {
context.getSections().add(section);
blocksPerPage.keySet().forEach(pageNumber -> addSectionNodeToPageNode(context, section, pageNumber));
List<Integer> tocId;
if (parentNode == null) {
tocId = context.getDocumentTree().createNewMainEntryAndReturnId(NodeType.SECTION, section);
} else {
tocId = context.getDocumentTree().createNewChildEntryAndReturnId(parentNode, NodeType.SECTION, section);
}
section.setTreeId(tocId);
section.setTreeId(getTreeId(parentNode, context, section));
addFirstHeadlineDirectlyToSection(pageBlocks, context, section);
if (containsTablesAndTextBlocks(pageBlocks)) {
splitPageBlocksIntoSubSections(pageBlocks).forEach(subSectionPageBlocks -> addSection(section, subSectionPageBlocks, emptyList(), context));
} else {
addTablesOrParagraphsOrHeadlinesToSection(pageBlocks, context, section);
addTablesAndParagraphsAndHeadlinesToSection(pageBlocks, context, section);
if (listIsTextBlockOnly(pageBlocks)) {
section.setExcludesTables(true);
}
@ -58,7 +51,17 @@ public class SectionNodeFactory {
}
private static boolean listIsTextBlockOnly(List<AbstractPageBlock> pageBlocks) {
private static List<Integer> getTreeId(GenericSemanticNode parentNode, DocumentGraphFactory.Context context, Section section) {
if (parentNode == null) {
return context.getDocumentTree().createNewMainEntryAndReturnId(NodeType.SECTION, section);
} else {
return context.getDocumentTree().createNewChildEntryAndReturnId(parentNode, NodeType.SECTION, section);
}
}
private boolean listIsTextBlockOnly(List<AbstractPageBlock> pageBlocks) {
return pageBlocks.stream().allMatch(abstractPageBlock -> abstractPageBlock instanceof ClassificationTextBlock);
}
@ -66,14 +69,14 @@ public class SectionNodeFactory {
private void addFirstHeadlineDirectlyToSection(List<AbstractPageBlock> pageBlocks, DocumentGraphFactory.Context context, Section section) {
if (pageBlockIsHeadline(pageBlocks.get(0))) {
addTablesOrParagraphsOrHeadlinesToSection(List.of(pageBlocks.get(0)), context, section);
if (pageBlocks.get(0).isHeadline()) {
addTablesAndParagraphsAndHeadlinesToSection(List.of(pageBlocks.get(0)), context, section);
pageBlocks.remove(0);
}
}
private void addTablesOrParagraphsOrHeadlinesToSection(List<AbstractPageBlock> pageBlocks, DocumentGraphFactory.Context context, Section section) {
private void addTablesAndParagraphsAndHeadlinesToSection(List<AbstractPageBlock> pageBlocks, DocumentGraphFactory.Context context, Section section) {
Set<AbstractPageBlock> alreadyMerged = new HashSet<>();
List<AbstractPageBlock> remainingBlocks = new LinkedList<>(pageBlocks);
@ -100,7 +103,7 @@ public class SectionNodeFactory {
}
private static boolean containsTablesAndTextBlocks(List<AbstractPageBlock> pageBlocks) {
private boolean containsTablesAndTextBlocks(List<AbstractPageBlock> pageBlocks) {
return pageBlocks.stream().anyMatch(pageBlock -> pageBlock instanceof TablePageBlock) && pageBlocks.stream()
.anyMatch(pageBlock -> pageBlock instanceof ClassificationTextBlock);
