Compare commits

...
86 Commits
Author SHA1 Message Date
Kilian Schuettler 8e4cfa2047 RED-10380: wip add table of contents node type 2024-11-04 16:39:08 +01:00
Kilian Schuettler 124afb3623 RED-10249: regex found incorrectly due to wrong text sorting 2024-11-04 12:39:17 +01:00
Kilian Schuettler ca2f3512d2 RED-10249: regex found incorrectly due to wrong text sorting 2024-11-04 12:16:44 +01:00
Kilian Schuettler 7338e06fb0 RED-10249: regex found incorrectly due to wrong text sorting 2024-11-04 11:58:36 +01:00
Kilian Schüttler 4683c696a5 Merge branch 'RED-10247' into 'main'
RED-10247: dictionary entry not found in footer due to wrong text sorting

See merge request fforesight/layout-parser!251
2024-10-25 18:30:35 +02:00
Kilian Schuettler 95c02ce3cf RED-10247: dictionary entry not found in footer due to wrong text sorting 2024-10-25 17:18:14 +02:00
Kilian Schüttler b2d62e32fe Merge branch 'RED-10270-fp' into 'main'
RED-10270: fix NumberFormatException

See merge request fforesight/layout-parser!248
2024-10-24 17:14:47 +02:00
Kilian Schuettler 65c1f03ea3 RED-10270: fix NumberFormatException 2024-10-24 10:59:05 +02:00
Kilian Schüttler 2219519a2b Merge branch 'RED-10127' into 'main'
RED-10127: rename TextPositionSequence to Word

See merge request fforesight/layout-parser!244
2024-10-18 12:20:15 +02:00
Kilian Schüttler af05218e37 RED-10127: rename TextPositionSequence to Word 2024-10-18 12:20:15 +02:00
Kilian Schüttler 736f531df3 Merge branch 'hotfix' into 'main'
Hotfix

See merge request fforesight/layout-parser!243
2024-10-18 12:12:15 +02:00
Kilian Schüttler c64445d54b Hotfix 2024-10-18 12:12:15 +02:00
Kilian Schüttler af29233b10 Merge branch 'feature/RED-10127' into 'main'
RED-10127: add more units

See merge request fforesight/layout-parser!242
2024-10-15 09:57:21 +02:00
Kilian Schuettler 5f04b45554 RED-10127: add more units 2024-10-15 09:47:39 +02:00
Kilian Schüttler 6c41533f0b Merge branch 'feature/RED-10127' into 'main'
RED-10127: improve list classification

See merge request fforesight/layout-parser!240
2024-10-14 17:34:33 +02:00
Kilian Schuettler 9d2596e5ef RED-10127: improve list classification
* add one more format to list identification
* add 'ppb' to known units
* special case for headlines continuing with 14C after the identifier (quite often in some specific files)
2024-10-14 17:21:44 +02:00
Kilian Schüttler e7b01161ac Merge branch 'feature/RED-10127' into 'main'
RED-10127: add list classification

See merge request fforesight/layout-parser!237
2024-10-10 10:50:10 +02:00
Kilian Schüttler 7b073eb4f3 RED-10127: add list classification 2024-10-10 10:50:10 +02:00
Dominique Eifländer 4b0c041d84 Merge branch 'feature/RED-10127' into 'main'
RED-10127: improve headline detection

See merge request fforesight/layout-parser!235
2024-10-09 08:48:48 +02:00
Kilian Schüttler 6c7442ac6d RED-10127: improve headline detection 2024-10-09 08:48:48 +02:00
Maverick Studer 23e23328ee Merge branch 'RED-10126' into 'main'
RM-187: Footers are recognized in the middle of the page

See merge request fforesight/layout-parser!233
2024-10-08 14:27:45 +02:00
Maverick Studer 9d1ffdd779 RM-187: Footers are recognized in the middle of the page 2024-10-08 14:27:44 +02:00
Maverick Studer 3109a30ae1 Merge branch 'RED-9123-proto' into 'main'
RED-9123: Improve performance of re-analysis (Spike)

See merge request fforesight/layout-parser!232
2024-10-07 12:28:10 +02:00
Maverick Studer fe2ed1807e RED-9123: Improve performance of re-analysis (Spike) 2024-10-07 12:28:10 +02:00
Maverick Studer 31de229fa5 Merge branch 'feature/RED-9010' into 'main'
RED-9010: remove redaction log

See merge request fforesight/layout-parser!231
2024-09-19 11:34:32 +02:00
Maverick Studer 8a80abfff1 RED-9010: remove redaction log 2024-09-19 11:34:32 +02:00
Dominique Eifländer 7c08905eda Merge branch 'RED-9975-main' into 'main'
RED-9975: Fixed missing section numbers in layout grid

See merge request fforesight/layout-parser!230
2024-09-18 11:29:51 +02:00
Dominique Eifländer 4f40c9dbc9 RED-9975: Fixed missing section numbers in layout grid 2024-09-18 11:22:37 +02:00
Dominique Eifländer 32381b4472 Merge branch 'RED-9974' into 'main'
Red 9974: improce headline classification, fix font size calculation

See merge request fforesight/layout-parser!226
2024-09-16 14:06:48 +02:00
Kilian Schüttler 469da38952 Red 9974: improce headline classification, fix font size calculation 2024-09-16 14:06:48 +02:00
Dominique Eifländer 0f8c4674b3 Merge branch 'hotfix' into 'main'
hotfix: viewerDocService doesn't remove existing marked content

See merge request fforesight/layout-parser!225
2024-09-12 09:12:54 +02:00
Kilian Schuettler 8e165a41d7 hotfix: viewerDocService doesn't remove existing marked content 2024-09-11 16:34:21 +02:00
Kilian Schüttler ed7a701ad9 Merge branch 'RED-9975' into 'main'
RED-9975: improve SuperSection handling

See merge request fforesight/layout-parser!223
2024-09-11 13:38:09 +02:00
Kilian Schüttler 393103e074 RED-9975: improve SuperSection handling 2024-09-11 13:38:09 +02:00
Dominique Eifländer bd02066e2c Merge branch 'RED-9976-main' into 'main'
RED-9976: Removed sorting that scrambles text in PDFTextStripper

See merge request fforesight/layout-parser!222
2024-09-10 13:02:36 +02:00
Dominique Eifländer fec19f4afb RED-9976: Removed sorting that scrambles text in PDFTextStripper 2024-09-10 12:50:37 +02:00
Kilian Schüttler c726a643f0 Merge branch 'hotfix' into 'main'
Hotfix: unmerge super large tables

See merge request fforesight/layout-parser!220
2024-09-05 15:05:21 +02:00
Kilian Schüttler 519e95735c Hotfix: unmerge super large tables 2024-09-05 15:05:21 +02:00
Maverick Studer b52af2637f Merge branch 'RED-9942-2' into 'main'
RED-9942: File only with images not recognised

See merge request fforesight/layout-parser!218
2024-09-05 10:49:12 +02:00
Maverick Studer 46ea7edc4c RED-9942: File only with images not recognised 2024-09-05 10:49:12 +02:00
Kilian Schüttler 9650195afd Merge branch 'hotfix-fp' into 'main'
hotfix: add Java advanced imaging

See merge request fforesight/layout-parser!217
2024-09-04 15:43:56 +02:00
Kilian Schuettler ce628a99f7 hotfix: add Java advanced imaging 2024-09-04 15:18:12 +02:00
Maverick Studer b66afe135c Merge branch 'RED-9524' into 'main'
RED-9524: File processing does not annotate images

See merge request fforesight/layout-parser!214
2024-09-04 13:27:06 +02:00
Maverick Studer dc892d0fec RED-9524: File processing does not annotate images 2024-09-04 13:27:06 +02:00
Kilian Schüttler af45f2cd8c Merge branch 'RED-9964' into 'main'
RED-9964: fix errors with images

See merge request fforesight/layout-parser!212
2024-09-04 09:16:59 +02:00
Kilian Schuettler befb6b1df6 RED-9964: fix errors with images 2024-09-03 16:37:48 +02:00
Maverick Studer 61efb4cae9 Merge branch 'update-tc' into 'main'
Update tenant-commons for dlq fix

See merge request fforesight/layout-parser!211
2024-09-03 13:50:02 +02:00
maverickstuder 4a06059258 Update tenant-commons for dlq fix 2024-09-03 13:15:08 +02:00
Dominique Eifländer 292e5b215e Merge branch 'RED-9988-main' into 'main'
RED-9988: Fixed NPE when image representation is not present

See merge request fforesight/layout-parser!210
2024-09-02 09:56:53 +02:00
Dominique Eifländer 7c2db6c3c5 RED-9988: Fixed NPE when image representation is not present 2024-09-02 09:51:59 +02:00
Dominique Eifländer 4395074b21 Merge branch 'RED-9975' into 'main'
Red 9975: fix outline detection

See merge request fforesight/layout-parser!206
2024-09-02 09:02:36 +02:00
Kilian Schüttler 8e14b74da2 Red 9975: fix outline detection 2024-09-02 09:02:36 +02:00
Kilian Schüttler 3b91639ea9 Merge branch 'RED-9964-fp' into 'main'
RED-9964: don't merge tables on non-consecutive pages

See merge request fforesight/layout-parser!205
2024-08-30 14:00:48 +02:00
Kilian Schüttler c5178ea5c2 RED-9964: don't merge tables on non-consecutive pages 2024-08-30 14:00:48 +02:00
Dominique Eifländer cf39d4dfcc Merge branch 'RED-9974' into 'main'
RED-9974: Improved headline detection for documine old

See merge request fforesight/layout-parser!202
2024-08-30 10:57:20 +02:00
Dominique Eifländer bb40345f79 RED-9974: Improved headline detection for documine old 2024-08-30 10:36:22 +02:00
Kilian Schüttler e3e9d16145 Merge branch 'RED-9975' into 'main'
RED-9975: activate outline detection

See merge request fforesight/layout-parser!201
2024-08-29 14:27:00 +02:00
Kilian Schuettler f6ca5a3c17 RED-9975: activate outline detection 2024-08-29 14:18:29 +02:00
Maverick Studer 15e3dced35 Merge branch 'tenants-retry' into 'main'
Tenants retry logic and queue renames

See merge request fforesight/layout-parser!197
2024-08-29 13:46:54 +02:00
Maverick Studer 933054b332 Tenants retry logic and queue renames 2024-08-29 13:46:54 +02:00
Kilian Schüttler ab86714cb3 Merge branch 'RED-9975' into 'main'
RED-9975: activate outline detection

See merge request fforesight/layout-parser!198
2024-08-29 12:25:42 +02:00
Kilian Schuettler 8626b106d0 RED-9975: activate outline detection 2024-08-29 12:16:07 +02:00
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
Kilian Schuettler b0ae00aa02 hotfix: threshold adjustements 2024-08-12 14:52:18 +02:00
Dominique Eifländer 00bf9f279e Merge branch 'hotfix-reading-order' into 'main'
hotfix: use center coordinates

See merge request fforesight/layout-parser!191
2024-08-09 15:51:42 +02:00
Kilian Schuettler d16377a24a hotfix: line comparison with center coordinates 2024-08-09 15:45:23 +02:00
Dominique Eifländer 81179ee744 Merge branch 'RED-9760-dcold' into 'main'
RED-9760: Changed lineSeparation threshold for documine old

See merge request fforesight/layout-parser!190
2024-08-09 14:50:39 +02:00
Dominique Eifländer 1953b5924f RED-9760: Changed lineSeparation threshold for documine old 2024-08-09 14:42:14 +02:00
Kilian Schüttler 6f6e8d5d4e Merge branch 'hotfix-reading-order' into 'main'
hotfix reading order

See merge request fforesight/layout-parser!187
2024-08-09 11:49:12 +02:00
Kilian Schüttler 69bcd4f68d hotfix reading order 2024-08-09 11:49:12 +02:00
Timo Bejan b900cfaf31 Merge branch 'CLARI-140' into 'main'
CLARI-140 - case issue

See merge request fforesight/layout-parser!189
2024-08-08 21:49:40 +02:00
Timo Bejan cdc2081785 CLARI-140 - case issue 2024-08-08 22:40:11 +03:00
Timo Bejan a9287ec406 Merge branch 'CLARI-139' into 'main'
CLAR-139 - fixed outline error for unparsable object

See merge request fforesight/layout-parser!188
2024-08-08 16:36:41 +02:00
Timo Bejan 5b6a706c28 CLAR-139 - fixed outline error for unparsable object 2024-08-08 16:20:14 +03:00
Timo Bejan 28d8ad0a3f Merge branch 'CLARI-128' into 'main'
Fixed Index out of bounds exception in blockificationpostprocessingservice -...

See merge request fforesight/layout-parser!186
2024-07-30 16:56:21 +02:00
Timo Bejan 0c1583c1be Fixed Index out of bounds exception in blockificationpostprocessingservice - this could should be documented btw, there are also probably other use-cases where the code doesnt work 2024-07-30 17:45:05 +03:00
Andrei Isvoran 7633566d9b Merge branch 'RED-9607-fp' into 'main'
RED-9607 - Correctly determine text position sequence based on file rotation

