Compare commits

...
9 Commits
Author SHA1 Message Date
Maverick Studer 52e948e66c Merge branch 'RED-9331' into 'main'
RED-9331: Explore possibilities for fair upload / analysis processing per tenant

See merge request fforesight/layout-parser!182
2024-08-27 09:27:37 +02:00
Maverick Studer 3b33405cbf RED-9331: Explore possibilities for fair upload / analysis processing per tenant 2024-08-27 09:27:37 +02:00
Maverick Studer b2fa14dde2 Merge branch 'AZURE_NER' into 'main'
RED-9918: Azure entity recognition (Spike)

See merge request fforesight/layout-parser!196
2024-08-26 14:34:46 +02:00
Maverick Studer 62e07686d7 RED-9918: Azure entity recognition (Spike) 2024-08-26 14:34:46 +02:00
Dominique Eifländer 3eb97d614f Merge branch 'RED-9760-NPE' into 'main'
RED-9760: Fixed nullpointer in TextPageBlock

See merge request fforesight/layout-parser!194
2024-08-13 13:24:48 +02:00
Dominique Eifländer 81469413b0 RED-9760: Fixed nullpointer in TextPageBlock 2024-08-13 13:18:50 +02:00
Dominique Eifländer 2993676a6f Merge branch 'RED-9670' into 'main'
RED-9760: change compareDouble to something sensible

See merge request fforesight/layout-parser!193
2024-08-12 16:02:51 +02:00
Kilian Schüttler 8e115dcd8a RED-9760: change compareDouble to something sensible 2024-08-12 16:02:50 +02:00
Dominique Eifländer 173911b840 Merge branch 'hotfix-reading-order' into 'main'
hotfix: threshold adjustements

See merge request fforesight/layout-parser!192
2024-08-12 14:59:20 +02:00
17 changed files with 249 additions and 39 deletions
@@ -21,5 +21,14 @@ public class SimplifiedText {
@Schema(description = "A List of simplified Sections, which contains almost exclusively text.")
@Builder.Default
private List<SimplifiedSectionText> sectionTexts = new ArrayList<>();
@Schema(description = "A list of the main section numbers ")
@Builder.Default
private List<String> mainSectionNumbers = new ArrayList<>();
@Schema(description = "A list of the header section numbers ")
@Builder.Default
private List<String> headerSectionNumbers = new ArrayList<>();
@Schema(description = "A list of the footer section numbers ")
@Builder.Default
private List<String> footerSectionNumbers = new ArrayList<>();
}
@@ -2,6 +2,9 @@ package com.knecon.fforesight.service.layoutparser.internal.api.queue;
public class LayoutParsingQueueNames {
public static final String LAYOUT_PARSING_REQUEST_QUEUE = "layout_parsing_request_queue";
public static final String LAYOUT_PARSING_FINISHED_EVENT_QUEUE = "layout_parsing_response_queue";
public static final String LAYOUT_PARSING_REQUEST_QUEUE_PREFIX = "layout_parsing_request_queue";
public static final String LAYOUT_PARSING_REQUEST_EXCHANGE = "layout_parsing_request_exchange";
public static final String LAYOUT_PARSING_RESPONSE_QUEUE_PREFIX = "layout_parsing_response_queue";
public static final String LAYOUT_PARSING_RESPONSE_EXCHANGE = "layout_parsing_response_exchange";
public static final String LAYOUT_PARSING_DLQ = "layout_parsing_dlq";
}
@@ -16,7 +16,7 @@ dependencies {
exclude("org.springframework.boot", "spring-boot-starter-security")
exclude("org.springframework.boot", "spring-boot-starter-validation")
}
implementation("com.knecon.fforesight:tenant-commons:0.21.0")
implementation("com.knecon.fforesight:tenant-commons:0.28.0")
implementation("com.iqser.red.commons:storage-commons:2.45.0")
implementation("org.apache.pdfbox:pdfbox:${pdfBoxVersion}")
@@ -9,6 +9,7 @@ import java.io.IOException;
import java.nio.file.Files;
import java.util.ArrayList;
import java.util.HashMap;
import java.util.HashSet;
import java.util.List;
