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

Fix entity span in table rows and detection of headers in rotated tables

Browse Source
This commit is contained in:
Thierry Göckel 2020-08-25 13:50:34 +02:00
parent 6483e637c6
commit 848c506c3f
8 changed files with 143 additions and 111 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/redaction/model/CellValue.java Normal file
View File

@ -0,0 +1,16 @@
package com.iqser.red.service.redaction.v1.server.redaction.model;
import com.iqser.red.service.redaction.v1.server.classification.model.TextBlock;
import lombok.RequiredArgsConstructor;
import lombok.Value;
@Value
@RequiredArgsConstructor
public class CellValue {
TextBlock textBlock;
int rowSpanStart;
}

14
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/redaction/model/Section.java
View File

@ -32,7 +32,7 @@ public class Section {
private int sectionNumber;
private Map<String, TextBlock> tabularData;
private Map<String, CellValue> tabularData;
public boolean rowEquals(String headerName, String value){
@ -40,7 +40,8 @@ public class Section {
.replaceAll(" ", "")
.replaceAll("-", "");
return tabularData != null && tabularData.containsKey(cleanHeaderName) && tabularData.get(cleanHeaderName).getText().equals(value);
return tabularData != null && tabularData.containsKey(cleanHeaderName)
&& tabularData.get(cleanHeaderName).getTextBlock().getText().equals(value);
}
@ -177,15 +178,18 @@ public class Section {
.replaceAll(" ", "")
.replaceAll("-", "");
TextBlock value = tabularData.get(cleanHeaderName);
CellValue value = tabularData.get(cleanHeaderName);
if (value == null) {
log.warn("Could not find any data for {}.", cellHeader);
} else {
Entity entity = new Entity(value.getText(), type, 0, value.getText().length(), headline, sectionNumber);
Entity entity = new Entity(value.getTextBlock()
.getText(), type, value.getRowSpanStart(), value.getRowSpanStart() + value.getTextBlock()
.getText()
.length(), headline, sectionNumber);
entity.setRedaction(false);
entity.setMatchedRule(ruleNumber);
entity.setRedactionReason(cellHeader);
entity.setTargetSequences(value.getSequences()); // Make sure no other cells with same content are highlighted
entity.setTargetSequences(value.getTextBlock().getSequences()); // Make sure no other cells with same content are highlighted
entities.add(entity);
}

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

@ -18,6 +18,7 @@ import com.iqser.red.service.redaction.v1.model.Rectangle;
import com.iqser.red.service.redaction.v1.server.classification.model.Document;
import com.iqser.red.service.redaction.v1.server.classification.model.Paragraph;
import com.iqser.red.service.redaction.v1.server.classification.model.TextBlock;
import com.iqser.red.service.redaction.v1.server.redaction.model.CellValue;
import com.iqser.red.service.redaction.v1.server.redaction.model.Entity;
import com.iqser.red.service.redaction.v1.server.redaction.model.EntityPositionSequence;
import com.iqser.red.service.redaction.v1.server.redaction.model.SearchableText;
@ -53,26 +54,27 @@ public class EntityRedactionService {
for (Table table : tables) {
for (List<Cell> row : table.getRows()) {
SearchableText searchableRow = new SearchableText();
Map<String, TextBlock> tabularData = new HashMap<>();
Map<String, CellValue> tabularData = new HashMap<>();
int start = 0;
for (Cell cell : row) {
if (cell.isHeaderCell() || CollectionUtils.isEmpty(cell.getTextBlocks())) {
continue;
}
addSectionToManualRedactions(cell.getTextBlocks(), manualRedactions, table.getHeadline(), sectionNumber);
int cellStart = start;
cell.getHeaderCells().forEach(headerCell -> {
StringBuilder headerBuilder = new StringBuilder();
headerCell.getTextBlocks().forEach(textBlock -> headerBuilder.append(textBlock.getText()));
String headerName = headerBuilder.toString()
.replaceAll("\n", "")
.replaceAll(" ", "")
.replaceAll("-", "");
tabularData.put(headerName, cell.getTextBlocks().get(0));
tabularData.put(headerName, new CellValue(cell.getTextBlocks().get(0), cellStart));
});
start = start + cell.getTextBlocks().get(0).toString().length();
for (TextBlock textBlock : cell.getTextBlocks()) {
searchableRow.addAll(textBlock.getSequences());
}
}
Set<Entity> rowEntities = findEntities(searchableRow, table.getHeadline(), sectionNumber);
@ -142,7 +144,7 @@ public class EntityRedactionService {
private Set<Entity> findEntities(SearchableText searchableText, String headline, int sectionNumber) {
Set<Entity> found = new HashSet<>();
if (StringUtils.isEmpty(searchableText.toString()) && StringUtils.isEmpty(headline)) {
if (StringUtils.isEmpty(searchableText.toString())) {
return found;
}

