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 
* implemented table merging when no headers present
This commit is contained in:
Kilian Schuettler 2023-05-04 13:08:24 +02:00
parent 06df8bd47f
commit bdd0692df9
9 changed files with 122 additions and 39 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/classification/model/table/Cell.java
View File

@ -27,6 +27,8 @@ public class Cell extends Rectangle {
private static final int MIN_SIZE = 1;
private int pageNumber;
public Cell(Point2D topLeft, Point2D bottomRight) {

13
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/factory/SectionNodeFactory.java
View File

@ -18,6 +18,7 @@ import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.no
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.nodes.PageNode;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.nodes.SectionNode;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.nodes.SemanticNode;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.utils.TableMergingUtility;
import lombok.experimental.UtilityClass;
@ -75,19 +76,25 @@ public class SectionNodeFactory {
private void addTablesOrParagraphsOrHeadlinesToSection(List<AbstractPageBlock> pageBlocks, DocumentGraphFactory.Context context, SectionNode sectionNode) {
Set<AbstractPageBlock> alreadyMerged = new HashSet<>();
List<AbstractPageBlock> remainingBlocks = new LinkedList<>(pageBlocks);
for (AbstractPageBlock abstractPageBlock : pageBlocks) {
if (alreadyMerged.contains(abstractPageBlock)) {
continue;
}
remainingBlocks.removeAll(alreadyMerged);
if (abstractPageBlock instanceof ClassificationTextBlock) {
List<ClassificationTextBlock> textBlocks = findTextBlocksWithSameClassificationAndAlignsY(abstractPageBlock, pageBlocks);
List<ClassificationTextBlock> textBlocks = findTextBlocksWithSameClassificationAndAlignsY(abstractPageBlock, remainingBlocks);
alreadyMerged.addAll(textBlocks);
DocumentGraphFactory.addParagraphOrHeadline(sectionNode, (ClassificationTextBlock) abstractPageBlock, context, textBlocks);
}
if (abstractPageBlock instanceof TablePageBlock) {
TableNodeFactory.addTable(sectionNode, (TablePageBlock) abstractPageBlock, context);
if (abstractPageBlock instanceof TablePageBlock tablePageBlock) {
List<TablePageBlock> tablesToMerge = TableMergingUtility.findConsecutiveTablesWithSameColCountAndSameHeaders(tablePageBlock, remainingBlocks);
alreadyMerged.addAll(tablesToMerge);
TableNodeFactory.addTable(sectionNode, tablesToMerge, context);
}
}
}

49
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/factory/TableNodeFactory.java
View File

@ -2,7 +2,10 @@ package com.iqser.red.service.redaction.v1.server.layoutparsing.document.factory
import static java.util.Collections.emptyList;
import java.util.Collection;
import java.util.List;
import java.util.Set;
import java.util.stream.Collectors;
import com.iqser.red.service.redaction.v1.server.layoutparsing.classification.model.AbstractPageBlock;
import com.iqser.red.service.redaction.v1.server.layoutparsing.classification.model.table.Cell;
@ -23,23 +26,40 @@ public class TableNodeFactory {
public static final double TABLE_CELL_MERGE_CONTENTS_SIZE_THRESHOLD = 0.05;
public void addTable(SemanticNode parentNode, TablePageBlock table, DocumentGraphFactory.Context context) {
public void addTable(SemanticNode parentNode, List<TablePageBlock> tablesToMerge, DocumentGraphFactory.Context context) {
PageNode page = context.getPage(table.getPage());
TableNode tableNode = TableNode.builder().tableOfContents(context.getTableOfContents()).numberOfCols(table.getColCount()).numberOfRows(table.getRowCount()).build();
setPageNumberInCellTextBlocks(tablesToMerge);
Set<PageNode> 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();
TableNode tableNode = TableNode.builder().tableOfContents(context.getTableOfContents()).numberOfCols(mergedRows.get(0).size()).numberOfRows(mergedRows.size()).build();
pages.forEach(page -> addTableToPage(page, parentNode, tableNode));
List<Integer> tocId = context.getTableOfContents().createNewChildEntryAndReturnId(parentNode.getTocId(), NodeType.TABLE, tableNode);
tableNode.setTocId(tocId);
addTableCells(mergedRows, tableNode, context);
IfTableHasNoHeadersAssumeFirstRowAreHeaders(tableNode);
}
private static void setPageNumberInCellTextBlocks(List<TablePageBlock> tablesToMerge) {
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()))));
}
private static void addTableToPage(PageNode page, SemanticNode parentNode, TableNode tableNode) {
if (!page.getMainBody().contains(parentNode)) {
parentNode.getPages().add(page);
}
page.getMainBody().add(tableNode);
List<Integer> tocId = context.getTableOfContents().createNewChildEntryAndReturnId(parentNode.getTocId(), NodeType.TABLE, tableNode);
tableNode.setTocId(tocId);
addTableCells(table.getRows(), tableNode, context, table.getPage());
IfTableHasNoHeadersAssumeFirstRowAreHeaders(tableNode);
}
@ -51,21 +71,20 @@ public class TableNodeFactory {
}
private void addTableCells(List<List<Cell>> rows, TableNode tableNode, DocumentGraphFactory.Context context, int pageNumber) {
private void addTableCells(List<List<Cell>> rows, TableNode tableNode, DocumentGraphFactory.Context context) {
for (int rowIndex = 0; rowIndex < rows.size(); rowIndex++) {
for (int colIndex = 0; colIndex < rows.get(rowIndex).size(); colIndex++) {
addTableCell(rows.get(rowIndex).get(colIndex), rowIndex, colIndex, tableNode, pageNumber, context);
addTableCell(rows.get(rowIndex).get(colIndex), rowIndex, colIndex, tableNode, context);
}
}
}
@SuppressWarnings("PMD.UnusedPrivateMethod") // PMD actually flags this wrong
private void addTableCell(Cell cell, int rowIndex, int colIndex, SemanticNode parentNode, int pageNumber, DocumentGraphFactory.Context context) {
private void addTableCell(Cell cell, int rowIndex, int colIndex, SemanticNode parentNode, DocumentGraphFactory.Context context) {
PageNode page = context.getPage(pageNumber);
cell.getTextBlocks().stream().filter(tb -> tb.getPage() == 0).forEach(tb -> tb.setPage(pageNumber));
PageNode page = context.getPage(cell.getPageNumber());
TableCellNode tableCellNode = TableCellNode.builder()
.tableOfContents(context.getTableOfContents())