import java.util.Map;
import java.util.concurrent.atomic.AtomicReference;
@@ -118,18 +119,14 @@ public class LayoutParsingPipeline {
log.info("Starting layout parsing for {}", layoutParsingRequest.identifier());
File originFile = layoutParsingStorageService.getOriginFile(layoutParsingRequest.originFileStorageId());
File viewerDocumentFile = layoutParsingStorageService.getViewerDocFile(layoutParsingRequest.viewerDocumentStorageId())
.orElse(originFile);
File viewerDocumentFile = layoutParsingStorageService.getViewerDocFile(layoutParsingRequest.viewerDocumentStorageId()).orElse(originFile);
VisualLayoutParsingResponse visualLayoutParsingResponse = layoutParsingRequest.visualLayoutParsingFileId()
.map(layoutParsingStorageService::getVisualLayoutParsingFile)
.orElse(new VisualLayoutParsingResponse());
.map(layoutParsingStorageService::getVisualLayoutParsingFile).orElse(new VisualLayoutParsingResponse());
ImageServiceResponse imageServiceResponse = layoutParsingRequest.imagesFileStorageId()
.map(layoutParsingStorageService::getImagesFile)
.orElse(new ImageServiceResponse());
.map(layoutParsingStorageService::getImagesFile).orElse(new ImageServiceResponse());
TableServiceResponse tableServiceResponse = layoutParsingRequest.tablesFileStorageId()
.map(layoutParsingStorageService::getTablesFile)
.orElse(new TableServiceResponse());
.map(layoutParsingStorageService::getTablesFile).orElse(new TableServiceResponse());
ClassificationDocument classificationDocument = parseLayout(settings.getLayoutParsingTypeOverride() == null //
? layoutParsingRequest.layoutParsingType() : settings.getLayoutParsingTypeOverride(),
@@ -278,7 +275,9 @@ public class LayoutParsingPipeline {
stripper.getText(originDocument);
List<TextPositionSequence> words = stripper.getTextPositionSequences();
if (layoutParsingType.equals(LayoutParsingType.DOCUMINE_OLD)) {
words = TextPositionOperations.sort(words);
var lines = TextPositionOperations.groupByLine(new HashSet<>(words));
classificationDocument.getLayoutDebugLayer().addLineVisualizationsFromNestedTextPosition(lines, pageNumber);
words = TextPositionOperations.sortLines(lines);
}
classificationDocument.getLayoutDebugLayer().addTextVisualizations(words, pageNumber);
@@ -7,9 +7,12 @@ public class DoubleUtils {
if (Double.isNaN(d1) || Double.isNaN(d2)) {
return Double.compare(d1, d2);
}
long i1 = Math.round(d1 / (precision == 0 ? 1 : precision));
long i2 = Math.round(d2 / (precision == 0 ? 1 : precision));
return Long.compare(i1, i2);
if (Math.abs(d1 - d2) < precision) {
return 0;
}
return Double.compare(d1, d2);
}
}
@@ -49,7 +49,9 @@ public class TextPageBlock extends AbstractPageBlock {
public TextPageBlock(List<TextPositionSequence> sequences) {
this.sequences = sequences;
calculateFrequencyCounters();
if (!sequences.isEmpty()) {
calculateFrequencyCounters();
}
calculateBBox();
}
@@ -34,7 +34,22 @@ public class SimplifiedSectionTextService {
List<SimplifiedSectionText> simplifiedText = Stream.of(simplifiedMainSectionsList, simplifiedHeadersList, simplifiedFootersList)
.flatMap(List::stream)
.collect(Collectors.toList());
return SimplifiedText.builder().numberOfPages(document.getNumberOfPages()).sectionTexts(simplifiedText).build();
return SimplifiedText.builder()
.numberOfPages(document.getNumberOfPages())
.sectionTexts(simplifiedText)
.mainSectionNumbers(document.getAllSections()
.stream()
.map(this::getSectionNumber)
.toList())
.headerSectionNumbers(document.getHeaders()
.stream()
.map(this::getSectionNumber)
.toList())
.footerSectionNumbers(document.getFooters()
.stream()