See merge request fforesight/layout-parser!184
2024-07-25 14:11:00 +02:00
Andrei Isvoran cc4f09711e RED-9607 - Correctly determine text position sequence based on file rotation 2024-07-24 16:35:11 +03:00
201 changed files with 17249 additions and 2109 deletions
+1
View File
@@ -21,5 +21,6 @@ deploy:
dotenv: version.env
rules:
- if: $CI_COMMIT_BRANCH == $CI_DEFAULT_BRANCH
- if: $CI_COMMIT_BRANCH =~ /^feature/ && $CI_COMMIT_TAG == ""
- if: $CI_COMMIT_BRANCH =~ /^release/
- if: $CI_COMMIT_TAG
@@ -42,6 +42,19 @@ tasks.jacocoTestReport {
}
allprojects {
tasks.withType<Javadoc> {
options {
this as StandardJavadocDocletOptions
addBooleanOption("Xdoclint:none", true)
addStringOption("Xmaxwarns", "1")
}
}
pmd {
setConsoleOutput(true)
}
publishing {
publications {
create<MavenPublication>(name) {
@@ -64,6 +77,7 @@ java {
withJavadocJar()
}
repositories {
mavenLocal()
mavenCentral()
@@ -7,4 +7,5 @@ description = "layoutparser-service-internal-api"
dependencies {
implementation("io.swagger.core.v3:swagger-annotations:2.2.15")
implementation("com.google.protobuf:protobuf-java-util:4.27.1")
}
@@ -2,6 +2,11 @@ package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
import java.io.Serializable;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentPageProto.AllDocumentPages;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentPositionDataProto.AllDocumentPositionData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentStructureProto.DocumentStructure;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentTextDataProto.AllDocumentTextData;
import io.swagger.v3.oas.annotations.media.Schema;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
@@ -13,16 +18,26 @@ import lombok.experimental.FieldDefaults;
@Builder
@AllArgsConstructor
@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
@Schema(description = "Object containing the complete document layout parsing information. It is split into 4 categories, structure, text, positions and pages: " + "The document tree structure of SemanticNodes such as Section, Paragraph, Headline, etc. " + "The text, which is stored as separate blocks of data. " + "The text positions, which are also stored as separate blocks. The Blocks are equal to the text blocks in length and order. " + "The page information.")
@Schema(description = "Object containing the complete document layout parsing information. It is split into 4 categories, structure, text, positions and pages: "
+ "The document tree structure of SemanticNodes such as Section, Paragraph, Headline, etc. "
+ "The text, which is stored as separate blocks of data. "
+ "The text positions, which are also stored as separate blocks. The Blocks are equal to the text blocks in length and order. "
+ "The page information.")
public class DocumentData implements Serializable {
@Schema(description = "Contains information about the document's pages.")
DocumentPage[] documentPages;
AllDocumentPages documentPages;
@Schema(description = "Contains information about the document's text.")
DocumentTextData[] documentTextData;
AllDocumentTextData documentTextData;
@Schema(description = "Contains information about the document's text positions.")
DocumentPositionData[] documentPositions;
AllDocumentPositionData documentPositions;
@Schema(description = "Contains information about the document's semantic structure.")
DocumentStructure documentStructure;
DocumentStructureWrapper documentStructureWrapper;
public DocumentStructure getDocumentStructure() {
return documentStructureWrapper.getDocumentStructure();
}
}
@@ -10,6 +10,7 @@ import lombok.Data;
import lombok.NoArgsConstructor;
import lombok.experimental.FieldDefaults;
@Deprecated
@Data
@Builder
@NoArgsConstructor
@@ -10,6 +10,7 @@ import lombok.Data;
import lombok.NoArgsConstructor;
import lombok.experimental.FieldDefaults;
@Deprecated
@Data
@Builder
@NoArgsConstructor
@@ -16,6 +16,7 @@ import lombok.Data;
import lombok.NoArgsConstructor;
import lombok.experimental.FieldDefaults;
@Deprecated
@Data
@Builder
@NoArgsConstructor
@@ -0,0 +1,799 @@
package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
// Generated by the protocol buffer compiler. DO NOT EDIT!
// NO CHECKED-IN PROTOBUF GENCODE
// source: DocumentStructure.proto
// Protobuf Java Version: 4.27.1
@SuppressWarnings("all")
public final class DocumentStructureProto {
private DocumentStructureProto() {}
static {
com.google.protobuf.RuntimeVersion.validateProtobufGencodeVersion(com.google.protobuf.RuntimeVersion.RuntimeDomain.PUBLIC,
/* major= */ 4,
/* minor= */ 27,
/* patch= */ 1,
/* suffix= */ "", DocumentStructureProto.class.getName());
}
public static void registerAllExtensions(com.google.protobuf.ExtensionRegistryLite registry) {
}
public static void registerAllExtensions(com.google.protobuf.ExtensionRegistry registry) {
registerAllExtensions((com.google.protobuf.ExtensionRegistryLite) registry);
}
public interface DocumentStructureOrBuilder extends
// @@protoc_insertion_point(interface_extends:DocumentStructure)
com.google.protobuf.MessageOrBuilder {
/**
* <pre>
* The root EntryData represents the Document.
* </pre>
*
* <code>.EntryData root = 1;</code>
*
* @return Whether the root field is set.
*/
boolean hasRoot();
/**
* <pre>
* The root EntryData represents the Document.
* </pre>
*
* <code>.EntryData root = 1;</code>
*
* @return The root.
*/
EntryDataProto.EntryData getRoot();
/**
* <pre>
* The root EntryData represents the Document.
* </pre>
*
* <code>.EntryData root = 1;</code>
*/
EntryDataProto.EntryDataOrBuilder getRootOrBuilder();
}
/**
* Protobuf type {@code DocumentStructure}
*/
public static final class DocumentStructure extends com.google.protobuf.GeneratedMessage implements
// @@protoc_insertion_point(message_implements:DocumentStructure)
DocumentStructureOrBuilder {
private static final long serialVersionUID = 0L;
static {
com.google.protobuf.RuntimeVersion.validateProtobufGencodeVersion(com.google.protobuf.RuntimeVersion.RuntimeDomain.PUBLIC,
/* major= */ 4,
/* minor= */ 27,
/* patch= */ 1,
/* suffix= */ "", DocumentStructure.class.getName());
}
// Use DocumentStructure.newBuilder() to construct.
private DocumentStructure(com.google.protobuf.GeneratedMessage.Builder<?> builder) {
super(builder);
}
private DocumentStructure() {
}
public static final com.google.protobuf.Descriptors.Descriptor getDescriptor() {
return DocumentStructureProto.internal_static_DocumentStructure_descriptor;
}
@Override
protected FieldAccessorTable internalGetFieldAccessorTable() {
return DocumentStructureProto.internal_static_DocumentStructure_fieldAccessorTable.ensureFieldAccessorsInitialized(DocumentStructure.class, Builder.class);
}
private int bitField0_;
public static final int ROOT_FIELD_NUMBER = 1;
private EntryDataProto.EntryData root_;
/**
* <pre>
* The root EntryData represents the Document.
* </pre>
*
* <code>.EntryData root = 1;</code>
*
* @return Whether the root field is set.
*/
@Override
public boolean hasRoot() {
return ((bitField0_ & 0x00000001) != 0);
}
/**
* <pre>
* The root EntryData represents the Document.
* </pre>
*
* <code>.EntryData root = 1;</code>
*
* @return The root.
*/
@Override
public EntryDataProto.EntryData getRoot() {
return root_ == null ? EntryDataProto.EntryData.getDefaultInstance() : root_;
}
/**
* <pre>
* The root EntryData represents the Document.
* </pre>
*
* <code>.EntryData root = 1;</code>
*/
@Override
public EntryDataProto.EntryDataOrBuilder getRootOrBuilder() {
return root_ == null ? EntryDataProto.EntryData.getDefaultInstance() : root_;
}
private byte memoizedIsInitialized = -1;
@Override
public final boolean isInitialized() {
byte isInitialized = memoizedIsInitialized;
if (isInitialized == 1) {
return true;
}
if (isInitialized == 0) {
return false;
}
memoizedIsInitialized = 1;
return true;
}
@Override
public void writeTo(com.google.protobuf.CodedOutputStream output) throws java.io.IOException {
if (((bitField0_ & 0x00000001) != 0)) {
output.writeMessage(1, getRoot());
}
getUnknownFields().writeTo(output);
}
@Override
public int getSerializedSize() {
int size = memoizedSize;
if (size != -1) {
return size;
}
size = 0;
if (((bitField0_ & 0x00000001) != 0)) {
size += com.google.protobuf.CodedOutputStream.computeMessageSize(1, getRoot());
}
size += getUnknownFields().getSerializedSize();
memoizedSize = size;
return size;
}
@Override
public boolean equals(final Object obj) {
if (obj == this) {
return true;
}
if (!(obj instanceof DocumentStructure)) {
return super.equals(obj);
}
DocumentStructure other = (DocumentStructure) obj;
if (hasRoot() != other.hasRoot()) {
return false;
}
if (hasRoot()) {
if (!getRoot().equals(other.getRoot())) {
return false;
}
}
if (!getUnknownFields().equals(other.getUnknownFields())) {
return false;
}
return true;
}
@Override
public int hashCode() {
if (memoizedHashCode != 0) {
return memoizedHashCode;
}
int hash = 41;
hash = (19 * hash) + getDescriptor().hashCode();
if (hasRoot()) {
hash = (37 * hash) + ROOT_FIELD_NUMBER;
hash = (53 * hash) + getRoot().hashCode();
}
hash = (29 * hash) + getUnknownFields().hashCode();
memoizedHashCode = hash;
return hash;
}
public static DocumentStructure parseFrom(java.nio.ByteBuffer data) throws com.google.protobuf.InvalidProtocolBufferException {
return PARSER.parseFrom(data);
}
public static DocumentStructure parseFrom(java.nio.ByteBuffer data,
com.google.protobuf.ExtensionRegistryLite extensionRegistry) throws com.google.protobuf.InvalidProtocolBufferException {
return PARSER.parseFrom(data, extensionRegistry);
}
public static DocumentStructure parseFrom(com.google.protobuf.ByteString data) throws com.google.protobuf.InvalidProtocolBufferException {
return PARSER.parseFrom(data);
}
public static DocumentStructure parseFrom(com.google.protobuf.ByteString data,
com.google.protobuf.ExtensionRegistryLite extensionRegistry) throws com.google.protobuf.InvalidProtocolBufferException {
return PARSER.parseFrom(data, extensionRegistry);
}
public static DocumentStructure parseFrom(byte[] data) throws com.google.protobuf.InvalidProtocolBufferException {
return PARSER.parseFrom(data);
}
public static DocumentStructure parseFrom(byte[] data,
com.google.protobuf.ExtensionRegistryLite extensionRegistry) throws com.google.protobuf.InvalidProtocolBufferException {
return PARSER.parseFrom(data, extensionRegistry);
}
public static DocumentStructure parseFrom(java.io.InputStream input) throws java.io.IOException {
return com.google.protobuf.GeneratedMessage.parseWithIOException(PARSER, input);
}
public static DocumentStructure parseFrom(java.io.InputStream input, com.google.protobuf.ExtensionRegistryLite extensionRegistry) throws java.io.IOException {
return com.google.protobuf.GeneratedMessage.parseWithIOException(PARSER, input, extensionRegistry);
}
public static DocumentStructure parseDelimitedFrom(java.io.InputStream input) throws java.io.IOException {
return com.google.protobuf.GeneratedMessage.parseDelimitedWithIOException(PARSER, input);
}
public static DocumentStructure parseDelimitedFrom(java.io.InputStream input, com.google.protobuf.ExtensionRegistryLite extensionRegistry) throws java.io.IOException {
return com.google.protobuf.GeneratedMessage.parseDelimitedWithIOException(PARSER, input, extensionRegistry);
}
public static DocumentStructure parseFrom(com.google.protobuf.CodedInputStream input) throws java.io.IOException {
return com.google.protobuf.GeneratedMessage.parseWithIOException(PARSER, input);
}
public static DocumentStructure parseFrom(com.google.protobuf.CodedInputStream input,
com.google.protobuf.ExtensionRegistryLite extensionRegistry) throws java.io.IOException {
return com.google.protobuf.GeneratedMessage.parseWithIOException(PARSER, input, extensionRegistry);
}
@Override
public Builder newBuilderForType() {return newBuilder();}
public static Builder newBuilder() {
return DEFAULT_INSTANCE.toBuilder();
}
public static Builder newBuilder(DocumentStructure prototype) {
return DEFAULT_INSTANCE.toBuilder().mergeFrom(prototype);
}
@Override
public Builder toBuilder() {
return this == DEFAULT_INSTANCE ? new Builder() : new Builder().mergeFrom(this);
}
@Override
protected Builder newBuilderForType(BuilderParent parent) {
Builder builder = new Builder(parent);
return builder;
}
/**
* Protobuf type {@code DocumentStructure}
*/
public static final class Builder extends com.google.protobuf.GeneratedMessage.Builder<Builder> implements
// @@protoc_insertion_point(builder_implements:DocumentStructure)
DocumentStructureOrBuilder {
public static final com.google.protobuf.Descriptors.Descriptor getDescriptor() {
return DocumentStructureProto.internal_static_DocumentStructure_descriptor;
}
@Override
protected FieldAccessorTable internalGetFieldAccessorTable() {
return DocumentStructureProto.internal_static_DocumentStructure_fieldAccessorTable.ensureFieldAccessorsInitialized(DocumentStructure.class, Builder.class);
}
// Construct using DocumentStructureOuterClass.DocumentStructure.newBuilder()
private Builder() {
maybeForceBuilderInitialization();
}
private Builder(BuilderParent parent) {
super(parent);
maybeForceBuilderInitialization();
}
private void maybeForceBuilderInitialization() {
if (com.google.protobuf.GeneratedMessage.alwaysUseFieldBuilders) {
getRootFieldBuilder();
}
}
@Override
public Builder clear() {
super.clear();
bitField0_ = 0;
root_ = null;
if (rootBuilder_ != null) {
rootBuilder_.dispose();
rootBuilder_ = null;
}
return this;
}
@Override
public com.google.protobuf.Descriptors.Descriptor getDescriptorForType() {
return DocumentStructureProto.internal_static_DocumentStructure_descriptor;
}
@Override
public DocumentStructure getDefaultInstanceForType() {
return DocumentStructure.getDefaultInstance();
}
@Override
public DocumentStructure build() {
DocumentStructure result = buildPartial();
if (!result.isInitialized()) {
throw newUninitializedMessageException(result);
}
return result;
}
@Override
public DocumentStructure buildPartial() {
DocumentStructure result = new DocumentStructure(this);
if (bitField0_ != 0) {
buildPartial0(result);
}
onBuilt();
return result;
}
private void buildPartial0(DocumentStructure result) {
int from_bitField0_ = bitField0_;
int to_bitField0_ = 0;
if (((from_bitField0_ & 0x00000001) != 0)) {
result.root_ = rootBuilder_ == null ? root_ : rootBuilder_.build();
to_bitField0_ |= 0x00000001;
}
result.bitField0_ |= to_bitField0_;
}
@Override
public Builder mergeFrom(com.google.protobuf.Message other) {
if (other instanceof DocumentStructure) {
return mergeFrom((DocumentStructure) other);
} else {
super.mergeFrom(other);
return this;
}
}
public Builder mergeFrom(DocumentStructure other) {
if (other == DocumentStructure.getDefaultInstance()) {
return this;
}
if (other.hasRoot()) {
mergeRoot(other.getRoot());
}
this.mergeUnknownFields(other.getUnknownFields());
onChanged();
return this;
}
@Override
public final boolean isInitialized() {
return true;
}
@Override
public Builder mergeFrom(com.google.protobuf.CodedInputStream input, com.google.protobuf.ExtensionRegistryLite extensionRegistry) throws java.io.IOException {
if (extensionRegistry == null) {
throw new NullPointerException();
}
try {
boolean done = false;
while (!done) {
int tag = input.readTag();
switch (tag) {
case 0:
done = true;
break;
case 10: {
input.readMessage(getRootFieldBuilder().getBuilder(), extensionRegistry);
bitField0_ |= 0x00000001;
break;
} // case 10
default: {
if (!super.parseUnknownField(input, extensionRegistry, tag)) {
done = true; // was an endgroup tag
}
break;
} // default:
} // switch (tag)
} // while (!done)
} catch (com.google.protobuf.InvalidProtocolBufferException e) {
throw e.unwrapIOException();
} finally {
onChanged();
} // finally
return this;
}
private int bitField0_;
private EntryDataProto.EntryData root_;
private com.google.protobuf.SingleFieldBuilder<EntryDataProto.EntryData, EntryDataProto.EntryData.Builder, EntryDataProto.EntryDataOrBuilder> rootBuilder_;
/**
* <pre>
* The root EntryData represents the Document.
* </pre>
*
* <code>.EntryData root = 1;</code>
*
* @return Whether the root field is set.
*/
public boolean hasRoot() {
return ((bitField0_ & 0x00000001) != 0);
}
/**
* <pre>
* The root EntryData represents the Document.
* </pre>
*
* <code>.EntryData root = 1;</code>
*
* @return The root.
*/
public EntryDataProto.EntryData getRoot() {
if (rootBuilder_ == null) {
return root_ == null ? EntryDataProto.EntryData.getDefaultInstance() : root_;
} else {
return rootBuilder_.getMessage();
}
}
/**
* <pre>