@ -109,8 +112,8 @@ public class SectionNodeFactory {
/**
* This function splits the list of PageBlocks around TablePageBlocks, such that SubSections can be created, that don't include tables.
* First, the function splits the list into coherent Lists, which include only either type.
* If a Table is directly preceded by a Headline, it is moved to the Table list.
* This is needed so we can execute rules on sections, that do not contain tables.
* See: <a href="https://knecon.atlassian.net/wiki/spaces/RED/pages/14765218/Document+Structure">document structure wiki</a>
*
* @param pageBlocks a List of AbstractPageBlocks, which have at least one TablePageBlock and one ClassificationTextBlock
* @return List of Lists of AbstractPageBlocks, which include either a single Headline ClassificationTextBlock and a TablePageBlock or only ClassificationTextBlocks.
@ -123,13 +126,13 @@ public class SectionNodeFactory {
}
private static void movePrecedingHeadlineToTableList(List<List<AbstractPageBlock>> splitList) {
private void movePrecedingHeadlineToTableList(List<List<AbstractPageBlock>> splitList) {
for (int i = 0; i < splitList.size(); i++) {
if (listIsTableOnly(splitList.get(i)) && i > 0) {
if (listIsTablesOnly(splitList.get(i)) && i > 0) {
List<AbstractPageBlock> previousList = splitList.get(i - 1);
AbstractPageBlock lastPageBlockInPreviousList = previousList.get(previousList.size() - 1);
if (pageBlockIsHeadline(lastPageBlockInPreviousList)) {
if (lastPageBlockInPreviousList.isHeadline()) {
previousList.remove(i - 1);
splitList.get(i).add(0, lastPageBlockInPreviousList);
}
@ -138,7 +141,7 @@ public class SectionNodeFactory {
}
private static boolean listIsTableOnly(List<AbstractPageBlock> abstractPageBlocks) {
private boolean listIsTablesOnly(List<AbstractPageBlock> abstractPageBlocks) {
return abstractPageBlocks.stream().allMatch(abstractPageBlock -> abstractPageBlock instanceof TablePageBlock);
}
@ -148,7 +151,7 @@ public class SectionNodeFactory {
* @param pageBlocks a List of AbstractPageBlocks, which have at least one TablePageBlock and one ClassificationTextBlock
* @return List of Lists of AbstractPageBlocks, which are exclusively of type ClassificationTextBlock or TablePageBlock
*/
private static List<List<AbstractPageBlock>> splitIntoCoherentList(List<AbstractPageBlock> pageBlocks) {
private List<List<AbstractPageBlock>> splitIntoCoherentList(List<AbstractPageBlock> pageBlocks) {
List<List<AbstractPageBlock>> splitList = new LinkedList<>();
List<AbstractPageBlock> currentList = new LinkedList<>();
@ -169,7 +172,7 @@ public class SectionNodeFactory {
}
private static List<ClassificationTextBlock> findTextBlocksWithSameClassificationAndAlignsY(AbstractPageBlock atc, List<AbstractPageBlock> pageBlocks) {
private List<ClassificationTextBlock> findTextBlocksWithSameClassificationAndAlignsY(AbstractPageBlock atc, List<AbstractPageBlock> pageBlocks) {
return pageBlocks.stream()
.filter(abstractTextContainer -> !abstractTextContainer.equals(atc))