.map(this::getSectionNumber)
.toList())
.build();
}
@@ -49,4 +64,13 @@ public class SimplifiedSectionTextService {
.build();
}
private String getSectionNumber(SemanticNode semanticNode) {
return semanticNode.getTreeId()
.stream()
.map(String::valueOf)
.collect(Collectors.joining("."));
}
}
@@ -27,7 +27,7 @@ public class TextPositionOperations {
public static final double MAX_LINE_HEIGHT_FACTOR = 0.66; // multiplied with max word height
public static final double MAX_WORD_DISTANCE_FACTOR = 3.5; // multiplied with max word width
private static final double Y_THRESHOLD = 6;
private static final double Y_THRESHOLD = 5;
private static final double X_THRESHOLD = 3;
private static final Comparator<TextBoundingBox> COMPARATOR_DIR_ADJ = //
Comparator.comparing(TextBoundingBox::getDir)
@@ -53,13 +53,19 @@ public class TextPositionOperations {
private List<TextPositionSequence> sortUsingLineDetection(Set<TextPositionSequence> sequences) {
return groupByLine(sequences).stream()
return sortLines(groupByLine(sequences));
}
public List<TextPositionSequence> sortLines(Collection<Set<TextPositionSequence>> lines) {
return lines.stream()
.map(TextPositionOperations::sortByXDirAdj)
.filter(line -> !line.isEmpty())
.sorted(Comparator.comparing(line -> line.get(0), COMPARATOR_DIR_ADJ))
.flatMap(Collection::stream)
.toList();
}
@@ -71,7 +77,7 @@ public class TextPositionOperations {
}
private Collection<Set<TextPositionSequence>> groupByLine(Set<TextPositionSequence> sequences) {
public Collection<Set<TextPositionSequence>> groupByLine(Set<TextPositionSequence> sequences) {
double maxLineDistance = sequences.stream()
.map(TextPositionSequence::getBBoxDirAdj)
@@ -6,6 +6,7 @@ import java.awt.geom.Point2D;
import java.awt.geom.Rectangle2D;
import java.util.Collection;
import java.util.List;
import java.util.Set;
import java.util.concurrent.atomic.AtomicInteger;
import org.apache.pdfbox.pdmodel.documentinterchange.markedcontent.PDMarkedContent;
@@ -150,6 +151,23 @@ public class LayoutDebugLayer extends LayoutDebugLayerConfig {
}
public void addLineVisualizationsFromNestedTextPosition(Collection<Set<TextPositionSequence>> lines, int pageNumber) {
if (!active) {
return;
}
VisualizationsOnPage visualizationsOnPage = getOrCreateVisualizationsOnPage(pageNumber, this.lines);
visualizationsOnPage.getColoredRectangles()
.addAll(lines.stream()
.map(line -> line.stream()
.map(BoundingBox::getBBoxPdf)
.collect(RectangleTransformations.collectBBox()))
.map(line -> new ColoredRectangle(line, LINES_COLOR, 0.5f))
.toList());
}
public void addTextBlockVisualizations(List<TextPageBlock> textPageBlocks, int page) {
if (!active) {
@@ -235,4 +253,5 @@ public class LayoutDebugLayer extends LayoutDebugLayerConfig {
}
}
@@ -30,7 +30,7 @@ dependencies {
implementation(project(":layoutparser-service-internal-api"))
implementation("com.iqser.red.commons:storage-commons:2.45.0")
implementation("com.knecon.fforesight:tenant-commons:0.21.0")
implementation("com.knecon.fforesight:tenant-commons:0.28.0")
implementation("com.knecon.fforesight:tracing-commons:0.5.0")
implementation("com.knecon.fforesight:lifecycle-commons:0.6.0")
implementation("org.springframework.boot:spring-boot-starter-actuator:${springBootStarterVersion}")
@@ -0,0 +1,36 @@
package com.knecon.fforesight.service.layoutparser.server.configuration;
import static com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingQueueNames.LAYOUT_PARSING_DLQ;
import org.springframework.amqp.core.DirectExchange;
import org.springframework.amqp.core.Queue;
import org.springframework.amqp.core.QueueBuilder;