16
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/services/EntityCreationService.java
View File

@ -1,7 +1,7 @@
package com.iqser.red.service.redaction.v1.server.layoutparsing.document.services;
import static com.iqser.red.service.redaction.v1.server.layoutparsing.document.utils.RedactionSearchUtils.getExpandedEndByRegex;
import static com.iqser.red.service.redaction.v1.server.layoutparsing.document.utils.RedactionSearchUtils.getExpandedStartByRegex;
import static com.iqser.red.service.redaction.v1.server.layoutparsing.document.utils.RedactionSearchUtility.getExpandedEndByRegex;
import static com.iqser.red.service.redaction.v1.server.layoutparsing.document.utils.RedactionSearchUtility.getExpandedStartByRegex;
import static com.iqser.red.service.redaction.v1.server.redaction.utils.SeparatorUtils.boundaryIsSurroundedBySeparators;
import static com.iqser.red.service.redaction.v1.server.redaction.utils.SeparatorUtils.isWhiteSpacesOrSeparatorsOnly;
@ -27,7 +27,7 @@ import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.no
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.nodes.SemanticNode;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.nodes.TableCellNode;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.textblock.TextBlock;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.utils.RedactionSearchUtils;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.utils.RedactionSearchUtility;
import com.iqser.red.service.redaction.v1.server.redaction.adapter.NerEntities;
import com.iqser.red.service.redaction.v1.server.redaction.model.EntityType;
import com.iqser.red.service.redaction.v1.server.redaction.model.dictionary.SearchImplementation;
@ -47,8 +47,8 @@ public class EntityCreationService {
public Stream<RedactionEntity> betweenStrings(String start, String stop, String type, EntityType entityType, SemanticNode node) {
TextBlock textBlock = node.buildTextBlock();
List<Boundary> startBoundaries = RedactionSearchUtils.findBoundariesByString(start, textBlock);
List<Boundary> stopBoundaries = RedactionSearchUtils.findBoundariesByString(stop, textBlock);
List<Boundary> startBoundaries = RedactionSearchUtility.findBoundariesByString(start, textBlock);
List<Boundary> stopBoundaries = RedactionSearchUtility.findBoundariesByString(stop, textBlock);
List<Boundary> entityBoundaries = new LinkedList<>();
if (startBoundaries.isEmpty() || stopBoundaries.isEmpty()) {
@ -90,7 +90,7 @@ public class EntityCreationService {
public Stream<RedactionEntity> lineAfterString(String string, String type, EntityType entityType, SemanticNode node) {
TextBlock textBlock = node.buildTextBlock();
return RedactionSearchUtils.findBoundariesByString(string, textBlock)
return RedactionSearchUtility.findBoundariesByString(string, textBlock)
.stream()
.map(boundary -> toLineAfterBoundary(textBlock, boundary))
.filter(boundary -> isValidEntityBoundary(textBlock, boundary))
@ -100,13 +100,13 @@ public class EntityCreationService {
public Stream<RedactionEntity> byRegex(String regexPattern, String type, EntityType entityType, SemanticNode node) {
return RedactionSearchUtils.findBoundariesByRegex(regexPattern, node.buildTextBlock()).stream().map(boundary -> byBoundary(boundary, type, entityType, node));
return RedactionSearchUtility.findBoundariesByRegex(regexPattern, node.buildTextBlock()).stream().map(boundary -> byBoundary(boundary, type, entityType, node));
}
public Stream<RedactionEntity> byString(String keyword, String type, EntityType entityType, SemanticNode node) {
return RedactionSearchUtils.findBoundariesByString(keyword, node.buildTextBlock()).stream().map(boundary -> byBoundary(boundary, type, entityType, node));
return RedactionSearchUtility.findBoundariesByString(keyword, node.buildTextBlock()).stream().map(boundary -> byBoundary(boundary, type, entityType, node));
}