5
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/segmentation/SectionsBuilderService.java
View File

@ -85,8 +85,7 @@ public class SectionsBuilderService {
}).collect(Collectors.toList());
}
if (previousTableNonHeaderRow.size() == tableNonHeaderRow.size()) {
for (int i = currentTable.getRows()
.size() - 1; i >= 0; i--) { // Non header rows are most likely at bottom of table
for (int i = currentTable.getRowCount() - 1; i >= 0; i--) { // Non header rows are most likely at bottom of table
List<Cell> row = currentTable.getRows().get(i);
if (row.size() == tableNonHeaderRow.size() && row.stream()
.allMatch(cell -> cell.getHeaderCells().isEmpty())) {
@ -185,7 +184,7 @@ public class SectionsBuilderService {
private List<Cell> getRowWithNonHeaderCells(Table table) {
for (int i = table.getRows().size() - 1; i >= 0; i--) { // Non header rows are most likely at bottom of table
for (int i = table.getRowCount() - 1; i >= 0; i--) { // Non header rows are most likely at bottom of table
List<Cell> row = table.getRows().get(i);
boolean allNonHeader = true;
for (Cell cell : row) {

167
redaction-service-v1/redaction-service-server-v1/src/main/java/com/iqser/red/service/redaction/v1/server/tableextraction/model/Table.java
View File