import org.springframework.context.annotation.Bean;
import org.springframework.context.annotation.Configuration;
import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingQueueNames;
@Configuration
public class MessagingConfiguration {
@Bean
public DirectExchange layoutParsingResponseExchange() {
return new DirectExchange(LayoutParsingQueueNames.LAYOUT_PARSING_RESPONSE_EXCHANGE);
}
@Bean
public DirectExchange layoutParsingRequestExchange() {
return new DirectExchange(LayoutParsingQueueNames.LAYOUT_PARSING_REQUEST_EXCHANGE);
}
@Bean
public Queue layoutParsingDLQ() {
return QueueBuilder.durable(LAYOUT_PARSING_DLQ).build();
}
}
@@ -0,0 +1,11 @@
package com.knecon.fforesight.service.layoutparser.server.configuration;
import org.springframework.context.annotation.Configuration;
import com.knecon.fforesight.tenantcommons.queue.TenantMessagingConfiguration;
@Configuration
public class TenantMessagingConfigurationImpl extends TenantMessagingConfiguration {
}
@@ -18,6 +18,7 @@ import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsi
import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingRequest;
import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingType;
import com.knecon.fforesight.service.layoutparser.processor.LayoutParsingPipeline;
import com.knecon.fforesight.tenantcommons.TenantContext;
import lombok.RequiredArgsConstructor;
import lombok.SneakyThrows;
@@ -28,15 +29,17 @@ import lombok.extern.slf4j.Slf4j;
@RequiredArgsConstructor
public class MessageHandler {
public static final String LAYOUT_PARSING_REQUEST_LISTENER_ID = "layout-parsing-request-listener";
private final LayoutParsingPipeline layoutParsingPipeline;
private final ObjectMapper objectMapper;
private final RabbitTemplate rabbitTemplate;
private final static String X_PIPELINE_PREFIX = "X-PIPE-";
@RabbitHandler
@RabbitListener(queues = LayoutParsingQueueNames.LAYOUT_PARSING_REQUEST_QUEUE)
@SneakyThrows
@RabbitHandler
@RabbitListener(id = LAYOUT_PARSING_REQUEST_LISTENER_ID)
public void receiveLayoutParsingRequest(Message message) {
LayoutParsingRequest layoutParsingRequest = objectMapper.readValue(message.getBody(), LayoutParsingRequest.class);
@@ -57,7 +60,7 @@ public class MessageHandler {
public void sendLayoutParsingFinishedEvent(LayoutParsingFinishedEvent layoutParsingFinishedEvent, Message message) {
Arrays.stream(layoutParsingFinishedEvent.message().split("\n")).forEach(log::info);
rabbitTemplate.convertAndSend(LayoutParsingQueueNames.LAYOUT_PARSING_FINISHED_EVENT_QUEUE, layoutParsingFinishedEvent, m -> {
rabbitTemplate.convertAndSend(LayoutParsingQueueNames.LAYOUT_PARSING_RESPONSE_EXCHANGE, TenantContext.getTenantId(), layoutParsingFinishedEvent, m -> {
var forwardHeaders = message.getMessageProperties()
.getHeaders()
.entrySet()
@@ -0,0 +1,64 @@
package com.knecon.fforesight.service.layoutparser.server.queue;
import java.util.Set;
import org.springframework.amqp.rabbit.annotation.RabbitHandler;
import org.springframework.amqp.rabbit.annotation.RabbitListener;
import org.springframework.boot.context.event.ApplicationReadyEvent;
import org.springframework.context.event.EventListener;
import org.springframework.stereotype.Service;
import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingQueueNames;
import com.knecon.fforesight.tenantcommons.TenantProvider;
import com.knecon.fforesight.tenantcommons.model.TenantCreatedEvent;
import com.knecon.fforesight.tenantcommons.model.TenantQueueConfiguration;
import com.knecon.fforesight.tenantcommons.model.TenantResponse;