5
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/utils/RedactionSearchUtils.java → redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/utils/RedactionSearchUtility.java
View File

@ -12,7 +12,10 @@ import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.en
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.textblock.TextBlock;
import com.iqser.red.service.redaction.v1.server.redaction.utils.Patterns;
public class RedactionSearchUtils {
import lombok.experimental.UtilityClass;
@UtilityClass
public class RedactionSearchUtility {
public static boolean anyMatch(CharSequence charSequence, String regexPattern) {

42
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/layoutparsing/document/utils/TableMergingUtility.java Normal file
View File

@ -0,0 +1,42 @@
package com.iqser.red.service.redaction.v1.server.layoutparsing.document.utils;
import java.util.Collection;
import java.util.LinkedList;
import java.util.List;
import java.util.stream.Stream;
import com.iqser.red.service.redaction.v1.server.layoutparsing.classification.model.AbstractPageBlock;
import com.iqser.red.service.redaction.v1.server.layoutparsing.classification.model.table.Cell;
import com.iqser.red.service.redaction.v1.server.layoutparsing.classification.model.table.TablePageBlock;
import lombok.experimental.UtilityClass;
@UtilityClass
public class TableMergingUtility {
public List<TablePageBlock> findConsecutiveTablesWithSameColCountAndSameHeaders(TablePageBlock originalTablePageBlock, List<AbstractPageBlock> pageBlocks) {
List<TablePageBlock> consecutiveTables = pageBlocks.stream()
.map(abstractPageBlock -> (TablePageBlock) abstractPageBlock)
.filter(tablePageBlock -> !tablePageBlock.equals(originalTablePageBlock))
.toList();
assert consecutiveTables.size() == pageBlocks.size() - 1;
List<TablePageBlock> consecutiveTablesWithSameColCountAndHeaders = new LinkedList<>();
for (TablePageBlock consecutiveTable : consecutiveTables) {
if (consecutiveTable.getColCount() == originalTablePageBlock.getColCount() && !hasTableHeader(consecutiveTable)) {
consecutiveTablesWithSameColCountAndHeaders.add(consecutiveTable);
} else {
break;
}
}
return Stream.concat(Stream.of(originalTablePageBlock), consecutiveTablesWithSameColCountAndHeaders.stream()).toList();
}
private boolean hasTableHeader(TablePageBlock table) {
return table.getRows().stream().flatMap(Collection::stream).anyMatch(Cell::isHeaderCell);
}
}

6
redaction-service-v1/redaction-service-server-v1/src/test/resources/drools/manual_redaction_rules.drl
View File

@ -1,8 +1,8 @@
package drools
import static java.lang.String.format;
import static com.iqser.red.service.redaction.v1.server.layoutparsing.document.utils.RedactionSearchUtils.anyMatch;
import static com.iqser.red.service.redaction.v1.server.layoutparsing.document.utils.RedactionSearchUtils.exactMatch;
import static com.iqser.red.service.redaction.v1.server.layoutparsing.document.utils.RedactionSearchUtility.anyMatch;
import static com.iqser.red.service.redaction.v1.server.layoutparsing.document.utils.RedactionSearchUtility.exactMatch;
import static com.iqser.red.service.redaction.v1.server.layoutparsing.document.data.mapper.PropertiesMapper.parseImageType;
import java.util.List;
@ -37,7 +37,7 @@ import com.iqser.red.service.redaction.v1.server.layoutparsing.document.graph.Bo
import java.util.stream.Collectors
import java.util.Collection
import java.util.stream.Stream
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.utils.RedactionSearchUtils;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.utils.RedactionSearchUtility;
global DocumentGraph document
global ManualRedactionApplicationService manualRedactionApplicationService