@ -29,11 +29,13 @@ public class Table extends AbstractTextContainer {
@Setter
private String headline;
@Getter
private int rowCount;
private int unrotatedRowCount;
@Getter
private int colCount;
private int unrotatedColCount;
private int rowCount = -1;
private int colCount = -1;
private final int rotation;
@ -65,6 +67,25 @@ public class Table extends AbstractTextContainer {
}
public int getRowCount() {
if (rowCount == -1) {
rowCount = getRows().size();
}
return rowCount;
}
public int getColCount() {
if (colCount == -1) {
colCount = getRows().stream().mapToInt(List::size).max().orElse(0);
}
return colCount;
}
/**
* Detect header cells (either first row or first column):
* Column is marked as header if cell text is bold and row cell text is not bold.
@ -72,100 +93,50 @@ public class Table extends AbstractTextContainer {
*/
private void computeHeaders() {
if (rows == null) {
rows = computeRows();
}
// A bold cell is a header cell as long as every cell to the left/top is bold, too
cells.forEach((position, cell) -> {
List<Cell> cellsToTheLeft = getCellsToTheLeft(position);
Cell lastHeaderCell = null;
for (Cell leftCell : cellsToTheLeft) {
if (CollectionUtils.isNotEmpty(leftCell.getTextBlocks()) && leftCell.getTextBlocks()
// we move from left to right and top to bottom
for (int rowIndex = 0; rowIndex < rows.size(); rowIndex++) {
List<Cell> rowCells = rows.get(rowIndex);
for (int colIndex = 0; colIndex < rowCells.size(); colIndex++) {
Cell cell = rowCells.get(colIndex);
List<Cell> cellsToTheLeft = rowCells.subList(0, colIndex);
Cell lastHeaderCell = null;
for (Cell leftCell : cellsToTheLeft) {
if (leftCell.isHeaderCell()) {
lastHeaderCell = leftCell;
} else {
break;
}
}
if (lastHeaderCell != null) {
cell.getHeaderCells().add(lastHeaderCell);
}
List<Cell> cellsToTheTop = new ArrayList<>();
for (int i = rowIndex - 1; i >= 0; i--) {
cellsToTheTop.add(rows.get(i).get(colIndex));
}
for (Cell topCell : cellsToTheTop) {
if (topCell.isHeaderCell()) {
lastHeaderCell = topCell;
} else {
break;
}
}
if (lastHeaderCell != null) {
cell.getHeaderCells().add(lastHeaderCell);
}
if (CollectionUtils.isNotEmpty(cell.getTextBlocks()) && cell.getTextBlocks()
.get(0)
.getMostPopularWordStyle()
.equals("bold")) {
lastHeaderCell = leftCell;
} else {
break;
cell.setHeaderCell(true);
}
}
if (lastHeaderCell != null) {
cell.getHeaderCells().add(lastHeaderCell);
}
lastHeaderCell = null;
List<Cell> cellsToTheTop = getCellToTheTop(position);
for (Cell topCell : cellsToTheTop) {
if (CollectionUtils.isNotEmpty(topCell.getTextBlocks()) && topCell.getTextBlocks()
.get(0)
.getMostPopularWordStyle()
.equals("bold")) {
lastHeaderCell = topCell;
} else {
break;
}
}
if (lastHeaderCell != null) {
cell.getHeaderCells().add(lastHeaderCell);
}
if (CollectionUtils.isNotEmpty(cell.getTextBlocks()) && cell.getTextBlocks()
.get(0)
.getMostPopularWordStyle()
.equals("bold")) {
cell.setHeaderCell(true);
}
});
}
private List<Cell> getCellsToTheLeft(CellPosition cellPosition) {
List<Cell> result = new ArrayList<>();
if (cellPosition.getCol() == 0) {
return result;
}
int row = cellPosition.getRow();
for (int i = cellPosition.getCol() - 1; i >= 0; i--) {
if (cells.get(new CellPosition(row, i)) != null) {
result.add(cells.get(new CellPosition(row, i)));
} else {
Cell spanningCell = null;
while (spanningCell == null && row >= 0) {
row--;
spanningCell = cells.get(new CellPosition(row, i));
}
if (spanningCell != null) {
result.add(spanningCell);
}
row = cellPosition.getRow();
}
}
Collections.reverse(result);
return result;
}
private List<Cell> getCellToTheTop(CellPosition cellPosition) {
List<Cell> result = new ArrayList<>();
if (cellPosition.getRow() == 0) {
return result;
}
int col = cellPosition.getCol();
for (int i = cellPosition.getRow() - 1; i >= 0; i--) {
if (cells.get(new CellPosition(i, col)) != null) {
result.add(cells.get(new CellPosition(i, col)));
} else {
Cell spanningCell = null;
while (spanningCell == null && col >= 0) {
col--;
spanningCell = cells.get(new CellPosition(i, col));
}
if (spanningCell != null) {
result.add(spanningCell);
}
col = cellPosition.getCol();
}
}
Collections.reverse(result);
return result;
}
@ -173,9 +144,9 @@ public class Table extends AbstractTextContainer {
List<List<Cell>> rows = new ArrayList<>();
if (rotation == 90) {
for (int i = 0; i < colCount; i++) { // rows
for (int i = 0; i < unrotatedColCount; i++) { // rows
List<Cell> lastRow = new ArrayList<>();
for (int j = rowCount - 1; j >= 0; j--) { // cols
for (int j = unrotatedRowCount - 1; j >= 0; j--) { // cols
Cell cell = cells.get(new CellPosition(j, i));
if (cell != null) {
lastRow.add(cell);
@ -184,9 +155,9 @@ public class Table extends AbstractTextContainer {
rows.add(lastRow);
}
} else if (rotation == 270) {
for (int i = colCount - 1; i >= 0; i--) { // rows
for (int i = unrotatedColCount - 1; i >= 0; i--) { // rows
List<Cell> lastRow = new ArrayList<>();
for (int j = 0; j < rowCount; j++) { // cols
for (int j = 0; j < unrotatedRowCount; j++) { // cols
Cell cell = cells.get(new CellPosition(i, j));
if (cell != null) {
lastRow.add(cell);
@ -195,9 +166,9 @@ public class Table extends AbstractTextContainer {
rows.add(lastRow);
}
} else {
for (int i = 0; i < rowCount; i++) {
for (int i = 0; i < unrotatedRowCount; i++) {
List<Cell> lastRow = new ArrayList<>();
for (int j = 0; j < colCount; j++) {
for (int j = 0; j < unrotatedColCount; j++) {
Cell cell = cells.get(new CellPosition(i, j)); // JAVA_8 use getOrDefault()
if (cell != null) {
lastRow.add(cell);
@ -214,8 +185,8 @@ public class Table extends AbstractTextContainer {
private void add(Cell chunk, int row, int col) {
rowCount = Math.max(rowCount, row + 1);
colCount = Math.max(colCount, col + 1);
unrotatedRowCount = Math.max(unrotatedRowCount, row + 1);
unrotatedColCount = Math.max(unrotatedColCount, col + 1);
CellPosition cp = new CellPosition(row, col);
cells.put(cp, chunk);