import com.knecon.fforesight.tenantcommons.queue.RabbitQueueFromExchangeService;
import com.knecon.fforesight.tenantcommons.queue.TenantExchangeMessageReceiver;
@Service
public class TenantExchangeMessageReceiverImpl extends TenantExchangeMessageReceiver {
public TenantExchangeMessageReceiverImpl(RabbitQueueFromExchangeService rabbitQueueService, TenantProvider tenantProvider) {
super(rabbitQueueService, tenantProvider);
}
@Override
protected Set<TenantQueueConfiguration> getTenantQueueConfigs() {
return Set.of(TenantQueueConfiguration.builder()
.listenerId(MessageHandler.LAYOUT_PARSING_REQUEST_LISTENER_ID)
.exchangeName(LayoutParsingQueueNames.LAYOUT_PARSING_REQUEST_EXCHANGE)
.queuePrefix(LayoutParsingQueueNames.LAYOUT_PARSING_REQUEST_QUEUE_PREFIX)
.dlqName(LayoutParsingQueueNames.LAYOUT_PARSING_DLQ)
.build());
}
@EventListener(ApplicationReadyEvent.class)
public void onApplicationReady() {
System.out.println("application ready invoked");
super.initializeQueues();
}
@RabbitHandler
@RabbitListener(queues = "#{tenantMessagingConfigurationImpl.getTenantCreatedQueueName()}")
public void reactToTenantCreation(TenantCreatedEvent tenantCreatedEvent) {
super.reactToTenantCreation(tenantCreatedEvent);
}
@RabbitHandler
@RabbitListener(queues = "#{tenantMessagingConfigurationImpl.getTenantDeletedQueueName()}")
public void reactToTenantDeletion(TenantResponse tenantResponse) {
super.reactToTenantDeletion(tenantResponse);
}
}
@@ -10,7 +10,9 @@ import java.util.stream.Collectors;
import org.assertj.core.api.Assertions;
import org.junit.jupiter.api.Test;
import org.junit.jupiter.api.extension.ExtendWith;
import org.springframework.amqp.rabbit.core.RabbitAdmin;
import org.springframework.amqp.rabbit.core.RabbitTemplate;
import org.springframework.amqp.rabbit.listener.RabbitListenerEndpointRegistry;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.boot.autoconfigure.EnableAutoConfiguration;
import org.springframework.boot.autoconfigure.amqp.RabbitAutoConfiguration;
@@ -55,6 +57,12 @@ public class HeadlinesGoldStandardIntegrationTest {
@MockBean
private RabbitTemplate rabbitTemplate;
@MockBean
private RabbitAdmin rabbitAdmin;
@MockBean
private RabbitListenerEndpointRegistry rabbitListenerEndpointRegistry;
@Autowired
private ObjectMapper objectMapper;
@@ -70,13 +78,18 @@ public class HeadlinesGoldStandardIntegrationTest {
List<Metrics> metrics = new ArrayList<>();
metrics.add(getMetrics("files/syngenta/CustomerFiles/01 - CGA100251 - Acute Oral Toxicity (Up and Down Procedure) - Rat (1).pdf",
"files/headlineTest/01 - CGA100251 - Acute Oral Toxicity (Up and Down Procedure) - Rat (1)_REDACTION_LOG.json"));
"files/headlineTest/01 - CGA100251 - Acute Oral Toxicity (Up and Down Procedure) - Rat (1)_REDACTION_LOG.json"));
metrics.add(getMetrics("files/syngenta/CustomerFiles/91 Trinexapac-ethyl_RAR_01_Volume_1_2018-02-23.pdf",
"files/headlineTest/91 Trinexapac-ethyl_RAR_01_Volume_1_2018-02-23_REDACTION_LOG.json"));
metrics.add(getMetrics("files/syngenta/CustomerFiles/S-Metolachlor_RAR_01_Volume_1_2018-09-06.pdf", "files/headlineTest/S-Metolachlor_RAR_01_Volume_1_2018-09-06_REDACTION_LOG.json"));
"files/headlineTest/91 Trinexapac-ethyl_RAR_01_Volume_1_2018-02-23_REDACTION_LOG.json"));
metrics.add(getMetrics("files/syngenta/CustomerFiles/S-Metolachlor_RAR_01_Volume_1_2018-09-06.pdf",
"files/headlineTest/S-Metolachlor_RAR_01_Volume_1_2018-09-06_REDACTION_LOG.json"));
double precision = metrics.stream().mapToDouble(Metrics::getPrecision).average().orElse(1.0);
double recall = metrics.stream().mapToDouble(Metrics::getRecall).average().orElse(1.0);