* The root EntryData represents the Document.
* </pre>
*
* <code>.EntryData root = 1;</code>
*/
public Builder setRoot(EntryDataProto.EntryData value) {
if (rootBuilder_ == null) {
if (value == null) {
throw new NullPointerException();
}
root_ = value;
} else {
rootBuilder_.setMessage(value);
}
bitField0_ |= 0x00000001;
onChanged();
return this;
}
/**
* <pre>
* The root EntryData represents the Document.
* </pre>
*
* <code>.EntryData root = 1;</code>
*/
public Builder setRoot(EntryDataProto.EntryData.Builder builderForValue) {
if (rootBuilder_ == null) {
root_ = builderForValue.build();
} else {
rootBuilder_.setMessage(builderForValue.build());
}
bitField0_ |= 0x00000001;
onChanged();
return this;
}
/**
* <pre>
* The root EntryData represents the Document.
* </pre>
*
* <code>.EntryData root = 1;</code>
*/
public Builder mergeRoot(EntryDataProto.EntryData value) {
if (rootBuilder_ == null) {
if (((bitField0_ & 0x00000001) != 0) && root_ != null && root_ != EntryDataProto.EntryData.getDefaultInstance()) {
getRootBuilder().mergeFrom(value);
} else {
root_ = value;
}
} else {
rootBuilder_.mergeFrom(value);
}
if (root_ != null) {
bitField0_ |= 0x00000001;
onChanged();
}
return this;
}
/**
* <pre>
* The root EntryData represents the Document.
* </pre>
*
* <code>.EntryData root = 1;</code>
*/
public Builder clearRoot() {
bitField0_ = (bitField0_ & ~0x00000001);
root_ = null;
if (rootBuilder_ != null) {
rootBuilder_.dispose();
rootBuilder_ = null;
}
onChanged();
return this;
}
/**
* <pre>
* The root EntryData represents the Document.
* </pre>
*
* <code>.EntryData root = 1;</code>
*/
public EntryDataProto.EntryData.Builder getRootBuilder() {
bitField0_ |= 0x00000001;
onChanged();
return getRootFieldBuilder().getBuilder();
}
/**
* <pre>
* The root EntryData represents the Document.
* </pre>
*
* <code>.EntryData root = 1;</code>
*/
public EntryDataProto.EntryDataOrBuilder getRootOrBuilder() {
if (rootBuilder_ != null) {
return rootBuilder_.getMessageOrBuilder();
} else {
return root_ == null ? EntryDataProto.EntryData.getDefaultInstance() : root_;
}
}
/**
* <pre>
* The root EntryData represents the Document.
* </pre>
*
* <code>.EntryData root = 1;</code>
*/
private com.google.protobuf.SingleFieldBuilder<EntryDataProto.EntryData, EntryDataProto.EntryData.Builder, EntryDataProto.EntryDataOrBuilder> getRootFieldBuilder() {
if (rootBuilder_ == null) {
rootBuilder_ = new com.google.protobuf.SingleFieldBuilder<EntryDataProto.EntryData, EntryDataProto.EntryData.Builder, EntryDataProto.EntryDataOrBuilder>(getRoot(),
getParentForChildren(),
isClean());
root_ = null;
}
return rootBuilder_;
}
// @@protoc_insertion_point(builder_scope:DocumentStructure)
}
// @@protoc_insertion_point(class_scope:DocumentStructure)
private static final DocumentStructure DEFAULT_INSTANCE;
static {
DEFAULT_INSTANCE = new DocumentStructure();
}
public static DocumentStructure getDefaultInstance() {
return DEFAULT_INSTANCE;
}
private static final com.google.protobuf.Parser<DocumentStructure> PARSER = new com.google.protobuf.AbstractParser<DocumentStructure>() {
@Override
public DocumentStructure parsePartialFrom(com.google.protobuf.CodedInputStream input,
com.google.protobuf.ExtensionRegistryLite extensionRegistry) throws com.google.protobuf.InvalidProtocolBufferException {
Builder builder = newBuilder();
try {
builder.mergeFrom(input, extensionRegistry);
} catch (com.google.protobuf.InvalidProtocolBufferException e) {
throw e.setUnfinishedMessage(builder.buildPartial());
} catch (com.google.protobuf.UninitializedMessageException e) {
throw e.asInvalidProtocolBufferException().setUnfinishedMessage(builder.buildPartial());
} catch (java.io.IOException e) {
throw new com.google.protobuf.InvalidProtocolBufferException(e).setUnfinishedMessage(builder.buildPartial());
}
return builder.buildPartial();
}
};
public static com.google.protobuf.Parser<DocumentStructure> parser() {
return PARSER;
}
@Override
public com.google.protobuf.Parser<DocumentStructure> getParserForType() {
return PARSER;
}
@Override
public DocumentStructure getDefaultInstanceForType() {
return DEFAULT_INSTANCE;
}
}
private static final com.google.protobuf.Descriptors.Descriptor internal_static_DocumentStructure_descriptor;
private static final com.google.protobuf.GeneratedMessage.FieldAccessorTable internal_static_DocumentStructure_fieldAccessorTable;
public static com.google.protobuf.Descriptors.FileDescriptor getDescriptor() {
return descriptor;
}
private static com.google.protobuf.Descriptors.FileDescriptor descriptor;
static {
String[] descriptorData = {"\n\027DocumentStructure.proto\032\017EntryData.pro"
+ "to\"-\n\021DocumentStructure\022\030\n\004root\030\001 \001(\0132\n."
+ "EntryDatab\006proto3"};
descriptor = com.google.protobuf.Descriptors.FileDescriptor.internalBuildGeneratedFileFrom(descriptorData,
new com.google.protobuf.Descriptors.FileDescriptor[]{EntryDataProto.getDescriptor(),});
internal_static_DocumentStructure_descriptor = getDescriptor().getMessageTypes()
.get(0);
internal_static_DocumentStructure_fieldAccessorTable = new com.google.protobuf.GeneratedMessage.FieldAccessorTable(internal_static_DocumentStructure_descriptor,
new String[]{"Root",});
descriptor.resolveAllFeaturesImmutable();
EntryDataProto.getDescriptor();
}
// @@protoc_insertion_point(outer_class_scope)
}
@@ -0,0 +1,126 @@
package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
import static com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentStructureProto.DocumentStructure;
import java.awt.geom.Rectangle2D;
import java.io.ObjectStreamException;
import java.io.Serializable;
import java.util.Arrays;
import java.util.List;
import java.util.stream.Stream;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.EntryDataProto.EntryData;
import io.swagger.v3.oas.annotations.media.Schema;
import lombok.AllArgsConstructor;
import lombok.Getter;
@Getter
@AllArgsConstructor
public class DocumentStructureWrapper implements Serializable {
private final DocumentStructure documentStructure;
@Schema(description = "Object containing the extra field names, a table has in its properties field.")
public static class TableProperties implements Serializable {
public static final String NUMBER_OF_ROWS = "numberOfRows";
public static final String NUMBER_OF_COLS = "numberOfCols";
}
@Schema(description = "Object containing the extra field names, an Image has in its properties field.")
public static class ImageProperties implements Serializable {
public static final String TRANSPARENT = "transparent";
public static final String IMAGE_TYPE = "imageType";
public static final String POSITION = "position";
public static final String ID = "id";
public static final String REPRESENTATION_HASH = "representationHash";
}
@Schema(description = "Object containing the extra field names, a table cell has in its properties field.")
public static class TableCellProperties implements Serializable {
public static final String B_BOX = "bBox";
public static final String ROW = "row";
public static final String COL = "col";
public static final String HEADER = "header";
}
@Schema(description = "Object containing the extra field names, a duplicate paragraph has in its properties field.")
public static class DuplicateParagraphProperties implements Serializable {
public static final String UNSORTED_TEXTBLOCK_ID = "utbid";
}
public static final String RECTANGLE_DELIMITER = ";";
public static Rectangle2D parseRectangle2D(String bBox) {
List<Float> floats = Arrays.stream(bBox.split(RECTANGLE_DELIMITER))
.map(Float::parseFloat)
.toList();
return new Rectangle2D.Float(floats.get(0), floats.get(1), floats.get(2), floats.get(3));
}
public static double[] parseRepresentationVector(String representationHash) {
String[] stringArray = representationHash.split("[,\\s]+");
double[] doubleArray = new double[stringArray.length];
for (int i = 0; i < stringArray.length; i++) {
doubleArray[i] = Double.parseDouble(stringArray[i]);
}
return doubleArray;
}
public EntryData get(List<Integer> tocId) {
if (tocId.isEmpty()) {
return documentStructure.getRoot();
}
EntryData entry = documentStructure.getRoot().getChildrenList()
.get(tocId.get(0));
for (int id : tocId.subList(1, tocId.size())) {
entry = entry.getChildrenList()
.get(id);
}
return entry;
}
public Stream<EntryData> streamAllEntries() {
return Stream.concat(Stream.of(documentStructure.getRoot()),
documentStructure.getRoot().getChildrenList()
.stream())
.flatMap(DocumentStructureWrapper::flatten);
}
public String toString() {
return String.join("\n",
streamAllEntries().map(EntryData::toString)
.toList());
}
private static Stream<EntryData> flatten(EntryData entry) {
return Stream.concat(Stream.of(entry),
entry.getChildrenList()
.stream()
.flatMap(DocumentStructureWrapper::flatten));
}
}
@@ -10,6 +10,7 @@ import lombok.Data;
import lombok.NoArgsConstructor;
import lombok.experimental.FieldDefaults;
@Deprecated
@Data
@Builder
@NoArgsConstructor
@@ -1,5 +1,6 @@
package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
@Deprecated
public enum LayoutEngine {
ALGORITHM,
AI,
@@ -0,0 +1,193 @@
package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
// Generated by the protocol buffer compiler. DO NOT EDIT!
// NO CHECKED-IN PROTOBUF GENCODE
// source: LayoutEngine.proto
// Protobuf Java Version: 4.27.1
@SuppressWarnings("all")
public final class LayoutEngineProto {
private LayoutEngineProto() {}
static {
com.google.protobuf.RuntimeVersion.validateProtobufGencodeVersion(com.google.protobuf.RuntimeVersion.RuntimeDomain.PUBLIC,
/* major= */ 4,
/* minor= */ 27,
/* patch= */ 1,
/* suffix= */ "", LayoutEngineProto.class.getName());
}
public static void registerAllExtensions(com.google.protobuf.ExtensionRegistryLite registry) {
}
public static void registerAllExtensions(com.google.protobuf.ExtensionRegistry registry) {
registerAllExtensions((com.google.protobuf.ExtensionRegistryLite) registry);
}
/**
* Protobuf enum {@code LayoutEngine}
*/
public enum LayoutEngine implements com.google.protobuf.ProtocolMessageEnum {
/**
* <code>ALGORITHM = 0;</code>
*/
ALGORITHM(0),
/**
* <code>AI = 1;</code>
*/
AI(1),
/**
* <code>OUTLINE = 2;</code>
*/
OUTLINE(2),
UNRECOGNIZED(-1),
;
static {
com.google.protobuf.RuntimeVersion.validateProtobufGencodeVersion(com.google.protobuf.RuntimeVersion.RuntimeDomain.PUBLIC,
/* major= */ 4,
/* minor= */ 27,
/* patch= */ 1,
/* suffix= */ "", LayoutEngine.class.getName());
}
/**
* <code>ALGORITHM = 0;</code>
*/
public static final int ALGORITHM_VALUE = 0;
/**
* <code>AI = 1;</code>
*/
public static final int AI_VALUE = 1;
/**
* <code>OUTLINE = 2;</code>
*/
public static final int OUTLINE_VALUE = 2;
public final int getNumber() {
if (this == UNRECOGNIZED) {
throw new IllegalArgumentException("Can't get the number of an unknown enum value.");
}
return value;
}
/**
* @param value The numeric wire value of the corresponding enum entry.
* @return The enum associated with the given numeric wire value.
* @deprecated Use {@link #forNumber(int)} instead.
*/
@Deprecated
public static LayoutEngine valueOf(int value) {
return forNumber(value);
}
/**
* @param value The numeric wire value of the corresponding enum entry.
* @return The enum associated with the given numeric wire value.
*/
public static LayoutEngine forNumber(int value) {
switch (value) {
case 0:
return ALGORITHM;
case 1:
return AI;
case 2:
return OUTLINE;
default:
return null;
}
}
public static com.google.protobuf.Internal.EnumLiteMap<LayoutEngine> internalGetValueMap() {
return internalValueMap;
}
private static final com.google.protobuf.Internal.EnumLiteMap<LayoutEngine> internalValueMap = new com.google.protobuf.Internal.EnumLiteMap<LayoutEngine>() {
public LayoutEngine findValueByNumber(int number) {
return LayoutEngine.forNumber(number);
}
};
public final com.google.protobuf.Descriptors.EnumValueDescriptor getValueDescriptor() {
if (this == UNRECOGNIZED) {
throw new IllegalStateException("Can't get the descriptor of an unrecognized enum value.");
}
return getDescriptor().getValues()
.get(ordinal());
}
public final com.google.protobuf.Descriptors.EnumDescriptor getDescriptorForType() {
return getDescriptor();
}
public static final com.google.protobuf.Descriptors.EnumDescriptor getDescriptor() {
return LayoutEngineProto.getDescriptor().getEnumTypes()
.get(0);
}
private static final LayoutEngine[] VALUES = values();
public static LayoutEngine valueOf(com.google.protobuf.Descriptors.EnumValueDescriptor desc) {
if (desc.getType() != getDescriptor()) {
throw new IllegalArgumentException("EnumValueDescriptor is not for this type.");
}
if (desc.getIndex() == -1) {
return UNRECOGNIZED;
}
return VALUES[desc.getIndex()];
}
private final int value;
private LayoutEngine(int value) {
this.value = value;
}
// @@protoc_insertion_point(enum_scope:LayoutEngine)
}
public static com.google.protobuf.Descriptors.FileDescriptor getDescriptor() {
return descriptor;
}
private static com.google.protobuf.Descriptors.FileDescriptor descriptor;
static {
String[] descriptorData = {"\n\022LayoutEngine.proto*2\n\014LayoutEngine\022\r\n\t" + "ALGORITHM\020\000\022\006\n\002AI\020\001\022\013\n\007OUTLINE\020\002b\006proto3"};
descriptor = com.google.protobuf.Descriptors.FileDescriptor.internalBuildGeneratedFileFrom(descriptorData, new com.google.protobuf.Descriptors.FileDescriptor[]{});
descriptor.resolveAllFeaturesImmutable();
}
// @@protoc_insertion_point(outer_class_scope)
}
@@ -3,6 +3,7 @@ package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
import java.io.Serializable;
import java.util.Locale;
@Deprecated
public enum NodeType implements Serializable {
DOCUMENT,
SECTION,
@@ -0,0 +1,274 @@
package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
import java.util.Locale;
// Generated by the protocol buffer compiler. DO NOT EDIT!
// NO CHECKED-IN PROTOBUF GENCODE
// source: NodeType.proto
// Protobuf Java Version: 4.27.1
@SuppressWarnings("all")
public final class NodeTypeProto {
private NodeTypeProto() {}
static {
com.google.protobuf.RuntimeVersion.validateProtobufGencodeVersion(com.google.protobuf.RuntimeVersion.RuntimeDomain.PUBLIC,
/* major= */ 4,
/* minor= */ 27,
/* patch= */ 1,
/* suffix= */ "", NodeTypeProto.class.getName());
}
public static void registerAllExtensions(com.google.protobuf.ExtensionRegistryLite registry) {
}
public static void registerAllExtensions(com.google.protobuf.ExtensionRegistry registry) {
registerAllExtensions((com.google.protobuf.ExtensionRegistryLite) registry);
}
/**
* Protobuf enum {@code NodeType}
*/
public enum NodeType implements com.google.protobuf.ProtocolMessageEnum {
/**
* <code>DOCUMENT = 0;</code>
*/
DOCUMENT(0),
/**
* <code>SECTION = 1;</code>
*/
SECTION(1),
/**
* <code>SUPER_SECTION = 2;</code>
*/
SUPER_SECTION(2),
/**
* <code>HEADLINE = 3;</code>
*/
HEADLINE(3),
/**
* <code>PARAGRAPH = 4;</code>
*/
PARAGRAPH(4),
/**
* <code>TABLE = 5;</code>
*/
TABLE(5),
/**
* <code>TABLE_CELL = 6;</code>
*/
TABLE_CELL(6),
/**
* <code>IMAGE = 7;</code>
*/
IMAGE(7),
/**
* <code>HEADER = 8;</code>
*/
HEADER(8),
/**
* <code>FOOTER = 9;</code>
*/
FOOTER(9),
UNRECOGNIZED(-1),
;
public String toString() {
return this.name().charAt(0) + this.name().substring(1).toLowerCase(Locale.ROOT);
}
static {
com.google.protobuf.RuntimeVersion.validateProtobufGencodeVersion(com.google.protobuf.RuntimeVersion.RuntimeDomain.PUBLIC,
/* major= */ 4,
/* minor= */ 27,
/* patch= */ 1,
/* suffix= */ "", NodeType.class.getName());
}
/**
* <code>DOCUMENT = 0;</code>
*/
public static final int DOCUMENT_VALUE = 0;
/**
* <code>SECTION = 1;</code>
*/
public static final int SECTION_VALUE = 1;
/**
* <code>SUPER_SECTION = 2;</code>
*/
public static final int SUPER_SECTION_VALUE = 2;
/**
* <code>HEADLINE = 3;</code>
*/
public static final int HEADLINE_VALUE = 3;
/**
* <code>PARAGRAPH = 4;</code>
*/
public static final int PARAGRAPH_VALUE = 4;
/**
* <code>TABLE = 5;</code>
*/
public static final int TABLE_VALUE = 5;
/**
* <code>TABLE_CELL = 6;</code>
*/
public static final int TABLE_CELL_VALUE = 6;
/**
* <code>IMAGE = 7;</code>
*/
public static final int IMAGE_VALUE = 7;
/**
* <code>HEADER = 8;</code>
*/
public static final int HEADER_VALUE = 8;
/**
* <code>FOOTER = 9;</code>
*/
public static final int FOOTER_VALUE = 9;
public final int getNumber() {
if (this == UNRECOGNIZED) {
throw new IllegalArgumentException("Can't get the number of an unknown enum value.");
}
return value;
}
/**
* @param value The numeric wire value of the corresponding enum entry.
* @return The enum associated with the given numeric wire value.