40
redaction-service-v1/redaction-service-server-v1/src/test/java/com/iqser/red/service/redaction/v1/server/segmentation/PdfSegmentationServiceTest.java
View File

@ -146,4 +146,44 @@ public class PdfSegmentationServiceTest {
}
}
@Test
public void testHeaderCellsForRotatedTable() throws IOException {
ClassPathResource pdfFileResource = new ClassPathResource("files/Minimal Examples/Rotated Table Headers.pdf");
try (PDDocument pdDocument = PDDocument.load(pdfFileResource.getInputStream())) {
Document document = pdfSegmentationService.parseDocument(pdDocument);
assertThat(document.getParagraphs()
.stream()
.flatMap(paragraph -> paragraph.getTables().stream())
.collect(Collectors.toList())).isNotEmpty();
Table firstTable = document.getParagraphs()
.stream()
.flatMap(paragraph -> paragraph.getTables().stream())
.collect(Collectors.toList())
.get(0);
assertThat(firstTable.getColCount()).isEqualTo(8);
assertThat(firstTable.getRowCount()).isEqualTo(1);
Table secondTable = document.getParagraphs()
.stream()
.flatMap(paragraph -> paragraph.getTables().stream())
.collect(Collectors.toList())
.get(1);
assertThat(secondTable.getColCount()).isEqualTo(8);
assertThat(secondTable.getRowCount()).isEqualTo(6);
List<List<Cell>> firstTableHeaderCells = firstTable.getRows()
.get(0)
.stream()
.map(Collections::singletonList)
.collect(Collectors.toList());
assertThat(secondTable.getRows().stream()
.allMatch(row -> row.stream()
.map(Cell::getHeaderCells)
.collect(Collectors.toList())
.equals(firstTableHeaderCells)))
.isTrue();
}
}
}

BIN
redaction-service-v1/redaction-service-server-v1/src/test/resources/files/Minimal Examples/Empty Tabular Data.pdf Normal file
View File

Binary file not shown.

BIN
redaction-service-v1/redaction-service-server-v1/src/test/resources/files/Minimal Examples/Rotated Table Headers.pdf Normal file
View File

Binary file not shown.