double precision = metrics.stream()
.mapToDouble(Metrics::getPrecision).average()
.orElse(1.0);
double recall = metrics.stream()
.mapToDouble(Metrics::getRecall).average()
.orElse(1.0);
System.out.println("Precision is: " + precision + " recall is: " + recall);
@@ -93,21 +106,28 @@ public class HeadlinesGoldStandardIntegrationTest {
Set<Headline> goldStandardHeadlines = new HashSet<>();
var goldStandardLog = objectMapper.readValue(redactionLogResource.getInputStream(), RedactionLog.class);
goldStandardLog.getRedactionLogEntry().removeIf(r -> !r.isRedacted() || r.getChanges().get(r.getChanges().size() - 1).getType().equals(ChangeType.REMOVED));
goldStandardLog.getRedactionLogEntry().forEach(e -> goldStandardHeadlines.add(new Headline(e.getPositions().get(0).getPage(), e.getValue())));
goldStandardLog.getRedactionLogEntry()
.removeIf(r -> !r.isRedacted() || r.getChanges()
.get(r.getChanges().size() - 1).getType().equals(ChangeType.REMOVED));
goldStandardLog.getRedactionLogEntry()
.forEach(e -> goldStandardHeadlines.add(new Headline(e.getPositions()
.get(0).getPage(), e.getValue())));
Document documentGraph = DocumentGraphFactory.buildDocumentGraph(LayoutParsingType.DOCUMINE,
layoutParsingPipeline.parseLayout(LayoutParsingType.DOCUMINE,
pdfFileResource.getFile(),
new ImageServiceResponse(),
new TableServiceResponse(),
new VisualLayoutParsingResponse(),
Map.of("file",filePath)));
pdfFileResource.getFile(),
new ImageServiceResponse(),
new TableServiceResponse(),
new VisualLayoutParsingResponse(),
Map.of("file", filePath)));
var foundHeadlines = documentGraph.streamAllSubNodes()
.map(SemanticNode::getHeadline)
.distinct()
.map(headlineNode -> new Headline(headlineNode.getPages().stream().findFirst().get().getNumber(), headlineNode.getTextBlock().getSearchText().stripTrailing()))
.map(headlineNode -> new Headline(headlineNode.getPages()
.stream()
.findFirst()
.get().getNumber(), headlineNode.getTextBlock().getSearchText().stripTrailing()))
.toList();
Set<Headline> correct = new HashSet<>();
@@ -121,7 +141,9 @@ public class HeadlinesGoldStandardIntegrationTest {
}
}
missing = goldStandardHeadlines.stream().filter(h -> !correct.contains(h)).collect(Collectors.toSet());
missing = goldStandardHeadlines.stream()
.filter(h -> !correct.contains(h))
.collect(Collectors.toSet());
float precision = (float) correct.size() / (float) foundHeadlines.size();
float recall = (float) correct.size() / ((float) correct.size() + (float) missing.size());
@@ -10,7 +10,9 @@ import java.util.Optional;
import org.junit.jupiter.api.AfterEach;
import org.junit.jupiter.api.BeforeEach;
import org.junit.jupiter.api.extension.ExtendWith;
import org.springframework.amqp.rabbit.core.RabbitAdmin;
import org.springframework.amqp.rabbit.core.RabbitTemplate;
import org.springframework.amqp.rabbit.listener.RabbitListenerEndpointRegistry;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.boot.autoconfigure.EnableAutoConfiguration;
import org.springframework.boot.autoconfigure.amqp.RabbitAutoConfiguration;
@@ -53,6 +55,11 @@ public abstract class AbstractTest {
@MockBean
private RabbitTemplate rabbitTemplate;
@MockBean
private RabbitAdmin rabbitAdmin;
@MockBean
private RabbitListenerEndpointRegistry rabbitListenerEndpointRegistry;
protected final static String ORIGIN_FILE_ID = "origin";
protected final static String VISUAL_LAYOUT_FILE = "visual";
@@ -39,3 +39,5 @@ management:
prometheus.enabled: ${monitoring.enabled:false}
health.enabled: true
endpoints.web.exposure.include: prometheus, health
POD_NAME: layoutparser-service