* @deprecated Use {@link #forNumber(int)} instead.
*/
@Deprecated
public static NodeType valueOf(int value) {
return forNumber(value);
}
/**
* @param value The numeric wire value of the corresponding enum entry.
* @return The enum associated with the given numeric wire value.
*/
public static NodeType forNumber(int value) {
switch (value) {
case 0:
return DOCUMENT;
case 1:
return SECTION;
case 2:
return SUPER_SECTION;
case 3:
return HEADLINE;
case 4:
return PARAGRAPH;
case 5:
return TABLE;
case 6:
return TABLE_CELL;
case 7:
return IMAGE;
case 8:
return HEADER;
case 9:
return FOOTER;
default:
return null;
}
}
public static com.google.protobuf.Internal.EnumLiteMap<NodeType> internalGetValueMap() {
return internalValueMap;
}
private static final com.google.protobuf.Internal.EnumLiteMap<NodeType> internalValueMap = new com.google.protobuf.Internal.EnumLiteMap<NodeType>() {
public NodeType findValueByNumber(int number) {
return NodeType.forNumber(number);
}
};
public final com.google.protobuf.Descriptors.EnumValueDescriptor getValueDescriptor() {
if (this == UNRECOGNIZED) {
throw new IllegalStateException("Can't get the descriptor of an unrecognized enum value.");
}
return getDescriptor().getValues()
.get(ordinal());
}
public final com.google.protobuf.Descriptors.EnumDescriptor getDescriptorForType() {
return getDescriptor();
}
public static final com.google.protobuf.Descriptors.EnumDescriptor getDescriptor() {
return NodeTypeProto.getDescriptor().getEnumTypes()
.get(0);
}
private static final NodeType[] VALUES = values();
public static NodeType valueOf(com.google.protobuf.Descriptors.EnumValueDescriptor desc) {
if (desc.getType() != getDescriptor()) {
throw new IllegalArgumentException("EnumValueDescriptor is not for this type.");
}
if (desc.getIndex() == -1) {
return UNRECOGNIZED;
}
return VALUES[desc.getIndex()];
}
private final int value;
private NodeType(int value) {
this.value = value;
}
// @@protoc_insertion_point(enum_scope:NodeType)
}
public static com.google.protobuf.Descriptors.FileDescriptor getDescriptor() {
return descriptor;
}
private static com.google.protobuf.Descriptors.FileDescriptor descriptor;
static {
String[] descriptorData = {"\n\016NodeType.proto*\223\001\n\010NodeType\022\014\n\010DOCUMEN"
+ "T\020\000\022\013\n\007SECTION\020\001\022\021\n\rSUPER_SECTION\020\002\022\014\n\010H"
+ "EADLINE\020\003\022\r\n\tPARAGRAPH\020\004\022\t\n\005TABLE\020\005\022\016\n\nT"
+ "ABLE_CELL\020\006\022\t\n\005IMAGE\020\007\022\n\n\006HEADER\020\010\022\n\n\006FO"
+ "OTER\020\tb\006proto3"};
descriptor = com.google.protobuf.Descriptors.FileDescriptor.internalBuildGeneratedFileFrom(descriptorData, new com.google.protobuf.Descriptors.FileDescriptor[]{});
descriptor.resolveAllFeaturesImmutable();
}
// @@protoc_insertion_point(outer_class_scope)
}
@@ -19,6 +19,16 @@ public class SimplifiedText {
@Schema(description = "Number of pages in the entire document.")
private int numberOfPages;
@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";
public static final String LAYOUT_PARSING_REQUEST_EXCHANGE = "layout_parsing_request_exchange";
public static final String LAYOUT_PARSING_RESPONSE_QUEUE_PREFIX = "layout_parsing_response";
public static final String LAYOUT_PARSING_RESPONSE_EXCHANGE = "layout_parsing_response_exchange";
public static final String LAYOUT_PARSING_DLQ = "layout_parsing_error";
}
@@ -0,0 +1,21 @@
syntax = "proto3";
message AllDocumentPages {
repeated DocumentPage documentPages = 1;
}
message DocumentPage {
// The page number, starting with 1.
int32 number = 1;
// The page height in PDF user units.
int32 height = 2;
// The page width in PDF user units.
int32 width = 3;
// The page rotation as specified by the PDF.
int32 rotation = 4;
}
@@ -0,0 +1,25 @@
syntax = "proto3";
message AllDocumentPositionData {
repeated DocumentPositionData documentPositionData = 1;
}
message DocumentPositionData {
// Identifier of the text block.
int64 id = 1;
// For each string coordinate in the search text of the text block, the array contains an entry relating the string coordinate to the position coordinate.
// This is required due to the text and position coordinates not being equal.
repeated int32 stringIdxToPositionIdx = 2;
// The bounding box for each glyph as a rectangle. This matrix is of size (n,4), where n is the number of glyphs in the text block.
// The second dimension specifies the rectangle with the value x, y, width, height, with x, y specifying the lower left corner.
// In order to access this information, the stringIdxToPositionIdx array must be used to transform the coordinates.
repeated Position positions = 3;
// Definition of a BoundingBox that contains x, y, width, and height.
message Position {
repeated float value = 1;
}
}
@@ -0,0 +1,8 @@
syntax = "proto3";
import "EntryData.proto";
message DocumentStructure {
// The root EntryData represents the Document.
EntryData root = 1;
}
@@ -0,0 +1,29 @@
syntax = "proto3";
message AllDocumentTextData {
repeated DocumentTextData documentTextData = 1;
}
message DocumentTextData {
// Identifier of the text block.
int64 id = 1;
// The page the text block occurs on.
int64 page = 2;
// The text of the text block.
string searchText = 3;
// Each text block is assigned a number on a page, starting from 0.
int32 numberOnPage = 4;
// The text blocks are ordered, this number represents the start of the text block as a string offset.
int32 start = 5;
// The text blocks are ordered, this number represents the end of the text block as a string offset.
int32 end = 6;
// The line breaks in the text of this semantic node in string offsets. They are exclusive end. At the end of each semantic node there is an implicit linebreak.
repeated int32 lineBreaks = 7;
}
@@ -0,0 +1,27 @@
syntax = "proto3";
import "LayoutEngine.proto";
import "NodeType.proto";
message EntryData {
// Type of the semantic node.
NodeType type = 1;
// Specifies the position in the parsed tree structure.
repeated int32 treeId = 2;
// Specifies the text block IDs associated with this semantic node.
repeated int64 atomicBlockIds = 3;
// Specifies the pages this semantic node appears on.
repeated int64 pageNumbers = 4;
// Some semantic nodes have additional information, this information is stored in this Map.
map<string, string> properties = 5;
// All child Entries of this Entry.
repeated EntryData children = 6;
// Describes the origin of the semantic node.
repeated LayoutEngine engines = 7;
}
@@ -0,0 +1,7 @@
syntax = "proto3";
enum LayoutEngine {
ALGORITHM = 0;
AI = 1;
OUTLINE = 2;
}
@@ -0,0 +1,14 @@
syntax = "proto3";
enum NodeType {
DOCUMENT = 0;
SECTION = 1;
SUPER_SECTION = 2;
HEADLINE = 3;
PARAGRAPH = 4;
TABLE = 5;
TABLE_CELL = 6;
IMAGE = 7;
HEADER = 8;
FOOTER = 9;
}
@@ -12,12 +12,14 @@ dependencies {
implementation(project(":layoutparser-service-internal-api"))
implementation(project(":viewer-doc-processor"))
implementation("com.iqser.red.service:persistence-service-shared-api-v1:2.144.0") {
implementation("com.iqser.red.service:persistence-service-shared-api-v1:2.564.0-RED9010.0") {
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.iqser.red.commons:storage-commons:2.45.0")
implementation("com.knecon.fforesight:tenant-commons:0.30.0") {
exclude("com.iqser.red.commons", "storage-commons")
}
implementation("com.iqser.red.commons:storage-commons:2.50.0")
implementation("org.apache.pdfbox:pdfbox:${pdfBoxVersion}")
implementation("org.apache.pdfbox:pdfbox-tools:${pdfBoxVersion}")
@@ -25,7 +27,14 @@ dependencies {
implementation("com.fasterxml.jackson.datatype:jackson-datatype-jsr310:${jacksonVersion}")
implementation("org.springframework.boot:spring-boot-starter-web:3.1.3")
implementation("org.jgrapht:jgrapht-core:1.5.2")
implementation("org.apache.pdfbox:jbig2-imageio:3.0.4")
implementation("com.github.jai-imageio:jai-imageio-core:1.4.0")
implementation("com.github.jai-imageio:jai-imageio-jpeg2000:1.4.0")
implementation("org.tinspin:tinspin-indexes:2.1.3")
implementation("org.commonmark:commonmark:0.22.0")
implementation("org.commonmark:commonmark-ext-gfm-tables:0.22.0")
implementation("com.pdftron:PDFNet:10.11.0")
implementation("org.apache.commons:commons-text:1.12.0")
implementation("com.google.protobuf:protobuf-java-util:4.27.1")
}
@@ -2,13 +2,15 @@ package com.knecon.fforesight.service.layoutparser.processor;
import static java.lang.String.format;
import java.awt.geom.Point2D;
import java.awt.geom.AffineTransform;
import java.awt.geom.Rectangle2D;
import java.io.File;
import java.io.IOException;
import java.nio.file.Files;
import java.util.ArrayList;
import java.util.Collection;
import java.util.HashMap;
import java.util.HashSet;
import java.util.List;
import java.util.Map;
import java.util.concurrent.atomic.AtomicReference;
@@ -20,11 +22,16 @@ import org.apache.pdfbox.pdmodel.common.PDRectangle;
import org.apache.pdfbox.pdmodel.documentinterchange.markedcontent.PDMarkedContent;
import org.springframework.stereotype.Service;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeTypeProto.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingFinishedEvent;
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.markdown.MarkdownMapper;
import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Character;
import com.knecon.fforesight.service.layoutparser.processor.model.table.Ruling;
import com.knecon.fforesight.service.layoutparser.processor.model.text.RedTextPosition;
import com.knecon.fforesight.service.layoutparser.processor.model.text.TextDirection;
import com.knecon.fforesight.service.layoutparser.processor.services.classification.ClassificationService;
import com.knecon.fforesight.service.layoutparser.processor.services.mapper.MarkdownMapper;
import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationDocument;
import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
@@ -39,7 +46,7 @@ import com.knecon.fforesight.service.layoutparser.processor.model.outline.TableO
import com.knecon.fforesight.service.layoutparser.processor.model.table.Cell;
import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRulings;
import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;
import com.knecon.fforesight.service.layoutparser.processor.python_api.adapter.CvTableParsingAdapter;
import com.knecon.fforesight.service.layoutparser.processor.python_api.adapter.ImageServiceResponseAdapter;
import com.knecon.fforesight.service.layoutparser.processor.python_api.adapter.VisualLayoutParsingAdapter;
@@ -47,7 +54,6 @@ import com.knecon.fforesight.service.layoutparser.processor.python_api.model.ima
import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.TableCells;
import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.TableServiceResponse;
import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.VisualLayoutParsingResponse;
import com.knecon.fforesight.service.layoutparser.processor.services.BodyTextFrameService;
import com.knecon.fforesight.service.layoutparser.processor.services.RulingCleaningService;
import com.knecon.fforesight.service.layoutparser.processor.services.SectionsBuilderService;
import com.knecon.fforesight.service.layoutparser.processor.services.SimplifiedSectionTextService;
@@ -57,9 +63,6 @@ import com.knecon.fforesight.service.layoutparser.processor.services.blockificat
import com.knecon.fforesight.service.layoutparser.processor.services.blockification.DocstrumBlockificationService;
import com.knecon.fforesight.service.layoutparser.processor.services.blockification.DocuMineBlockificationService;
import com.knecon.fforesight.service.layoutparser.processor.services.blockification.RedactManagerBlockificationService;
import com.knecon.fforesight.service.layoutparser.processor.services.classification.ClarifyndClassificationService;
import com.knecon.fforesight.service.layoutparser.processor.services.classification.DocuMineClassificationService;
import com.knecon.fforesight.service.layoutparser.processor.services.classification.RedactManagerClassificationService;
import com.knecon.fforesight.service.layoutparser.processor.services.factory.DocumentGraphFactory;
import com.knecon.fforesight.service.layoutparser.processor.services.graphics.Box;
import com.knecon.fforesight.service.layoutparser.processor.services.graphics.GraphicExtractorService;
@@ -69,6 +72,7 @@ import com.knecon.fforesight.service.layoutparser.processor.services.parsing.PDF
import com.knecon.fforesight.service.layoutparser.processor.services.visualization.LayoutGridService;
import com.knecon.fforesight.service.layoutparser.processor.utils.MarkedContentUtils;
import com.knecon.fforesight.service.layoutparser.processor.utils.PageInformation;
import com.knecon.fforesight.service.layoutparser.processor.utils.TextPositionOperations;
import io.micrometer.observation.Observation;
import io.micrometer.observation.ObservationRegistry;
@@ -90,10 +94,7 @@ public class LayoutParsingPipeline {
CvTableParsingAdapter cvTableParsingAdapter;
LayoutParsingStorageService layoutParsingStorageService;
SectionsBuilderService sectionsBuilderService;
RedactManagerClassificationService redactManagerClassificationService;
DocuMineClassificationService docuMineClassificationService;
SimplifiedSectionTextService simplifiedSectionTextService;
BodyTextFrameService bodyTextFrameService;
RulingCleaningService rulingCleaningService;
TableExtractionService tableExtractionService;
DocuMineBlockificationService docuMineBlockificationService;
@@ -103,12 +104,12 @@ public class LayoutParsingPipeline {
LayoutGridService layoutGridService;
ObservationRegistry observationRegistry;
VisualLayoutParsingAdapter visualLayoutParsingAdapter;
ClarifyndClassificationService clarifyndClassificationService;
GraphicExtractorService graphicExtractorService;
OutlineExtractorService outlineExtractorService;
OutlineValidationService outlineValidationService;
TOCEnrichmentService tocEnrichmentService;
LayoutparserSettings settings;
ClassificationService classificationService;
public LayoutParsingFinishedEvent parseLayoutAndSaveFilesToStorage(LayoutParsingRequest layoutParsingRequest) throws IOException {
@@ -137,17 +138,16 @@ public class LayoutParsingPipeline {
log.info("Building document graph for {}", layoutParsingRequest.identifier());
Document documentGraph = observeBuildDocumentGraph(settings.getLayoutParsingTypeOverride() == null //
? layoutParsingRequest.layoutParsingType() : settings.getLayoutParsingTypeOverride(),
classificationDocument);
? layoutParsingRequest.layoutParsingType() : settings.getLayoutParsingTypeOverride(), classificationDocument);
log.info("Creating viewer document for {}", layoutParsingRequest.identifier());
layoutGridService.addLayoutGrid(viewerDocumentFile, documentGraph, viewerDocumentFile, false, layoutParsingRequest.visualLayoutParsingFileId().isPresent());
layoutGridService.addLayoutGrid(viewerDocumentFile, documentGraph, viewerDocumentFile, false);
log.info("Storing resulting files for {}", layoutParsingRequest.identifier());
layoutParsingStorageService.storeDocumentData(layoutParsingRequest, DocumentDataMapper.toDocumentData(documentGraph));
if(layoutParsingRequest.documentMarkdownFileStorageId().isPresent()) {
if (layoutParsingRequest.documentMarkdownFileStorageId().isPresent()) {
layoutParsingStorageService.storeMarkdownFile(layoutParsingRequest.documentMarkdownFileStorageId().get(), new MarkdownMapper().toMarkdownContent(documentGraph));
}
layoutParsingStorageService.storeSimplifiedText(layoutParsingRequest, simplifiedSectionTextService.toSimplifiedText(documentGraph));
@@ -160,9 +160,9 @@ public class LayoutParsingPipeline {
}
if (!viewerDocumentFile.equals(originFile)) {
viewerDocumentFile.delete();
assert !viewerDocumentFile.exists() || viewerDocumentFile.delete();
}
originFile.delete();
assert !originFile.exists() || originFile.delete();
return LayoutParsingFinishedEvent.builder()
.identifier(layoutParsingRequest.identifier())
@@ -242,12 +242,8 @@ public class LayoutParsingPipeline {
}
List<ClassificationPage> classificationPages = new ArrayList<>();
OutlineObject lastProcessedOutlineObject = null;
// parsing the structure elements could be useful as well
if (layoutParsingType != LayoutParsingType.REDACT_MANAGER_OLD && layoutParsingType != LayoutParsingType.DOCUMINE_OLD) {
classificationDocument.setOutlineObjectTree(outlineExtractorService.getOutlineObjectTree(originDocument));
}
classificationDocument.setOutlineObjectTree(outlineExtractorService.getOutlineObjectTree(originDocument));