26
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/factory/TableNodeFactory.java
View File

@ -30,7 +30,7 @@ public class TableNodeFactory {
public void addTable(GenericSemanticNode parentNode, List<TablePageBlock> tablesToMerge, DocumentGraphFactory.Context context) {
setPageNumberInCellTextBlocks(tablesToMerge);
setPageNumberInCells(tablesToMerge);
Set<Page> pages = tablesToMerge.stream().map(AbstractPageBlock::getPage).map(context::getPage).collect(Collectors.toSet());
List<List<Cell>> mergedRows = tablesToMerge.stream().map(TablePageBlock::getRows).flatMap(Collection::stream).toList();
Table table = Table.builder().documentTree(context.getDocumentTree()).numberOfCols(mergedRows.get(0).size()).numberOfRows(mergedRows.size()).build();
@ -41,22 +41,32 @@ public class TableNodeFactory {
table.setTreeId(tocId);
addTableCells(mergedRows, table, context);
IfTableHasNoHeadersAssumeFirstRowAreHeaders(table);
ifTableHasNoHeadersAssumeFirstRowAreHeaders(table);
}
private static void setPageNumberInCellTextBlocks(List<TablePageBlock> tablesToMerge) {
private void setPageNumberInCells(List<TablePageBlock> tablesToMerge) {
// For some reason I can't figure out, in some table cells, the ClassificationTextBlocks have 0 as page number
// So I am fixing this here, but this should actually be fixed upstream.
tablesToMerge.forEach(table -> table.getRows()
.stream()
.flatMap(Collection::stream)
.peek(cell -> cell.setPageNumber(table.getPage()))
.forEach(cell -> cell.getTextBlocks().stream().filter(tb -> tb.getPage() == 0).forEach(tb -> tb.setPage(table.getPage()))));
.forEach(cell -> setPageNumberInTextBlocksWithPageNumberSetTo0(table, cell)));
}
private static void setPageNumberInTextBlocksWithPageNumberSetTo0(TablePageBlock table, Cell cell) {
cell.getTextBlocks().stream()//
.filter(tb -> tb.getPage() == 0)//
.forEach(tb -> tb.setPage(table.getPage()));
}
@SuppressWarnings("PMD.UnusedPrivateMethod") // PMD actually flags this wrong
private static void addTableToPage(Page page, SemanticNode parentNode, Table table) {
private void addTableToPage(Page page, SemanticNode parentNode, Table table) {
if (!page.getMainBody().contains(parentNode)) {
parentNode.getPages().add(page);
@ -66,7 +76,7 @@ public class TableNodeFactory {
}
private static void IfTableHasNoHeadersAssumeFirstRowAreHeaders(Table table) {
private void ifTableHasNoHeadersAssumeFirstRowAreHeaders(Table table) {
if (table.streamHeaders().findAny().isEmpty()) {
table.streamRow(0).forEach(tableCellNode -> tableCellNode.setHeader(true));
@ -124,13 +134,13 @@ public class TableNodeFactory {
}
private static boolean cellAreaIsSmallerThanPageAreaTimesThreshold(Cell cell, Page page) {
private boolean cellAreaIsSmallerThanPageAreaTimesThreshold(Cell cell, Page page) {
return cell.getArea() < TABLE_CELL_MERGE_CONTENTS_SIZE_THRESHOLD * page.getHeight() * page.getWidth();
}
private static boolean firstTextBlockIsHeadline(Cell cell) {
private boolean firstTextBlockIsHeadline(Cell cell) {
String classification = cell.getTextBlocks().get(0).getClassification();
return classification != null && classification.startsWith("H");

61
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/factory/TextBlockFactory.java
View File