22
redaction-service-v1/redaction-service-server-v1/src/test/resources/drools/rules.drl
View File

@ -1,8 +1,8 @@
package drools
import static java.lang.String.format;
import static com.iqser.red.service.redaction.v1.server.layoutparsing.document.utils.RedactionSearchUtils.anyMatch;
import static com.iqser.red.service.redaction.v1.server.layoutparsing.document.utils.RedactionSearchUtils.exactMatch;
import static com.iqser.red.service.redaction.v1.server.layoutparsing.document.utils.RedactionSearchUtility.anyMatch;
import static com.iqser.red.service.redaction.v1.server.layoutparsing.document.utils.RedactionSearchUtility.exactMatch;
import static com.iqser.red.service.redaction.v1.server.layoutparsing.document.data.mapper.PropertiesMapper.parseImageType;
import java.util.List;
@ -37,7 +37,7 @@ import com.iqser.red.service.redaction.v1.server.redaction.adapter.NerEntities;
import java.util.stream.Collectors;
import java.util.Collection;
import java.util.stream.Stream;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.utils.RedactionSearchUtils;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.utils.RedactionSearchUtility;
global DocumentGraph document
global EntityCreationService entityCreationService
@ -509,7 +509,7 @@ rule "102: Guidelines FileAttributes"
when
$section: SectionNode(excludesTables, (containsString("DATA REQUIREMENT(S):") || containsString("TEST GUIDELINE(S):")) && (containsString("OECD") || containsString("EPA") || containsString("OPPTS")))
then
RedactionSearchUtils.findBoundariesByRegex("OECD (No\\.? )?\\d{3}( \\(\\d{4}\\))?", $section.buildTextBlock()).stream()
RedactionSearchUtility.findBoundariesByRegex("OECD (No\\.? )?\\d{3}( \\(\\d{4}\\))?", $section.buildTextBlock()).stream()
.map(boundary -> $section.buildTextBlock().subSequence(boundary).toString())
.map(value -> FileAttribute.builder().label("OECD Number").value(value).build())
.forEach(fileAttribute -> insert(fileAttribute));
@ -640,17 +640,27 @@ rule "remove Entity of type RECOMMENDATION when contained by FALSE_RECOMMENDATIO
retract($recommendation);
end
rule "remove Entity of type RECOMMENDATION when contained by ENTITY"
rule "remove Entity of type RECOMMENDATION when intersected by ENTITY with same type"
salience 256
when
$entity: RedactionEntity($type: type, entityType == EntityType.ENTITY)
$recommendation: RedactionEntity(containedBy($entity), type == $type, entityType == EntityType.RECOMMENDATION, !resized, !skipRemoveEntitiesContainedInLarger)
$recommendation: RedactionEntity(intersects($entity), type == $type, entityType == EntityType.RECOMMENDATION, !resized, !skipRemoveEntitiesContainedInLarger)
then
$entity.addEngines($recommendation.getEngines());
$recommendation.removeFromGraph();
retract($recommendation);
end
rule "remove Entity of type RECOMMENDATION when contained by ENTITY"
salience 256
when
$entity: RedactionEntity(entityType == EntityType.ENTITY)
$recommendation: RedactionEntity(containedBy($entity), entityType == EntityType.RECOMMENDATION, !resized, !skipRemoveEntitiesContainedInLarger)
then
$recommendation.removeFromGraph();
retract($recommendation);
end
rule "remove Entity of lower rank, when equal boundaries and entityType"
salience 32
when

6
redaction-service-v1/redaction-service-server-v1/src/test/resources/drools/rules_v2.drl
View File

@ -1,8 +1,8 @@
package drools
import static java.lang.String.format;
import static com.iqser.red.service.redaction.v1.server.layoutparsing.document.utils.RedactionSearchUtils.anyMatch;
import static com.iqser.red.service.redaction.v1.server.layoutparsing.document.utils.RedactionSearchUtils.exactMatch;
import static com.iqser.red.service.redaction.v1.server.layoutparsing.document.utils.RedactionSearchUtility.anyMatch;
import static com.iqser.red.service.redaction.v1.server.layoutparsing.document.utils.RedactionSearchUtility.exactMatch;
import static com.iqser.red.service.redaction.v1.server.layoutparsing.document.data.mapper.PropertiesMapper.parseImageType;
import java.util.List;
@ -37,7 +37,7 @@ import com.iqser.red.service.redaction.v1.server.redaction.adapter.NerEntities;
import java.util.stream.Collectors;
import java.util.Collection;
import java.util.stream.Stream;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.utils.RedactionSearchUtils;
import com.iqser.red.service.redaction.v1.server.layoutparsing.document.utils.RedactionSearchUtility;
global DocumentGraph document
global EntityCreationService entityCreationService