long pageCount = originDocument.getNumberOfPages();
@@ -271,21 +267,23 @@ public class LayoutParsingPipeline {
stripper.setStartPage(pageNumber);
stripper.setEndPage(pageNumber);
stripper.setPdpage(pdPage);
if (layoutParsingType.equals(LayoutParsingType.DOCUMINE_OLD)) {
stripper.setSortByPosition(true);
}
stripper.getText(originDocument);
List<TextPositionSequence> words = stripper.getTextPositionSequences();
List<Word> words = stripper.getWords();
// rotateDirAdjExactly(words, pdPage); // works really well for many highly rotated documents (e.g. VV-331340.pdf), but it decreases the headline performance by 1.3%, so I am leaving it out for now
if (layoutParsingType.equals(LayoutParsingType.DOCUMINE_OLD)) {
var lines = TextPositionOperations.groupByLine(new HashSet<>(words));
classificationDocument.getLayoutDebugLayer().addLineVisualizationsFromNestedTextPosition(lines, pageNumber);
words = TextPositionOperations.sortWords(lines);
}
classificationDocument.getLayoutDebugLayer().addTextVisualizations(words, pageNumber);
PDRectangle pdr = pdPage.getMediaBox();
int rotation = pdPage.getRotation();
boolean isLandscape = pdr.getWidth() > pdr.getHeight() && (rotation == 0 || rotation == 180) || pdr.getHeight() > pdr.getWidth() && (rotation == 90 || rotation == 270);
PDRectangle cropbox = pdPage.getCropBox();
classificationDocument.getLayoutDebugLayer().addRulingVisualization(stripper.getRulings(), pageNumber);
CleanRulings cleanRulings = rulingCleaningService.deduplicateAndStraightenRulings(pdfTableCells.get(pageNumber), stripper.getRulings());
List<Ruling> rulings = stripper.getRulings();
classificationDocument.getLayoutDebugLayer().addRulingVisualization(rulings, pageNumber);
CleanRulings cleanRulings = rulingCleaningService.deduplicateAndStraightenRulings(pdfTableCells.get(pageNumber), rulings);
PageInformation pageInformation = PageInformation.fromPDPage(pageNumber, pdPage);
List<Cell> emptyTableCells = TableExtractionService.findCells(cleanRulings.getHorizontals(), cleanRulings.getVerticals(), pageInformation);
@@ -293,7 +291,7 @@ public class LayoutParsingPipeline {
TextRulingsClassifier.classifyUnderlinedAndStrikethroughText(words, cleanRulings);
List<Box> graphics = graphicExtractorService.extractPathElementGraphics(originDocument, pdPage, pageNumber, cleanRulings, stripper.getTextPositionSequences(), false);
List<Box> graphics = graphicExtractorService.extractPathElementGraphics(originDocument, pdPage, pageNumber, cleanRulings, stripper.getWords(), false);
pdfImages.computeIfAbsent(pageNumber, x -> new ArrayList<>())
.addAll(graphics.stream()
@@ -305,8 +303,7 @@ public class LayoutParsingPipeline {
.toList());
ClassificationPage classificationPage = switch (layoutParsingType) {
case REDACT_MANAGER_OLD ->
redactManagerBlockificationService.blockify(stripper.getTextPositionSequences(), cleanRulings, classificationDocument.getLayoutDebugLayer());
case REDACT_MANAGER_OLD -> redactManagerBlockificationService.blockify(stripper.getWords(), cleanRulings, classificationDocument.getLayoutDebugLayer());
case DOCUMINE_OLD -> docuMineBlockificationService.blockify(words, cleanRulings);
case DOCUMINE, REDACT_MANAGER, REDACT_MANAGER_PARAGRAPH_DEBUG, REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH ->
docstrumBlockificationService.blockify(words, cleanRulings, true, classificationDocument.getLayoutDebugLayer(), layoutParsingType);
@@ -314,26 +311,9 @@ public class LayoutParsingPipeline {
docstrumBlockificationService.blockify(words, cleanRulings, false, classificationDocument.getLayoutDebugLayer(), layoutParsingType);
};
classificationPage.setCleanRulings(cleanRulings);
classificationPage.setRotation(rotation);
classificationPage.setLandscape(isLandscape);
classificationPage.setPageNumber(pageNumber);
classificationPage.setPageWidth(cropbox.getWidth());
classificationPage.setPageHeight(cropbox.getHeight());
updateClassificationPage(pdPage, pdr, classificationPage, cleanRulings, pageNumber, pageInformation);
if (layoutParsingType != LayoutParsingType.REDACT_MANAGER_OLD && layoutParsingType != LayoutParsingType.DOCUMINE_OLD) {
List<OutlineObject> outlineObjects = classificationDocument.getOutlineObjectTree().getOutlineObjectsPerPage().getOrDefault(pageNumber - 1, new ArrayList<>());
OutlineObject notFoundOutlineObject = null;
if (lastProcessedOutlineObject != null && !lastProcessedOutlineObject.isFound()) {
lastProcessedOutlineObject.setPoint(new Point2D.Float(0, cropbox.getHeight()));
notFoundOutlineObject = lastProcessedOutlineObject;
}
if (!outlineObjects.isEmpty()) {
classificationPage.setOutlineObjects(outlineObjects);
lastProcessedOutlineObject = blockificationPostprocessingService.sanitizeOutlineBlocks(classificationPage, notFoundOutlineObject);
}
}
blockificationPostprocessingService.findHeadlinesFromOutline(classificationDocument, pageNumber, classificationPage, pageInformation);
classificationDocument.getLayoutDebugLayer().addMarkedContentVisualizations(stripper.getMarkedContents(), pageNumber);
// MarkedContent needs to be converted at this point, otherwise it leads to GC Problems in Pdfbox.
@@ -363,27 +343,9 @@ public class LayoutParsingPipeline {
originDocument.close();
log.info("Calculating BodyTextFrame for {}", identifier);
bodyTextFrameService.setBodyTextFrames(classificationDocument, layoutParsingType);
for (ClassificationPage page : classificationDocument.getPages()) {
classificationDocument.getLayoutDebugLayer().addCleanRulingVisualization(page.getCleanRulings(), page.getPageNumber());
}
log.info("Classify TextBlocks for {}", identifier);
switch (layoutParsingType) {
case REDACT_MANAGER, REDACT_MANAGER_PARAGRAPH_DEBUG, REDACT_MANAGER_OLD, CLARIFYND_PARAGRAPH_DEBUG, REDACT_MANAGER_WITHOUT_DUPLICATE_PARAGRAPH ->
redactManagerClassificationService.classifyDocument(classificationDocument);
case DOCUMINE_OLD, DOCUMINE -> docuMineClassificationService.classifyDocument(classificationDocument);
case CLARIFYND -> clarifyndClassificationService.classifyDocument(classificationDocument);
}
classificationService.classify(classificationDocument, layoutParsingType, identifier);
List<TextPageBlock> headlines = classificationDocument.getPages()
.stream()
.flatMap(classificationPage -> classificationPage.getTextBlocks()
.stream()
.filter(tb -> tb instanceof TextPageBlock && tb.getClassification() != null && tb.getClassification().isHeadline())
.map(tb -> (TextPageBlock) tb))
.toList();
TableOfContents tableOfContents = outlineValidationService.createToC(headlines);
TableOfContents tableOfContents = outlineValidationService.createToC(classificationDocument);
classificationDocument.setTableOfContents(tableOfContents);
log.info("Building Sections for {}", identifier);
@@ -397,6 +359,50 @@ public class LayoutParsingPipeline {
}
private static void updateClassificationPage(PDPage pdPage,
PDRectangle pdr,
ClassificationPage classificationPage,
CleanRulings cleanRulings,
int pageNumber,
PageInformation pageInformation) {
int rotation = pdPage.getRotation();
boolean isLandscape = pdr.getWidth() > pdr.getHeight() && (rotation == 0 || rotation == 180) || pdr.getHeight() > pdr.getWidth() && (rotation == 90 || rotation == 270);
classificationPage.setCleanRulings(cleanRulings);
classificationPage.setRotation(rotation);
classificationPage.setLandscape(isLandscape);
classificationPage.setPageNumber(pageNumber);
classificationPage.setPageWidth((float) pageInformation.width());
classificationPage.setPageHeight((float) pageInformation.height());
}
private static void rotateDirAdjExactly(List<Word> words, PDPage pdPage) {
for (TextDirection dir : TextDirection.values()) {
double averageRotation = words.stream()
.map(Word::getCharacters)
.flatMap(Collection::stream)
.map(Character::getTextPosition)
.filter(pos -> pos.getDir().equals(dir))
.mapToDouble(RedTextPosition::getExactDir).average().orElse(0);
if (averageRotation == 0) {
continue;
}
AffineTransform rotateInstance = AffineTransform.getRotateInstance(averageRotation, pdPage.getMediaBox().getWidth() / 2, pdPage.getMediaBox().getHeight() / 2);
for (Word word : words) {
if (!dir.equals(word.getDir())) {
continue;
}
word.transform(rotateInstance);
}
}
}
private void addNumberOfPagesToTrace(int numberOfPages, long size) {
if (observationRegistry.getCurrentObservation() != null) {
@@ -438,10 +444,10 @@ public class LayoutParsingPipeline {
// Collect all statistics for the classificationPage, except from blocks inside tables, as tables will always be added to BodyTextFrame.
for (AbstractPageBlock textBlock : classificationPage.getTextBlocks()) {
if (textBlock instanceof TextPageBlock) {
if (((TextPageBlock) textBlock).getSequences() == null) {
if (((TextPageBlock) textBlock).getWords() == null) {
continue;
}
for (TextPositionSequence word : ((TextPageBlock) textBlock).getSequences()) {
for (Word word : ((TextPageBlock) textBlock).getWords()) {
classificationPage.getTextHeightCounter().add(word.getTextHeight());
classificationPage.getFontCounter().add(word.getFont());
classificationPage.getFontSizeCounter().add(word.getFontSize());
@@ -11,7 +11,9 @@ import java.nio.file.Path;
import java.nio.file.Paths;
import java.nio.file.StandardOpenOption;
import java.util.Optional;
import java.util.concurrent.CompletableFuture;
import org.springframework.core.task.TaskExecutor;
import org.springframework.stereotype.Service;
import com.fasterxml.jackson.databind.ObjectMapper;
@@ -39,6 +41,8 @@ public class LayoutParsingStorageService {
private final StorageService storageService;
private final ObjectMapper objectMapper;
private final TaskExecutor taskExecutor;
@Observed(name = "LayoutParsingStorageService", contextualName = "get-origin-file")
public File getOriginFile(String storageId) throws IOException {
@@ -100,13 +104,35 @@ public class LayoutParsingStorageService {
}
@SneakyThrows
@Observed(name = "LayoutParsingStorageService", contextualName = "store-document-data")
public void storeDocumentData(LayoutParsingRequest layoutParsingRequest, DocumentData documentData) {
storageService.storeJSONObject(TenantContext.getTenantId(), layoutParsingRequest.structureFileStorageId(), documentData.getDocumentStructure());
storageService.storeJSONObject(TenantContext.getTenantId(), layoutParsingRequest.textBlockFileStorageId(), documentData.getDocumentTextData());
storageService.storeJSONObject(TenantContext.getTenantId(), layoutParsingRequest.positionBlockFileStorageId(), documentData.getDocumentPositions());
storageService.storeJSONObject(TenantContext.getTenantId(), layoutParsingRequest.pageFileStorageId(), documentData.getDocumentPages());
Runnable storeDocumentStructureRunnable = () -> storageService.storeProtoObject(TenantContext.getTenantId(),
layoutParsingRequest.structureFileStorageId(),
documentData.getDocumentStructure());
CompletableFuture<Void> storeDocumentStructureFuture = CompletableFuture.runAsync(storeDocumentStructureRunnable, taskExecutor);
Runnable storeDocumentTextDataRunnable = () -> storageService.storeProtoObject(TenantContext.getTenantId(),
layoutParsingRequest.textBlockFileStorageId(),
documentData.getDocumentTextData());
CompletableFuture<Void> storeDocumentTextDataFuture = CompletableFuture.runAsync(storeDocumentTextDataRunnable, taskExecutor);
Runnable storeDocumentPositionsRunnable = () -> storageService.storeProtoObject(TenantContext.getTenantId(),
layoutParsingRequest.positionBlockFileStorageId(),
documentData.getDocumentPositions());
CompletableFuture<Void> storeDocumentPositionsFuture = CompletableFuture.runAsync(storeDocumentPositionsRunnable, taskExecutor);
Runnable storeDocumentPagesRunnable = () -> storageService.storeProtoObject(TenantContext.getTenantId(),
layoutParsingRequest.pageFileStorageId(),
documentData.getDocumentPages());
CompletableFuture<Void> storeDocumentPagesFuture = CompletableFuture.runAsync(storeDocumentPagesRunnable, taskExecutor);
CompletableFuture.allOf(storeDocumentStructureFuture, storeDocumentTextDataFuture, storeDocumentPositionsFuture, storeDocumentPagesFuture).join();
}
@@ -1,6 +1,7 @@
package com.knecon.fforesight.service.layoutparser.processor.docstrum;
import java.util.ArrayList;
import java.util.EnumMap;
import java.util.List;
import java.util.stream.Collectors;
@@ -17,7 +18,7 @@ import com.knecon.fforesight.service.layoutparser.processor.docstrum.service.Zon
import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRulings;
import com.knecon.fforesight.service.layoutparser.processor.model.text.RedTextPosition;
import com.knecon.fforesight.service.layoutparser.processor.model.text.TextDirection;
import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;
import com.knecon.fforesight.service.layoutparser.processor.visualization.LayoutDebugLayer;
import lombok.RequiredArgsConstructor;
@@ -26,6 +27,7 @@ import lombok.RequiredArgsConstructor;
@RequiredArgsConstructor
public class DocstrumSegmentationService {
public static final double SAME_DIRECTION_THRESHOLD = 0.9;
private final NearestNeighbourService nearestNeighbourService;
private final SpacingService spacingService;
private final LineBuilderService lineBuilderService;
@@ -33,30 +35,57 @@ public class DocstrumSegmentationService {
private final ReadingOrderService readingOrderService;
public List<Zone> segmentPage(List<TextPositionSequence> textPositions, boolean xyOrder, CleanRulings usedRulings, LayoutDebugLayer visualizations) {
public List<Zone> segmentPage(List<Word> textPositions, boolean xyOrder, CleanRulings usedRulings, LayoutDebugLayer visualizations) {
List<Zone> zones = new ArrayList<>();
zones.addAll(computeZones(textPositions, usedRulings, visualizations, TextDirection.ZERO));
zones.addAll(computeZones(textPositions, usedRulings, visualizations, TextDirection.QUARTER_CIRCLE));
zones.addAll(computeZones(textPositions, usedRulings, visualizations, TextDirection.HALF_CIRCLE));
zones.addAll(computeZones(textPositions, usedRulings, visualizations, TextDirection.THREE_QUARTER_CIRCLE));
EnumMap<TextDirection, Integer> directionCounts = new EnumMap<>(TextDirection.class);
return readingOrderService.resolve(zones, xyOrder);
List<Zone> newZones = computeZones(textPositions, usedRulings, visualizations, TextDirection.ZERO);
directionCounts.put(TextDirection.ZERO, newZones.size());
List<Zone> zones = new ArrayList<>(newZones);
newZones = computeZones(textPositions, usedRulings, visualizations, TextDirection.QUARTER_CIRCLE);
directionCounts.put(TextDirection.QUARTER_CIRCLE, newZones.size());
zones.addAll(newZones);
newZones = computeZones(textPositions, usedRulings, visualizations, TextDirection.HALF_CIRCLE);
directionCounts.put(TextDirection.HALF_CIRCLE, newZones.size());
zones.addAll(newZones);
newZones = computeZones(textPositions, usedRulings, visualizations, TextDirection.THREE_QUARTER_CIRCLE);
directionCounts.put(TextDirection.THREE_QUARTER_CIRCLE, newZones.size());
zones.addAll(newZones);
return readingOrderService.resolve(zones, xyOrder, mostSameDirection(directionCounts));
}
private List<Zone> computeZones(List<TextPositionSequence> textPositions, CleanRulings rulings, LayoutDebugLayer visualizations, TextDirection direction) {
private boolean mostSameDirection(EnumMap<TextDirection, Integer> directionCounts) {
List<RedTextPosition> positions = textPositions.stream()
int total = directionCounts.values()
.stream()
.mapToInt(i -> i).sum();
if ((double) directionCounts.get(TextDirection.ZERO) / total > SAME_DIRECTION_THRESHOLD) {
return true;
} else if ((double) directionCounts.get(TextDirection.QUARTER_CIRCLE) / total > SAME_DIRECTION_THRESHOLD) {
return true;
} else if ((double) directionCounts.get(TextDirection.HALF_CIRCLE) / total > SAME_DIRECTION_THRESHOLD) {
return true;
} else if ((double) directionCounts.get(TextDirection.THREE_QUARTER_CIRCLE) / total > SAME_DIRECTION_THRESHOLD) {
return true;
}
return false;
}
private List<Zone> computeZones(List<Word> textPositions, CleanRulings rulings, LayoutDebugLayer visualizations, TextDirection direction) {
List<Character> characters = textPositions.stream()
.filter(t -> t.getDir() == direction)
.map(TextPositionSequence::getTextPositions)
.map(Word::getCharacters)
.flatMap(List::stream)
.toList();
List<Character> characters = positions.stream()
.map(Character::new)
.collect(Collectors.toList());
nearestNeighbourService.findNearestNeighbors(characters);
double characterSpacing = spacingService.computeCharacterSpacing(characters);
@@ -15,10 +15,16 @@ public class AngleFilter {
public boolean matches(Neighbor neighbor) {
return matches(neighbor.getAngle());