@ -1,74 +1,53 @@
package com.iqser.red.service.redaction.v1.server.layoutparsing.document.factory;
import java.util.Collections;
import java.util.List;
import java.util.concurrent.atomic.AtomicInteger;
import java.util.concurrent.atomic.AtomicLong;
import com.iqser.red.service.redaction.v1.server.layoutparsing.classification.model.text.TextPositionSequence;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.Boundary;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.nodes.Page;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.nodes.SemanticNode;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.textblock.AtomicTextBlock;
import lombok.AccessLevel;
import lombok.experimental.FieldDefaults;
@FieldDefaults(level = AccessLevel.PRIVATE)
public class TextBlockFactory {
AtomicInteger stringOffset;
AtomicLong textBlockIdx;
public TextBlockFactory() {
stringOffset = new AtomicInteger();
textBlockIdx = new AtomicLong();
}
int stringOffset;
long textBlockIdx;
public AtomicTextBlock buildAtomicTextBlock(List<TextPositionSequence> sequences, SemanticNode parent, DocumentGraphFactory.Context context, Page page) {
Integer numberOnPage = context.getPages().get(page).getAndIncrement();
return buildAtomicTextBlock(sequences, parent, context, numberOnPage, page);
Integer numberOnPage = context.getAndIncrementTextBlockNumberOnPage(page);
return buildAtomicTextBlock(sequences, parent, numberOnPage, page);
}
public AtomicTextBlock buildAtomicTextBlock(List<TextPositionSequence> sequences, SemanticNode parent, DocumentGraphFactory.Context context, Integer numberOnPage, Page page) {
public AtomicTextBlock buildAtomicTextBlock(List<TextPositionSequence> sequences, SemanticNode parent, Integer numberOnPage, Page page) {
SearchTextWithTextPositionDto searchTextWithTextPositionDto = SearchTextWithTextPositionFactory.buildSearchTextToTextPositionModel(sequences);
int offset = stringOffset.getAndAdd(searchTextWithTextPositionDto.getSearchText().length());
return AtomicTextBlock.builder()
.id(textBlockIdx.getAndIncrement())
.parent(parent)
.searchText(searchTextWithTextPositionDto.getSearchText())
.numberOnPage(numberOnPage)
.page(page)
.lineBreaks(searchTextWithTextPositionDto.getLineBreaks())
.positions(searchTextWithTextPositionDto.getPositions())
.stringIdxToPositionIdx(searchTextWithTextPositionDto.getStringCoordsToPositionCoords())
.boundary(new Boundary(offset, offset + searchTextWithTextPositionDto.getSearchText().length()))
.build();
int offset = stringOffset;
stringOffset += searchTextWithTextPositionDto.getSearchText().length();
long idx = textBlockIdx;
textBlockIdx++;
return AtomicTextBlock.fromSearchTextWithTextPositionDto(searchTextWithTextPositionDto, parent, offset, idx, numberOnPage, page);
}
public AtomicTextBlock emptyTextBlock(SemanticNode parent, DocumentGraphFactory.Context context, Page page) {
return emptyTextBlock(parent, context.getPages().get(page).getAndIncrement(), page);
long idx = textBlockIdx;
textBlockIdx++;
return AtomicTextBlock.empty(idx, stringOffset, page, context.getAndIncrementTextBlockNumberOnPage(page), parent);
}
public AtomicTextBlock emptyTextBlock(SemanticNode parent, Integer numberOnPage, Page page) {
return AtomicTextBlock.builder()
.id(textBlockIdx.getAndIncrement())
.boundary(new Boundary(stringOffset.get(), stringOffset.get()))
.searchText("")
.lineBreaks(Collections.emptyList())
.page(page)
.numberOnPage(numberOnPage)
.stringIdxToPositionIdx(Collections.emptyList())
.positions(Collections.emptyList())
.parent(parent)
.build();
long idx = textBlockIdx;
textBlockIdx++;
return AtomicTextBlock.empty(idx, stringOffset, page, numberOnPage, parent);
}
}

7
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/graph/nodes/Document.java
View File

@ -38,6 +38,13 @@ public class Document implements GenericSemanticNode {
Set<RedactionEntity> entities = new HashSet<>();
@Override
public NodeType getType() {
return NodeType.DOCUMENT;
}
public TextBlock buildTextBlock() {
return streamTerminalTextBlocksInOrder().collect(new TextBlockCollector());

7
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/graph/nodes/Footer.java
View File

@ -37,6 +37,13 @@ public class Footer implements GenericSemanticNode {
Set<RedactionEntity> entities = new HashSet<>();
@Override
public NodeType getType() {
return NodeType.FOOTER;
}
@Override
public TextBlock buildTextBlock() {

7
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/graph/nodes/Header.java
View File

@ -37,6 +37,13 @@ public class Header implements GenericSemanticNode {
Set<RedactionEntity> entities = new HashSet<>();
@Override
public NodeType getType() {
return NodeType.HEADER;
}
@Override
public TextBlock buildTextBlock() {

7
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/graph/nodes/Headline.java
View File

@ -37,6 +37,13 @@ public class Headline implements GenericSemanticNode {
Set<RedactionEntity> entities = new HashSet<>();
@Override
public NodeType getType() {
return NodeType.HEADLINE;
}
@Override
public TextBlock buildTextBlock() {