}
public boolean matches(double angle) {
if (lowerAngle <= upperAngle) {
return lowerAngle <= neighbor.getAngle() && neighbor.getAngle() < upperAngle;
return lowerAngle <= angle && angle < upperAngle;
} else {
return lowerAngle <= neighbor.getAngle() || neighbor.getAngle() < upperAngle;
return lowerAngle <= angle || angle < upperAngle;
}
}
@@ -7,8 +7,12 @@ import java.util.List;
import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
import lombok.Data;
import lombok.NoArgsConstructor;
import lombok.experimental.SuperBuilder;
@Data
@SuperBuilder
@NoArgsConstructor
public abstract class BoundingBox {
// Java coordinate system: (0, 0) is always upper left, x is increasing left to right and y is increasing from top to bottom.
@@ -19,7 +23,7 @@ public abstract class BoundingBox {
// This rotates completely in 90 degree steps with page rotation.
// Needs to be used when writing to a PDF.
// Also, these are definitely correct and should be used whenever possible.
protected Rectangle2D bBoxInitialUserSpace;
protected Rectangle2D bBoxPdf;
protected static final float VERTICAL_COMPARISON_THRESHOLD = 0.4f;
@@ -50,25 +54,25 @@ public abstract class BoundingBox {
public double getPdfMinX() {
return bBoxInitialUserSpace.getMinX();
return bBoxPdf.getMinX();
}
public double getPdfMaxX() {
return bBoxInitialUserSpace.getMaxX();
return bBoxPdf.getMaxX();
}
public double getPdfMinY() {
return bBoxInitialUserSpace.getMinY();
return bBoxPdf.getMinY();
}
public double getPdfMaxY() {
return bBoxInitialUserSpace.getMaxY();
return bBoxPdf.getMaxY();
}
@@ -129,13 +133,31 @@ public abstract class BoundingBox {
}
public boolean intersectsY(BoundingBox other) {
public boolean intersectsX(BoundingBox other, float threshold) {
return this.getX() - threshold <= other.getMaxX() && this.getMaxX() + threshold >= other.getX();
}
public boolean intersectsPdf(BoundingBox other) {
return this.intersectsXPdf(other) && this.intersectsYPdf(other);
}
public boolean intersectsPdf(BoundingBox other, float yThreshold, float xThreshold) {
return this.intersectsXPdf(other, xThreshold) && this.intersectsYPdf(other, yThreshold);
}
public boolean intersectsYPdf(BoundingBox other) {
return this.getPdfMinY() <= other.getPdfMaxY() && this.getPdfMaxY() >= other.getPdfMinY();
}
public boolean intersectsYJava(BoundingBox other) {
public boolean intersectsY(BoundingBox other) {
return this.getY() <= other.getMaxY() && this.getMaxY() >= other.getY();
}
@@ -143,25 +165,31 @@ public abstract class BoundingBox {
public boolean intersectsY(BoundingBox other, float threshold) {
return this.getY() - threshold <= other.getMaxY() && this.getMaxY() + threshold >= other.getY();
}
public boolean intersectsYPdf(BoundingBox other, float threshold) {
return this.getPdfMinY() - threshold <= other.getPdfMaxY() && this.getPdfMaxY() + threshold >= other.getPdfMinY();
}
public boolean intersectsX(BoundingBox other) {
public boolean intersectsXPdf(BoundingBox other) {
return this.getPdfMinX() <= other.getPdfMaxX() && this.getPdfMaxX() >= other.getPdfMinX();
}
public boolean intersectsXJava(BoundingBox other) {
public boolean intersectsX(BoundingBox other) {
return this.getX() <= other.getMaxX() && this.getMaxX() >= other.getMinX();
}
public boolean intersectsX(BoundingBox other, float threshold) {
public boolean intersectsXPdf(BoundingBox other, float threshold) {
return this.getPdfMinX() - threshold <= other.getMaxX() && this.getMaxX() + threshold >= other.getPdfMinX();
return this.getPdfMinX() - threshold <= other.getPdfMaxX() && this.getMaxX() + threshold >= other.getPdfMinX();
}
@@ -170,8 +198,8 @@ public abstract class BoundingBox {
this.bBox = components.stream()
.map(BoundingBox::getBBox)
.collect(RectangleTransformations.collectBBox());
this.bBoxInitialUserSpace = components.stream()
.map(BoundingBox::getBBoxInitialUserSpace)
this.bBoxPdf = components.stream()
.map(BoundingBox::getBBoxPdf)
.collect(RectangleTransformations.collectBBox());
}
@@ -197,57 +225,55 @@ public abstract class BoundingBox {
public double horizontalDistance(BoundingBox other) {
Rectangle2D left;
Rectangle2D right;
if (this.leftOf(other)) {
left = this.getBBox();
right = other.getBBox();
} else {
left = other.getBBox();
right = this.getBBox();
}
double rect1Right = getMaxX();
double rect1Left = getMinX();
double rect2Right = other.getMaxX();
double rect2Left = other.getMinX();
return Math.max(0, right.getMinX() - left.getMaxX());
if (rect1Left > rect2Right || rect2Left > rect1Right) {
return Math.max(rect2Left - rect1Right, rect1Left - rect2Right);
} else {
return 0;
}
}
public double verticalDistance(BoundingBox other) {
Rectangle2D bottom;
Rectangle2D top;
if (this.isAbove(other)) {
top = this.getBBox();
bottom = other.getBBox();
} else {
bottom = this.getBBox();
top = other.getBBox();
}
double rect1Top = getMaxY();
double rect1Bottom = getMinY();
double rect2Top = other.getMaxY();
double rect2Bottom = other.getMinY();
return Math.max(0, bottom.getMinY() - top.getMaxY());
if (rect1Bottom > rect2Top || rect2Bottom > rect1Top) {
return Math.max(rect2Bottom - rect1Top, rect1Bottom - rect2Top);
} else {
return 0;
}
}
public boolean rightOf(BoundingBox other) {
return this.intersectsYJava(other) && other.getMaxX() <= this.getMinX();
return this.intersectsY(other) && other.getMaxX() <= this.getMinX();
}
public boolean leftOf(BoundingBox other) {
return this.intersectsYJava(other) && other.getMinX() >= this.getMaxX();
return this.intersectsY(other) && other.getMinX() >= this.getMaxX();
}
public boolean isAbove(BoundingBox other) {
return this.intersectsXJava(other) && other.getMinY() >= this.getMaxY();
return this.intersectsX(other) && other.getMinY() >= this.getMaxY();
}
public boolean isBelow(BoundingBox other) {
return this.intersectsXJava(other) && this.getMinY() >= other.getMaxY();
return this.intersectsX(other) && this.getMinY() >= other.getMaxY();
}
}
@@ -35,7 +35,7 @@ public class Character {
public double getHeight() {
return textPosition.getHeightDir();
return textPosition.getHeightDirAdj();
}
@@ -65,9 +65,9 @@ public class Character {
double s = Math.sin(-0);
double c = Math.cos(-0);
xs[0] = c * x - s * y;
xs[1] = c * (x + textPosition.getWidthDirAdj()) - s * (y + textPosition.getHeightDir());
xs[1] = c * (x + textPosition.getWidthDirAdj()) - s * (y + textPosition.getHeightDirAdj());
xs[2] = c * other.x - s * other.y;
xs[3] = c * (other.x + other.textPosition.getWidthDirAdj()) - s * (other.y + other.textPosition.getHeightDir());
xs[3] = c * (other.x + other.textPosition.getWidthDirAdj()) - s * (other.y + other.textPosition.getHeightDirAdj());
boolean overlapping = xs[1] >= xs[2] && xs[3] >= xs[0];
Arrays.sort(xs);
return Math.abs(xs[2] - xs[1]) * (overlapping ? 1 : -1);
@@ -1,18 +1,28 @@
package com.knecon.fforesight.service.layoutparser.processor.docstrum.model;
import static com.knecon.fforesight.service.layoutparser.processor.model.text.Word.BOLD;
import static com.knecon.fforesight.service.layoutparser.processor.model.text.Word.BOLD_ITALIC;
import static com.knecon.fforesight.service.layoutparser.processor.model.text.Word.ITALIC;
import static com.knecon.fforesight.service.layoutparser.processor.model.text.Word.STANDARD;
import java.util.ArrayList;
import java.util.Arrays;
import java.util.Comparator;
import java.util.EnumMap;
import java.util.List;
import java.util.Map;
import java.util.concurrent.atomic.AtomicInteger;
import com.knecon.fforesight.service.layoutparser.processor.docstrum.utils.FastAtan2;
import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;
import com.knecon.fforesight.service.layoutparser.processor.model.text.FontStyle;
import lombok.Data;
import lombok.EqualsAndHashCode;
@Data
@EqualsAndHashCode(onlyExplicitlyIncluded = true)
public class Line extends BoundingBox {
@EqualsAndHashCode(onlyExplicitlyIncluded = true, callSuper = false)
public class Line extends TextBoundingBox {
private static final double WORD_DISTANCE_MULTIPLIER = 0.18;
@@ -26,16 +36,13 @@ public class Line extends BoundingBox {
@EqualsAndHashCode.Include
private final double y1;
private final double height;
private FontStyle fontStyle;
private final List<Character> characters;
private final List<TextPositionSequence> words = new ArrayList<>();
private final List<Word> words;
public Line(List<Character> characters, double wordSpacing) {
this.characters = characters;
if (characters.size() >= 2) {
// linear regression
double sx = 0.0;
@@ -64,9 +71,43 @@ public class Line extends BoundingBox {
this.y0 = character.getY() - dy;
this.y1 = character.getY() + dy;
}
height = computeHeight();
computeWords(wordSpacing * WORD_DISTANCE_MULTIPLIER);
this.words = new ArrayList<>();
computeWords(characters, wordSpacing * WORD_DISTANCE_MULTIPLIER);
buildBBox();
computeFontStyle();
}
public Line(List<Word> words) {
this.words = words;
buildBBox();
x0 = getMinX();
y0 = getMinY();
x1 = getMaxX();
y1 = getMaxY();
computeFontStyle();
}
private void computeFontStyle() {
EnumMap<FontStyle, AtomicInteger> fontStyleCounter = new EnumMap<>(FontStyle.class);
for (FontStyle fontStyle : FontStyle.values()) {
fontStyleCounter.put(fontStyle, new AtomicInteger(0));
}
for (Word word : words) {
switch (word.getFontStyle()) {
case STANDARD -> fontStyleCounter.get(FontStyle.REGULAR).getAndIncrement();
case BOLD -> fontStyleCounter.get(FontStyle.BOLD).getAndIncrement();
case ITALIC -> fontStyleCounter.get(FontStyle.ITALIC).getAndIncrement();
case BOLD_ITALIC -> fontStyleCounter.get(FontStyle.BOLD_ITALIC).getAndIncrement();
}
}
fontStyle = fontStyleCounter.entrySet()
.stream()
.max(Comparator.comparing(entry -> entry.getValue().get()))
.map(Map.Entry::getKey).orElse(FontStyle.REGULAR);
}
@@ -82,14 +123,6 @@ public class Line extends BoundingBox {
}
private double computeHeight() {
return characters.stream()
.map(Character::getHeight)
.reduce(0d, Double::sum) / characters.size();
}
public double angularDifference(Line j) {
double diff = Math.abs(getAngle() - j.getAngle());
@@ -122,19 +155,19 @@ public class Line extends BoundingBox {
}
private void computeWords(double wordSpacing) {
private void computeWords(List<Character> characters, double wordSpacing) {
TextPositionSequence word = new TextPositionSequence();
Word word = new Word();
Character previous = null;
for (Character current : characters) {
if (previous != null) {
double dist = current.getTextPosition().getXDirAdj() - previous.getTextPosition().getXDirAdj() - previous.getTextPosition().getWidthDirAdj();
if (dist > wordSpacing) {
words.add(word);
word = new TextPositionSequence();
word = new Word();
}
}
word.getTextPositions().add(current.getTextPosition());
word.add(current);
previous = current;
}
words.add(word);
@@ -143,9 +176,7 @@ public class Line extends BoundingBox {
private void buildBBox() {
this.setToBBoxOfComponents(characters.stream()
.map(Character::getTextPosition)
.toList());
this.setToBBoxOfComponents(words);
}
@@ -0,0 +1,180 @@
package com.knecon.fforesight.service.layoutparser.processor.docstrum.model;
import java.awt.geom.Rectangle2D;
import java.util.List;
import java.util.Set;
import java.util.stream.Collectors;
import com.knecon.fforesight.service.layoutparser.processor.model.text.TextDirection;
import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
import lombok.EqualsAndHashCode;
import lombok.Getter;
import lombok.NoArgsConstructor;
import lombok.Setter;
import lombok.experimental.SuperBuilder;
@Getter
@Setter
@SuperBuilder
@NoArgsConstructor
@EqualsAndHashCode(callSuper = false)
public abstract class TextBoundingBox extends BoundingBox {
protected Rectangle2D bBoxDirAdj;
protected TextDirection dir;
@Override
public void setToBBoxOfComponents(List<? extends BoundingBox> components) {
super.setToBBoxOfComponents(components);
this.bBoxDirAdj = components.stream()
.filter(c -> c instanceof TextBoundingBox)
.map(c -> (TextBoundingBox) c)
.map(TextBoundingBox::getBBoxDirAdj)
.collect(RectangleTransformations.collectBBox());
Set<TextDirection> textDirections = components.stream()
.filter(c -> c instanceof TextBoundingBox)
.map(c -> (TextBoundingBox) c)
.map(TextBoundingBox::getDir)
.collect(Collectors.toSet());
if (textDirections.isEmpty()) {
dir = TextDirection.ZERO;
} else if (textDirections.size() > 1) {
throw new IllegalArgumentException("More than one text direction found");
} else {
dir = textDirections.iterator().next();
}
}
public double getXDirAdj() {
return this.bBoxDirAdj.getX();
}
public double getYDirAdj() {
return this.bBoxDirAdj.getY();
}
public double getWidthDirAdj() {
return this.bBoxDirAdj.getWidth();
}
public double getHeightDirAdj() {
return this.bBoxDirAdj.getHeight();
}
public double getMaxXDirAdj() {
return this.bBoxDirAdj.getMaxX();
}
public double getMaxYDirAdj() {
return this.bBoxDirAdj.getMaxY();
}
public double getCenterYDirAdj() {
return this.bBoxDirAdj.getCenterY();
}
public double getCenterXDirAdj() {
return this.bBoxDirAdj.getCenterX();
}
public double horizontalDistanceDirAdj(TextBoundingBox other) {
double rect1Right = getMaxXDirAdj();
double rect1Left = getXDirAdj();
double rect2Right = other.getMaxXDirAdj();
double rect2Left = other.getXDirAdj();
if (rect1Left > rect2Right || rect2Left > rect1Right) {
return Math.max(rect2Left - rect1Right, rect1Left - rect2Right);
} else {
return 0;
}
}
public double verticalDistanceDirAdj(TextBoundingBox other) {
double rect1Top = getMaxYDirAdj();
double rect1Bottom = getYDirAdj();
double rect2Top = other.getMaxYDirAdj();
double rect2Bottom = other.getYDirAdj();
if (rect1Bottom > rect2Top || rect2Bottom > rect1Top) {
return Math.max(rect2Bottom - rect1Top, rect1Bottom - rect2Top);
} else {
return 0;
}
}
public boolean intersectsDirAdj(TextBoundingBox other) {
return this.intersectsXDirAdj(other) && this.intersectsYDirAdj(other);
}
public boolean intersectsDirAdj(TextBoundingBox other, float yThreshold, float xThreshold) {
return this.intersectsXDirAdj(other, xThreshold) && this.intersectsYDirAdj(other, yThreshold);
}
public boolean intersectsXDirAdj(TextBoundingBox other, float threshold) {
return this.getXDirAdj() - threshold <= other.getMaxXDirAdj() && this.getMaxXDirAdj() + threshold >= other.getXDirAdj();
}
public boolean intersectsXDirAdj(TextBoundingBox other) {
return this.getXDirAdj() <= other.getMaxXDirAdj() && this.getMaxXDirAdj() >= other.getXDirAdj();
}
public boolean intersectsYDirAdj(TextBoundingBox other) {
return this.getYDirAdj() <= other.getMaxYDirAdj() && this.getMaxYDirAdj() >= other.getYDirAdj();
}
public boolean intersectsYDirAdj(TextBoundingBox other, float threshold) {
return this.getYDirAdj() - threshold <= other.getMaxYDirAdj() && this.getMaxYDirAdj() + threshold >= other.getYDirAdj();
}
public boolean isAboveDirAdj(TextBoundingBox other) {
return other.isBelow(this);
}
public boolean isBelowDirAdj(TextBoundingBox other) {
return this.intersectsXDirAdj(other) && this.getYDirAdj() >= other.getMaxYDirAdj();
}
}
@@ -28,4 +28,10 @@ public class UnionFind<T> extends org.jgrapht.alg.util.UnionFind<T> {
return setRep.values();
}
public Collection<T> getElements() {
return getParentMap().keySet();
}
}
@@ -6,9 +6,11 @@ import java.util.List;
import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
import lombok.Data;
import lombok.EqualsAndHashCode;
@Data
public class Zone extends BoundingBox {
@EqualsAndHashCode(callSuper = false)
public class Zone extends TextBoundingBox {
private List<Line> lines;
@@ -16,7 +18,6 @@ public class Zone extends BoundingBox {
@SuppressWarnings("PMD.ConstructorCallsOverridableMethod")
public Zone(List<Line> lines) {
lines.sort(Comparator.comparingDouble(Line::getY0));
this.lines = lines;
setToBBoxOfComponents(lines);
}
@@ -17,7 +17,7 @@ public class LineBuilderService {
private static final double CHARACTER_SPACING_DISTANCE_MULTIPLIER = 3.5;
private static final double LINE_SPACING_THRESHOLD_MULTIPLIER = 0.67;
private static final double ANGLE_TOLERANCE = Math.PI / 6;
private static final double ANGLE_TOLERANCE = Math.toRadians(5);
public List<Line> buildLines(List<Character> characters, double characterSpacing, double lineSpacing, CleanRulings rulings) {
@@ -1,15 +1,17 @@
package com.knecon.fforesight.service.layoutparser.processor.docstrum.service;
import java.awt.geom.Rectangle2D;
import java.util.ArrayList;
import java.util.Comparator;
import java.util.HashMap;
import java.util.List;
import java.util.ListIterator;
import java.util.Map;
import java.util.stream.Collectors;
import org.springframework.stereotype.Service;
import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.BoundingBox;
import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.TextBoundingBox;
import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Zone;
import com.knecon.fforesight.service.layoutparser.processor.docstrum.utils.DoubleUtils;
@@ -19,21 +21,30 @@ public class ReadingOrderService {
private static final double THRESHOLD = 5;
public static final double MULTI_COLUMN_DETECTION_THRESHOLD = 1.5;
private static final Comparator<TextBoundingBox> COMPARATOR = //
Comparator.comparing(TextBoundingBox::getY, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD))
.thenComparing(TextBoundingBox::getX, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD));
public List<Zone> resolve(List<Zone> zones, boolean xyReadingOrder) {
private static final Comparator<TextBoundingBox> COMPARATOR_DIR_ADJ = //
Comparator.comparing(TextBoundingBox::getYDirAdj, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD))
.thenComparing(TextBoundingBox::getXDirAdj, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD));
public List<Zone> resolve(List<Zone> zones, boolean xyReadingOrder, boolean useDirAdjCoords) {
if (zones.isEmpty() || zones.size() == 1) {
return zones;
}
if (xyReadingOrder) {
return resolveSingleColumnReadingOrder(zones);
return resolveSingleColumnReadingOrder(zones, useDirAdjCoords);
}
Map<Long, Integer> histogram = new HashMap<>();
for (Zone zone : zones) {
long minY = Math.round(zone.getBBox().getMinY());
long maxY = Math.round(zone.getBBox().getMaxY());
Rectangle2D bbox = useDirAdjCoords ? zone.getBBoxDirAdj() : zone.getBBox();
long minY = Math.round(bbox.getMinY());
long maxY = Math.round(bbox.getMaxY());
for (long i = minY; i <= maxY; i++) {
histogram.put(i, histogram.getOrDefault(i, 0) + 1);
}
@@ -43,24 +54,32 @@ public class ReadingOrderService {
.stream()
.mapToInt(Integer::intValue).average()
.orElse(1) < MULTI_COLUMN_DETECTION_THRESHOLD) {
return resolveSingleColumnReadingOrder(zones);
return resolveSingleColumnReadingOrder(zones, useDirAdjCoords);
} else {
return resolveMultiColumnReadingOder(zones);
return resolveMultiColumnReadingOder(zones, useDirAdjCoords);
}
}
private static List<Zone> resolveSingleColumnReadingOrder(List<Zone> zones) {
private static List<Zone> resolveSingleColumnReadingOrder(List<Zone> zones, boolean useDirAdjCoords) {
zones.sort(Comparator.comparing(BoundingBox::getY, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD))
.thenComparing(BoundingBox::getX, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD)));
if (useDirAdjCoords) {
return zones.stream()
.collect(Collectors.groupingBy(TextBoundingBox::getDir)).values()
.stream()
.flatMap(words -> words.stream()
.sorted(COMPARATOR_DIR_ADJ))
.toList();
}
zones.sort(COMPARATOR);
return zones;
}
private List<Zone> resolveMultiColumnReadingOder(List<Zone> zones) {
private List<Zone> resolveMultiColumnReadingOder(List<Zone> zones, boolean useDirAdjCoords) {
// Simple reading order resolver for multi column page layout as described here : https://pub.towardsai.net/advanced-rag-02-unveiling-pdf-parsing-b84ae866344e
// TODO implement a more fancy reading order resolver see https://github.com/BobLd/DocumentLayoutAnalysis/blob/master/README.md#reading-order
@@ -69,11 +88,12 @@ public class ReadingOrderService {
double maxX = Double.NEGATIVE_INFINITY;
for (Zone zone : zones) {
if (zone.getX() < minX) {
minX = zone.getX();
Rectangle2D bbox = useDirAdjCoords ? zone.getBBoxDirAdj() : zone.getBBox();
if (bbox.getX() < minX) {
minX = zone.getXDirAdj();
}
if (zone.getX() + zone.getWidth() > maxX) {
maxX = zone.getX() + zone.getWidth();
if (bbox.getMaxX() > maxX) {
maxX = zone.getMaxXDirAdj();
}
}
@@ -82,24 +102,27 @@ public class ReadingOrderService {
List<Zone> leftOf = new ArrayList<>();
List<Zone> rightOf = new ArrayList<>();
List<Zone> middle = new ArrayList<>();
for (Zone zone : zones) {
if (zone.getX() < midLineXCoordinate && zone.getX() + zone.getWidth() < midLineXCoordinate) {
Rectangle2D bbox = useDirAdjCoords ? zone.getBBoxDirAdj() : zone.getBBox();
if (bbox.getX() < midLineXCoordinate && bbox.getX() + bbox.getWidth() < midLineXCoordinate) {
leftOf.add(zone);
} else if (zone.getX() > midLineXCoordinate && zone.getX() + zone.getWidth() > midLineXCoordinate) {
} else if (bbox.getX() > midLineXCoordinate && bbox.getX() + bbox.getWidth() > midLineXCoordinate) {
rightOf.add(zone);
} else {
middle.add(zone);
}
}
leftOf.sort(Comparator.comparing(BoundingBox::getY, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD))
.thenComparing(BoundingBox::getX, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD)));
rightOf.sort(Comparator.comparing(BoundingBox::getY, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD))
.thenComparing(BoundingBox::getX, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD)));
middle.sort(Comparator.comparing(BoundingBox::getY, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD))
.thenComparing(BoundingBox::getX, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD)));
if (useDirAdjCoords) {
leftOf.sort(COMPARATOR_DIR_ADJ);
rightOf.sort(COMPARATOR_DIR_ADJ);
middle.sort(COMPARATOR_DIR_ADJ);
} else {
leftOf.sort(COMPARATOR);
rightOf.sort(COMPARATOR);
middle.sort(COMPARATOR);
}
/*
List<Zone> leftNotIntersecting = new ArrayList<>();
for (Zone leftZone : leftOf) {
@@ -151,8 +174,9 @@ public class ReadingOrderService {
while (itty.hasNext()) {
Zone current = itty.next();
Rectangle2D bbox = useDirAdjCoords ? current.getBBoxDirAdj() : current.getBBox();
for (int i = 0; i < sortedZones.size(); i++) {
if (current.getY() < sortedZones.get(i).getY()) {
if (bbox.getY() < sortedZones.get(i).getY()) {
sortedZones.add(i, current);
itty.remove();
break;
@@ -1,7 +1,7 @@
package com.knecon.fforesight.service.layoutparser.processor.docstrum.service;
import java.util.ArrayList;
import java.util.Comparator;
import java.util.Collection;
import java.util.HashSet;
import java.util.List;
import java.util.Set;
@@ -9,11 +9,12 @@ import java.util.stream.Collectors;
import org.springframework.stereotype.Service;
import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Character;
import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Line;
import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.UnionFind;
import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Zone;
import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRulings;
import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;
import com.knecon.fforesight.service.layoutparser.processor.utils.TextPositionOperations;
@Service
public class ZoneBuilderService {
@@ -21,7 +22,7 @@ public class ZoneBuilderService {
private static final double MIN_HORIZONTAL_DISTANCE_MULTIPLIER = -0.5;
private static final double MAX_VERTICAL_DISTANCE_MULTIPLIER = 1.2;
private static final double MIN_HORIZONTAL_MERGE_DISTANCE_MULTIPLIER = -3.0;
private static final double MIN_HORIZONTAL_MERGE_DISTANCE_MULTIPLIER = -7;
private static final double MAX_VERTICAL_MERGE_DISTANCE_MULTIPLIER = 0.5;
@@ -29,7 +30,7 @@ public class ZoneBuilderService {
private static final double MAX_LINE_SIZE_SCALE = 2.5;
private static final double ANGLE_TOLERANCE = Math.PI / 6;
private static final double ANGLE_TOLERANCE = Math.toRadians(5);
private static final double MAX_VERTICAL_MERGE_DISTANCE = 0.5;
@@ -38,7 +39,7 @@ public class ZoneBuilderService {
double minHorizontalDistance = characterSpacing * MIN_HORIZONTAL_DISTANCE_MULTIPLIER;
double maxVerticalDistance = lineSpacing * MAX_VERTICAL_DISTANCE_MULTIPLIER;
double minHorizontalMergeDistance = characterSpacing * MIN_HORIZONTAL_MERGE_DISTANCE_MULTIPLIER;
double minHorizontalMergeDistance = lineSpacing * MIN_HORIZONTAL_MERGE_DISTANCE_MULTIPLIER;
double maxVerticalMergeDistance = lineSpacing * MAX_VERTICAL_MERGE_DISTANCE_MULTIPLIER;
UnionFind<Line> unionFind = new UnionFind<>(new HashSet<>(lines));
@@ -54,11 +55,26 @@ public class ZoneBuilderService {
return;
}
double scale = Math.min(outerLine.getHeight(), innerLine.getHeight()) / meanHeight;
scale = Math.max(MIN_LINE_SIZE_SCALE, Math.min(scale, MAX_LINE_SIZE_SCALE));
// if (!innerLine.getFontStyle().equals(outerLine.getFontStyle()) //
// && !outerLine.intersectsY(innerLine, -2f)) {
// return;
// }
double horizontalDistance = outerLine.horizontalDistance(innerLine) / scale;
double verticalDistance = outerLine.verticalDistance(innerLine) / scale;
double horizontalScale = Math.min(outerLine.getHeightDirAdj(), innerLine.getHeightDirAdj()) / meanHeight;
horizontalScale = Math.max(MIN_LINE_SIZE_SCALE, Math.min(horizontalScale, MAX_LINE_SIZE_SCALE));
double verticalScale = horizontalScale;
// if (innerLine.toString().endsWith(":")
// || outerLine.toString().endsWith(":")
// || numericalIdentifierPattern.matcher(innerLine.toString()).matches()
// || numericalIdentifierPattern.matcher(outerLine.toString()).matches()) {
//
// horizontalScale *= 5;
// verticalScale /= 10;
// }
double horizontalDistance = outerLine.horizontalDistance(innerLine) / horizontalScale;
double verticalDistance = outerLine.verticalDistance(innerLine) / verticalScale;
if ((!(minHorizontalDistance <= horizontalDistance) || !(verticalDistance <= maxVerticalDistance)) //
&& (!(minHorizontalMergeDistance <= horizontalDistance) || !(verticalDistance <= maxVerticalMergeDistance))) {
@@ -87,7 +103,7 @@ public class ZoneBuilderService {
double weights = 0.0;
for (Line line : lines) {
double weight = line.getLength();
meanHeight += line.getHeight() * weight;
meanHeight += line.getHeightDirAdj() * weight;
weights += weight;
}
meanHeight /= weights;
@@ -97,64 +113,14 @@ public class ZoneBuilderService {
private Zone mergeLinesInZone(List<Line> lines, double characterSpacing, double lineSpacing) {
double maxHorizontalDistance = 0;
double minVerticalDistance = 0;
double maxVerticalDistance = lineSpacing * MAX_VERTICAL_MERGE_DISTANCE;
Set<Word> words = lines.stream()
.map(Line::getWords)
.flatMap(Collection::stream)
.collect(Collectors.toSet());
Collection<Set<Word>> groupedLines = TextPositionOperations.groupByLine(words);
UnionFind<Line> unionFind = new UnionFind<>(new HashSet<>(lines));
lines.forEach(outer -> {
lines.forEach(inner -> {
if (inner == outer) {
return;
}
double horizontalDistance = outer.horizontalDistance(inner);
double verticalDistance = outer.verticalDistance(inner);
if (horizontalDistance <= maxHorizontalDistance && minVerticalDistance <= verticalDistance && verticalDistance <= maxVerticalDistance) {
unionFind.union(outer, inner);
} else if (minVerticalDistance <= verticalDistance
&& verticalDistance <= maxVerticalDistance
&& Math.abs(horizontalDistance - Math.min(outer.getLength(), inner.getLength())) < 0.1) {
boolean characterOverlap = false;
int overlappingCount = 0;
for (Character outerCharacter : outer.getCharacters()) {
for (Character innerCharacter : inner.getCharacters()) {
double characterOverlapDistance = outerCharacter.overlappingDistance(innerCharacter);
if (characterOverlapDistance > 2) {
characterOverlap = true;
}
if (characterOverlapDistance > 0) {
overlappingCount++;
}
}
}
if (!characterOverlap && overlappingCount <= 2) {
unionFind.union(outer, inner);
}
}
});
});
List<Line> outputZone = new ArrayList<>();
for (Set<Line> group : unionFind.getGroups()) {
List<Character> characters = new ArrayList<>();
for (Line line : group) {
characters.addAll(line.getCharacters());
}
characters.sort(Comparator.comparingDouble(Character::getX));
outputZone.add(new Line(characters, characterSpacing));
}
return new Zone(outputZone.stream()
.sorted(Comparator.comparing(Line::getY0))
.collect(Collectors.toList()));
List<Line> sortedLines = TextPositionOperations.sortLines(groupedLines);
return new Zone(sortedLines);
}
}
@@ -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);
}
}
@@ -4,7 +4,7 @@ import java.util.HashSet;
import java.util.Set;
import com.fasterxml.jackson.annotation.JsonIgnore;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngineProto.LayoutEngine;
import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.BoundingBox;
import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
@@ -18,6 +18,7 @@ import lombok.RequiredArgsConstructor;
@Data
@RequiredArgsConstructor
public class ClassificationPage {
@NonNull
@@ -25,7 +26,7 @@ public class ClassificationPage {
private List<OutlineObject> outlineObjects = new ArrayList<>();
private List<AbstractPageBlock> headlines = new ArrayList<>();
private List<AbstractPageBlock> headlines = new ArrayList<>();
private List<ClassifiedImage> images = new ArrayList<>();
@@ -44,7 +45,7 @@ public class ClassificationPage {
private float pageWidth;
private float pageHeight;
CleanRulings cleanRulings;
private CleanRulings cleanRulings;
private Map<String, List<Rectangle2D>> markedContentBboxPerType = new HashMap<>();
@@ -1,6 +1,5 @@
package com.knecon.fforesight.service.layoutparser.processor.model;
import java.util.ArrayList;
import java.util.Collections;
import java.util.HashMap;
import java.util.List;
@@ -12,10 +11,14 @@ import lombok.Getter;
@Getter
public class FloatFrequencyCounter {
Map<Float, Integer> countPerValue = new HashMap<>();
Map<Double, Integer> countPerValue = new HashMap<>();
boolean changed;
Double mostPopularCache;
public void add(float value) {
public void add(double value) {
changed = true;
if (!countPerValue.containsKey(value)) {
countPerValue.put(value, 1);
@@ -25,9 +28,11 @@ public class FloatFrequencyCounter {
}
public void addAll(Map<Float, Integer> otherCounter) {
public void addAll(Map<Double, Integer> otherCounter) {
for (Map.Entry<Float, Integer> entry : otherCounter.entrySet()) {
changed = true;
for (Map.Entry<Double, Integer> entry : otherCounter.entrySet()) {
if (countPerValue.containsKey(entry.getKey())) {
countPerValue.put(entry.getKey(), countPerValue.get(entry.getKey()) + entry.getValue());
} else {
@@ -37,36 +42,36 @@ public class FloatFrequencyCounter {
}
public Float getMostPopular() {
public Double getMostPopular() {
Map.Entry<Float, Integer> mostPopular = null;
for (Map.Entry<Float, Integer> entry : countPerValue.entrySet()) {
if (mostPopular == null || entry.getValue() >= mostPopular.getValue()) {
mostPopular = entry;
if (changed || mostPopularCache == null) {
Map.Entry<Double, Integer> mostPopular = null;
for (Map.Entry<Double, Integer> entry : countPerValue.entrySet()) {
if (mostPopular == null || entry.getValue() >= mostPopular.getValue()) {
mostPopular = entry;
}
}
mostPopularCache = mostPopular != null ? mostPopular.getKey() : 0;
changed = false;
}
return mostPopular != null ? mostPopular.getKey() : null;
return mostPopularCache;
}
public List<Float> getHighterThanMostPopular() {
public List<Double> getValuesInReverseOrder() {
Float mostPopular = getMostPopular();
List<Float> higher = new ArrayList<>();
for (Float value : countPerValue.keySet()) {
if (value > mostPopular) {
higher.add(value);
}
}
return higher.stream().sorted(Collections.reverseOrder()).collect(Collectors.toList());
return countPerValue.keySet()
.stream()
.sorted(Collections.reverseOrder())
.collect(Collectors.toList());
}
public Float getHighest() {
public Double getHighest() {
Float highest = null;
for (Float value : countPerValue.keySet()) {
Double highest = null;
for (Double value : countPerValue.keySet()) {
if (highest == null || value > highest) {
highest = value;
}
@@ -3,7 +3,7 @@ package com.knecon.fforesight.service.layoutparser.processor.model;
import java.awt.geom.Rectangle2D;
import java.util.List;
import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
@@ -16,8 +16,8 @@ import lombok.experimental.FieldDefaults;
public class LineInformation {
List<Rectangle2D> lineBBox;
List<List<TextPositionSequence>> sequencesByLines;
List<List<Word>> sequencesByLines;
List<List<Rectangle2D>> bBoxWithGapsByLines;
List<List<List<TextPositionSequence>>> sequencesWithGapsByLines;
List<List<List<Word>>> sequencesWithGapsByLines;
}
@@ -9,12 +9,14 @@ public enum PageBlockType {
H6,
HEADER,
FOOTER,
TITLE,
PARAGRAPH,
PARAGRAPH_BOLD,
PARAGRAPH_ITALIC,
PARAGRAPH_UNKNOWN,
OTHER,
TABLE_OF_CONTENTS_HEADLINE,
TABLE_OF_CONTENTS_ITEM,
LIST_ITEM,
TABLE;
@@ -34,7 +36,7 @@ public enum PageBlockType {
public static int getHeadlineNumber(PageBlockType pageBlockType) {
return switch (pageBlockType) {
case H1 -> 1;
case H1, TABLE_OF_CONTENTS_HEADLINE -> 1;
case H2 -> 2;
case H3 -> 3;
case H4 -> 4;
@@ -46,6 +48,6 @@ public enum PageBlockType {
public boolean isHeadline() {
return this.equals(H1) || this.equals(H2) || this.equals(H3) || this.equals(H4) || this.equals(H5) || this.equals(H6);