7
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/graph/nodes/Image.java
View File

@ -55,6 +55,13 @@ public class Image implements GenericSemanticNode {
Set<RedactionEntity> entities = new HashSet<>();
@Override
public NodeType getType() {
return NodeType.IMAGE;
}
@Override
public TextBlock buildTextBlock() {

14
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/graph/nodes/Page.java
View File

@ -1,9 +1,11 @@
package com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.nodes;
import java.util.HashSet;
import java.util.LinkedList;
import java.util.List;
import java.util.Set;
import com.iqser.red.service.redaction.v1.server.layoutparsing.classification.model.ClassificationPage;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.entity.RedactionEntity;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.textblock.TextBlock;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.textblock.TextBlockCollector;
@ -44,6 +46,18 @@ public class Page {
Set<Image> images = new HashSet<>();
public static Page fromClassificationPage(ClassificationPage classificationPage) {
return Page.builder()
.height((int) classificationPage.getPageHeight())
.width((int) classificationPage.getPageWidth())
.number(classificationPage.getPageNumber())
.rotation(classificationPage.getRotation())
.mainBody(new LinkedList<>())
.build();
}
public TextBlock getMainBodyTextBlock() {
return mainBody.stream().filter(SemanticNode::isLeaf).map(SemanticNode::getLeafTextBlock).collect(new TextBlockCollector());

7
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/graph/nodes/Paragraph.java
View File

@ -35,6 +35,13 @@ public class Paragraph implements GenericSemanticNode {
Set<RedactionEntity> entities = new HashSet<>();
@Override
public NodeType getType() {
return NodeType.PARAGRAPH;
}
@Override
public TextBlock buildTextBlock() {

7
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/graph/nodes/Section.java
View File

@ -36,6 +36,13 @@ public class Section implements GenericSemanticNode {
Set<RedactionEntity> entities = new HashSet<>();
@Override
public NodeType getType() {
return NodeType.SECTION;
}
@Override
public TextBlock buildTextBlock() {

8
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/graph/nodes/SemanticNode.java
View File

@ -19,6 +19,14 @@ import com.iqser.red.service.redaction.v1.server.layoutparsing.document.utils.Re
public interface SemanticNode {
/**
* Returns the type of this node, such as Section, Paragraph, etc.
*
* @return NodeType of this node
*/
NodeType getType();
/**
* Searches all Nodes located underneath this Node in the DocumentTree and concatenates their AtomicTextBlocks into a single TextBlockEntity.
* So, for a Section all TextBlocks of Subsections, Paragraphs, and Tables are concatenated into a single TextBlockEntity

7
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/graph/nodes/Table.java
View File

@ -271,6 +271,13 @@ public class Table implements SemanticNode {
}
@Override
public NodeType getType() {
return NodeType.TABLE;
}
@Override
public TextBlock buildTextBlock() {

7
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/graph/nodes/TableCell.java
View File

@ -55,6 +55,13 @@ public class TableCell implements GenericSemanticNode {
}
@Override
public NodeType getType() {
return NodeType.TABLE_CELL;
}
@Override
public TextBlock buildTextBlock() {

43
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/graph/textblock/AtomicTextBlock.java
View File

@ -10,9 +10,9 @@ import java.util.HashMap;
import java.util.List;
import java.util.Map;
import com.google.common.primitives.Ints;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.data.AtomicPositionBlockData;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.data.AtomicTextBlockData;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.factory.SearchTextWithTextPositionDto;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.Boundary;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.nodes.Page;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.nodes.SemanticNode;
@ -55,6 +55,43 @@ public class AtomicTextBlock implements TextBlock {
}
public static AtomicTextBlock fromSearchTextWithTextPositionDto(SearchTextWithTextPositionDto searchTextWithTextPositionDto,
SemanticNode parent,
int stringOffset,
Long textBlockIdx,
Integer numberOnPage,
Page page) {
return AtomicTextBlock.builder()
.id(textBlockIdx)
.parent(parent)
.searchText(searchTextWithTextPositionDto.getSearchText())
.numberOnPage(numberOnPage)
.page(page)
.lineBreaks(searchTextWithTextPositionDto.getLineBreaks())
.positions(searchTextWithTextPositionDto.getPositions())
.stringIdxToPositionIdx(searchTextWithTextPositionDto.getStringCoordsToPositionCoords())
.boundary(new Boundary(stringOffset, stringOffset + searchTextWithTextPositionDto.getSearchText().length()))
.build();
}
public static AtomicTextBlock empty(Long textBlockIdx, int stringOffset, Page page, int numberOnPage, SemanticNode parent) {
return AtomicTextBlock.builder()
.id(textBlockIdx)
.boundary(new Boundary(stringOffset, stringOffset))
.searchText("")
.lineBreaks(Collections.emptyList())
.page(page)
.numberOnPage(numberOnPage)
.stringIdxToPositionIdx(Collections.emptyList())
.positions(Collections.emptyList())
.parent(parent)
.build();
}
public static AtomicTextBlock fromAtomicTextBlockData(AtomicTextBlockData atomicTextBlockData,
AtomicPositionBlockData atomicPositionBlockData,
SemanticNode parent,
@ -66,8 +103,8 @@ public class AtomicTextBlock implements TextBlock {
.page(page)
.boundary(new Boundary(atomicTextBlockData.getStart(), atomicTextBlockData.getEnd()))
.searchText(atomicTextBlockData.getSearchText())
.lineBreaks(Ints.asList(atomicTextBlockData.getLineBreaks()))
.stringIdxToPositionIdx(Ints.asList(atomicPositionBlockData.getStringIdxToPositionIdx()))
.lineBreaks(Arrays.stream(atomicTextBlockData.getLineBreaks()).boxed().toList())
.stringIdxToPositionIdx(Arrays.stream(atomicPositionBlockData.getStringIdxToPositionIdx()).boxed().toList())
.positions(toRectangle2DList(atomicPositionBlockData.getPositions()))
.parent(parent)
.build();

7
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/graph/textblock/ConcatenatedTextBlock.java
View File

@ -3,6 +3,7 @@ package com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.t
import static java.lang.String.format;
import java.awt.geom.Rectangle2D;
import java.util.Collections;
import java.util.HashMap;
import java.util.LinkedList;
import java.util.List;
@ -25,6 +26,12 @@ public class ConcatenatedTextBlock implements TextBlock {
Boundary boundary;
public static ConcatenatedTextBlock empty() {
return new ConcatenatedTextBlock(Collections.emptyList());
}
public ConcatenatedTextBlock(List<AtomicTextBlock> atomicTextBlocks) {
this.atomicTextBlocks = new LinkedList<>();

3
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/graph/textblock/TextBlockCollector.java
View File

@ -1,6 +1,5 @@
package com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.textblock;
import java.util.Collections;
import java.util.Set;
import java.util.function.BiConsumer;
import java.util.function.BinaryOperator;
@ -16,7 +15,7 @@ public class TextBlockCollector implements Collector<TextBlock, ConcatenatedText
@Override
public Supplier<ConcatenatedTextBlock> supplier() {
return () -> new ConcatenatedTextBlock(Collections.emptyList());
return ConcatenatedTextBlock::empty;
}

4
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/redaction/service/AnalyzeService.java
View File

@ -33,7 +33,7 @@ import com.iqser.red.service.redaction.v1.server.layoutparsing.classification.mo
import com.iqser.red.service.redaction.v1.server.layoutparsing.classification.model.text.SimplifiedSectionText;
import com.iqser.red.service.redaction.v1.server.layoutparsing.classification.model.text.SimplifiedText;
import com.iqser.red.service.redaction.v1.server.layoutparsing.classification.service.PdfSegmentationService;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.data.mapper.DocumentDataMapper;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.data.DocumentData;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.data.mapper.DocumentGraphMapper;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.factory.DocumentGraphFactory;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.nodes.Document;
@ -109,7 +109,7 @@ public class AnalyzeService {
sectionGridCreatorService.createSectionGrid(classifiedDoc, pageCount);
log.info("Store document graph, text, simplified text, and section grid for file {} in dossier {}", analyzeRequest.getFileId(), analyzeRequest.getDossierId());
redactionStorageService.storeObject(analyzeRequest.getDossierId(), analyzeRequest.getFileId(), FileType.TEXT, DocumentDataMapper.toDocumentData(document));
redactionStorageService.storeObject(analyzeRequest.getDossierId(), analyzeRequest.getFileId(), FileType.TEXT, DocumentData.fromDocument(document));
redactionStorageService.storeObject(analyzeRequest.getDossierId(), analyzeRequest.getFileId(), FileType.SIMPLIFIED_TEXT, toSimplifiedText(document));
redactionStorageService.storeObject(analyzeRequest.getDossierId(), analyzeRequest.getFileId(), FileType.SECTION_GRID, classifiedDoc.getSectionGrid());

12
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/redaction/utils/TextNormalizationUtilities.java
View File

@ -16,4 +16,16 @@ public final class TextNormalizationUtilities {
return text.replaceAll("([^\\s\\d\\-]{2,500})[\\-\\u00AD]\\R", "$1");
}
public static String removeLineBreaks(String text) {
return text.replaceAll("\n", " ");
}
public static String removeRepeatingWhitespaces(String text) {
return text.replaceAll(" {2}", " ");
}
}

3
redaction-service-v1/redaction-service-server-v1/src/test/java/com/iqser/red/service/redaction/v1/server/document/graph/DocumentMappingIntegrationTest.java
View File

@ -10,7 +10,6 @@ import com.iqser.red.service.redaction.v1.server.layoutparsing.document.data.Ato
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.data.DocumentData;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.data.DocumentTreeData;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.data.PageData;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.data.mapper.DocumentDataMapper;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.data.mapper.DocumentGraphMapper;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.data.mapper.PropertiesMapper;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.nodes.Document;
@ -29,7 +28,7 @@ public class DocumentMappingIntegrationTest extends BuildDocumentIntegrationTest
String filename = "files/new/crafted document";
Document document = buildGraph(filename);
DocumentData documentData = DocumentDataMapper.toDocumentData(document);
DocumentData documentData = DocumentData.fromDocument(document);
storageService.storeJSONObject(TenantContext.getTenantId(), filename + "_PAGES" + ".json", documentData.getPages());
storageService.storeJSONObject(TenantContext.getTenantId(), filename + "_TEXT" + ".json", documentData.getAtomicTextBlocks());

5
redaction-service-v1/redaction-service-server-v1/src/test/resources/drools/rules.drl
View File

@ -135,7 +135,6 @@ rule "4: Redact Names and Addresses if no_redaction_indicator and redaction_indi
end
rule "5: Do not redact Names and Addresses if published information found"
when
$section: Section(hasEntitiesOfType("vertebrate"),
hasEntitiesOfType("published_information"),
@ -663,8 +662,8 @@ rule "remove Entity of type RECOMMENDATION when contained by ENTITY"
rule "remove Entity of lower rank, when equal boundaries and entityType"
salience 32
when
$higherRank: RedactionEntity($type: type, $entityType: entityType, $boundary: boundary)
$lowerRank: RedactionEntity($boundary == boundary, type != $type, entityType == $entityType, dictionary.getDictionaryRank(type) < dictionary.getDictionaryRank($type), !redaction)
$higherRank: RedactionEntity($type: type, $entityType: entityType)
$lowerRank: RedactionEntity(intersects($higherRank), type != $type, entityType == $entityType, dictionary.getDictionaryRank(type) < dictionary.getDictionaryRank($type), !resized, !skipRemoveEntitiesContainedInLarger)
then
$lowerRank.removeFromGraph();
retract($lowerRank);