return this.equals(H1) || this.equals(H2) || this.equals(H3) || this.equals(H4) || this.equals(H5) || this.equals(H6) || this.equals(TABLE_OF_CONTENTS_HEADLINE);
}
}
@@ -4,7 +4,7 @@ import java.awt.geom.Rectangle2D;
import java.util.List;
import com.knecon.fforesight.service.layoutparser.processor.model.table.Ruling;
import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;
import lombok.AllArgsConstructor;
import lombok.Builder;
@@ -15,7 +15,7 @@ import lombok.Getter;
@AllArgsConstructor
public class PageContents {
List<TextPositionSequence> sortedTextPositionSequences;
List<Word> sortedWords;
Rectangle2D cropBox;
Rectangle2D mediaBox;
List<Ruling> rulings;
@@ -3,6 +3,7 @@ package com.knecon.fforesight.service.layoutparser.processor.model;
import java.util.Collections;
import java.util.LinkedList;
import java.util.List;
import java.util.Locale;
import java.util.regex.Matcher;
import java.util.regex.Pattern;
@@ -15,11 +16,13 @@ import lombok.experimental.FieldDefaults;
@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
public class SectionIdentifier {
static Pattern numericalIdentifierPattern = Pattern.compile("^[\\s]?(\\d+)[\\s.,;]?(\\d+)?[\\s.,;]?(\\d+)?[\\s.,;]?(\\d+)?");
public static Pattern numericalIdentifierPattern = Pattern.compile("^[\\s]?(\\d{1,2})[\\s.,;]?(\\d{1,2})?[\\s.,;]?(\\d{1,2})?[\\s.,;]?(\\d{1,2})?");
public static Pattern alphanumericIdentifierPattern = Pattern.compile("^[\\s]?[A-Za-z][\\s.,;]?(\\d{1,2})[\\s.,;]?(\\d{1,2})?[\\s.,;]?(\\d{1,2})?[\\s.,;]?(\\d{1,2})?[\\s.,;]?");
public enum Format {
EMPTY,
NUMERICAL,
ALPHANUMERIC,
DOCUMENT
}
@@ -41,6 +44,10 @@ public class SectionIdentifier {
if (numericalIdentifierMatcher.find()) {
return buildNumericalSectionIdentifier(headline, numericalIdentifierMatcher);
}
Matcher alphanumericIdentifierMatcher = alphanumericIdentifierPattern.matcher(headline);
if (alphanumericIdentifierMatcher.find()) {
return buildAlphanumericSectionIdentifier(headline, alphanumericIdentifierMatcher);
}
// more formats here
return SectionIdentifier.empty();
}
@@ -75,7 +82,36 @@ public class SectionIdentifier {
}
identifiers.add(Integer.parseInt(numericalIdentifier.trim()));
}
return new SectionIdentifier(Format.NUMERICAL, identifierString, identifiers.stream().toList(), false);
return new SectionIdentifier(Format.NUMERICAL,
identifierString,
identifiers.stream()
.toList(),
false);
}
private static SectionIdentifier buildAlphanumericSectionIdentifier(String headline, Matcher alphanumericIdentifierMatcher) {
String identifierString = headline.substring(alphanumericIdentifierMatcher.start(), alphanumericIdentifierMatcher.end());
String alphanumericIdentifier = alphanumericIdentifierMatcher.group(0).substring(0, 1).toUpperCase(Locale.ENGLISH);
int mappedCharacterValue = alphanumericIdentifier.charAt(0) - 'A' + 1;
List<Integer> identifiers = new LinkedList<>();
identifiers.add(mappedCharacterValue);
for (int i = 1; i <= 3; i++) {
String numericalIdentifier = alphanumericIdentifierMatcher.group(i);
if (numericalIdentifier == null || numericalIdentifier.equals("0") || numericalIdentifier.isEmpty() || numericalIdentifier.isBlank()) {
break;
}
identifiers.add(Integer.parseInt(numericalIdentifier.trim()));
}
return new SectionIdentifier(Format.ALPHANUMERIC,
identifierString,
identifiers.stream()
.toList(),
false);
}
@@ -123,4 +159,22 @@ public class SectionIdentifier {
return identifierString;
}
public boolean isEmpty() {
return this.format.equals(Format.EMPTY);
}
public int level() {
return identifiers.size();
}
protected List<Integer> getIdentifiers() {
return identifiers;
}
}
@@ -11,6 +11,8 @@ import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Se
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.SuperSection;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Table;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.TableCell;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.TableOfContents;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.TableOfContentsItem;
public abstract class AbstractNodeVisitor implements NodeVisitor {
@@ -83,8 +85,20 @@ public abstract class AbstractNodeVisitor implements NodeVisitor {
visitChildren(tableCell);
}
@Override
public void visit(TableOfContents toc) {
private void visitChildren(SemanticNode semanticNode) {
visitChildren(toc);
}
@Override
public void visit(TableOfContentsItem toci) {
visitChildren(toci);
}
protected void visitChildren(SemanticNode semanticNode) {
semanticNode.streamChildren()
.forEach(node -> node.accept(this));
@@ -8,7 +8,7 @@ import java.util.LinkedList;
import java.util.List;
import java.util.stream.Stream;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeTypeProto.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.GenericSemanticNode;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.SemanticNode;
@@ -10,6 +10,10 @@ import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Se
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.SuperSection;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Table;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.TableCell;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.TableOfContents;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.TableOfContentsItem;
import software.amazon.awssdk.utils.builder.ToCopyableBuilder;
public interface NodeVisitor {
@@ -42,4 +46,10 @@ public interface NodeVisitor {
void visit(TableCell tableCell);
void visit(TableOfContents tableOfContents);
void visit(TableOfContentsItem tableOfContentsItem);
}
@@ -98,10 +98,10 @@ public class TextRange implements Comparable<TextRange> {
public List<TextRange> split(List<Integer> splitIndices) {
if (splitIndices.stream()
.anyMatch(idx -> !this.containsExclusive(idx))) {
.anyMatch(idx -> !this.contains(idx))) {
throw new IndexOutOfBoundsException(format("%s splitting indices are out of range for %s",
splitIndices.stream()
.filter(idx -> !this.containsExclusive(idx))
.filter(idx -> !this.contains(idx))
.toList(),
this));
}
@@ -116,7 +116,9 @@ public class TextRange implements Comparable<TextRange> {
splitBoundaries.add(new TextRange(previousIndex, splitIndex));
previousIndex = splitIndex;
}
splitBoundaries.add(new TextRange(previousIndex, end));
if (previousIndex != end) {
splitBoundaries.add(new TextRange(previousIndex, end));
}
return splitBoundaries;
}
@@ -6,7 +6,7 @@ import java.util.List;
import java.util.Map;
import java.util.Set;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngineProto.LayoutEngine;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.TextEntity;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
@@ -9,7 +9,7 @@ import java.util.Set;
import java.util.stream.Collectors;
import java.util.stream.Stream;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeTypeProto.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
@@ -25,11 +25,4 @@ public class DuplicatedParagraph extends Paragraph {
}
@Override
public String toString() {
return super.toString();
}
}
@@ -1,6 +1,6 @@
package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeTypeProto.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
@@ -1,6 +1,6 @@
package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeTypeProto.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
@@ -1,6 +1,6 @@
package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeTypeProto.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
@@ -6,7 +6,7 @@ import java.util.HashMap;
import java.util.Map;
import java.util.Set;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeTypeProto.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
@@ -1,12 +1,15 @@
package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
import java.util.Comparator;
import java.util.HashSet;
import java.util.LinkedList;
import java.util.List;
import java.util.Set;
import java.util.stream.Stream;
import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.TextEntity;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.AtomicTextBlock;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlockCollector;
@@ -29,9 +32,8 @@ public class Page {
Integer height;
Integer width;
Integer rotation;
@EqualsAndHashCode.Exclude
List<SemanticNode> mainBody;
List<AtomicTextBlock> textBlocksOnPage;
@EqualsAndHashCode.Exclude
Header header;
@EqualsAndHashCode.Exclude
@@ -53,20 +55,44 @@ public class Page {
.width((int) classificationPage.getPageWidth())
.number(classificationPage.getPageNumber())
.rotation(classificationPage.getRotation())
.mainBody(new LinkedList<>())
.textBlocksOnPage(new LinkedList<>())
.build();
}
/**
* Constructs and returns a {@link TextBlock} representing the concatenated text of all leaf semantic nodes in the main body.
*
* @return The main body text block.
*/
public TextBlock getMainBodyTextBlock() {
return mainBody.stream()
.filter(SemanticNode::isLeaf)
.map(SemanticNode::getLeafTextBlock)
return textBlocksOnPage.stream()
.filter(atb -> !atb.isEmpty())
.collect(new TextBlockCollector());
}
public List<SemanticNode> getMainBody() {
return textBlocksOnPage.stream()
.map(AtomicTextBlock::getParent)
.map(this::getHighestParentOnPage)
.distinct()
.toList();
}
private SemanticNode getHighestParentOnPage(SemanticNode node) {
SemanticNode currentNode = node;
while (currentNode.hasParent() && currentNode.getParent().onlyOnPage(this)) {
currentNode = currentNode.getParent();
}
return currentNode;
}
@Override
public String toString() {
@@ -1,6 +1,6 @@
package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeTypeProto.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
@@ -1,6 +1,7 @@
package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeTypeProto.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
import lombok.AccessLevel;
@@ -12,8 +12,8 @@ import java.util.Set;
import java.util.stream.Collectors;
import java.util.stream.Stream;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngineProto.LayoutEngine;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeTypeProto.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.TextRange;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.EntityType;
@@ -74,7 +74,8 @@ public interface SemanticNode {
return getTextBlock().getPages()
.stream()
.min(Comparator.comparingInt(Page::getNumber)).orElseThrow(() -> new IllegalStateException("SemanticNode has no Page!"));
.min(Comparator.comparingInt(Page::getNumber))
.orElseThrow(() -> new IllegalStateException("SemanticNode has no Page!"));
}
@@ -504,4 +505,17 @@ public interface SemanticNode {
void accept(NodeVisitor visitor);
/**
* Checks wether this SemanticNode appears on a single page only, and if that page is the provided one.
*
* @param page the page to check
* @return true, when SemanticNode is on a single page only and the page is the provided page. Otherwise, false.
*/
default boolean onlyOnPage(Page page) {
Set<Page> pages = getPages();
return pages.size() == 1 && pages.contains(page);
}
}
@@ -1,6 +1,6 @@
package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeTypeProto.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
import lombok.AccessLevel;
@@ -12,8 +12,8 @@ import java.util.Set;
import java.util.stream.IntStream;
import java.util.stream.Stream;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngine;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.LayoutEngineProto.LayoutEngine;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeTypeProto.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.TextEntity;
@@ -4,7 +4,7 @@ import java.awt.geom.Rectangle2D;
import java.util.HashMap;
import java.util.Map;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeTypeProto.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlockCollector;
@@ -0,0 +1,41 @@
package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeTypeProto;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Data;
import lombok.EqualsAndHashCode;
import lombok.experimental.FieldDefaults;
import lombok.experimental.SuperBuilder;
@Data
@SuperBuilder
@AllArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE)
@EqualsAndHashCode(callSuper = true)
public class TableOfContents extends AbstractSemanticNode {
@Override
public NodeTypeProto.NodeType getType() {
return NodeTypeProto.NodeType.TABLE_OF_CONTENTS;
}
public Headline getHeadline() {
return streamChildrenOfType(NodeTypeProto.NodeType.HEADLINE).map(node -> (Headline) node)
.findFirst()
.orElseGet(() -> getParent().getHeadline());
}
@Override
public void accept(NodeVisitor visitor) {
visitor.visit(this);
}
}
@@ -0,0 +1,51 @@
package com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeTypeProto;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.NodeVisitor;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Data;
import lombok.EqualsAndHashCode;
import lombok.experimental.FieldDefaults;
import lombok.experimental.SuperBuilder;
@Data
@SuperBuilder
@AllArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE)
@EqualsAndHashCode(callSuper = true)
public class TableOfContentsItem extends AbstractSemanticNode {
TextBlock leafTextBlock;
@Override
public NodeTypeProto.NodeType getType() {
return NodeTypeProto.NodeType.TABLE_OF_CONTENTS_ITEM;
}
@Override
public boolean isLeaf() {
return true;
}
@Override
public void accept(NodeVisitor visitor) {
visitor.visit(this);
}
@Override
public TextBlock getTextBlock() {
return leafTextBlock;
}
}
@@ -1,5 +1,6 @@
package com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock;
import static com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentTextDataProto.DocumentTextData;
import static java.lang.String.format;
import java.awt.geom.Rectangle2D;
@@ -13,8 +14,8 @@ import java.util.Map;
import java.util.Set;
import java.util.stream.Collectors;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentPositionData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentTextData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentPositionDataProto.DocumentPositionData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentPositionDataProto.DocumentPositionData.Position;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.TextRange;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Page;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.SemanticNode;
@@ -145,10 +146,7 @@ public class AtomicTextBlock implements TextBlock {
}
public static AtomicTextBlock fromAtomicTextBlockData(DocumentTextData documentTextData,
DocumentPositionData documentPositionData,
SemanticNode parent,
Page page) {
public static AtomicTextBlock fromAtomicTextBlockData(DocumentTextData documentTextData, DocumentPositionData documentPositionData, SemanticNode parent, Page page) {
return AtomicTextBlock.builder()
.id(documentTextData.getId())
@@ -156,9 +154,9 @@ public class AtomicTextBlock implements TextBlock {
.page(page)
.textRange(new TextRange(documentTextData.getStart(), documentTextData.getEnd()))
.searchText(documentTextData.getSearchText())
.lineBreaks(Arrays.stream(documentTextData.getLineBreaks()).boxed().toList())
.stringIdxToPositionIdx(Arrays.stream(documentPositionData.getStringIdxToPositionIdx()).boxed().toList())
.positions(toRectangle2DList(documentPositionData.getPositions()))
.lineBreaks(documentTextData.getLineBreaksList())
.stringIdxToPositionIdx(documentPositionData.getStringIdxToPositionIdxList())
.positions(toRectangle2DList(documentPositionData.getPositionsList()))
.parent(parent)
.build();
}
@@ -166,7 +164,17 @@ public class AtomicTextBlock implements TextBlock {
private static List<Rectangle2D> toRectangle2DList(float[][] positions) {
return Arrays.stream(positions).map(floatArr -> (Rectangle2D) new Rectangle2D.Float(floatArr[0], floatArr[1], floatArr[2], floatArr[3])).toList();
return Arrays.stream(positions)
.map(floatArr -> (Rectangle2D) new Rectangle2D.Float(floatArr[0], floatArr[1], floatArr[2], floatArr[3]))
.toList();
}
private static List<Rectangle2D> toRectangle2DList(List<Position> positions) {
return positions.stream()
.map(pos -> (Rectangle2D) new Rectangle2D.Float(pos.getValue(0), pos.getValue(1), pos.getValue(2), pos.getValue(3)))
.toList();
}
@@ -176,6 +184,9 @@ public class AtomicTextBlock implements TextBlock {
throw new IndexOutOfBoundsException(format("line %d out of range for AtomicTextBlock with %d lines", lineNumber, numberOfLines()));
}
if (lineNumber == 0) {
if (lineBreaks.isEmpty()) {
return searchText;
}
return subSequence(textRange.start(), lineBreaks.get(0) + textRange.start());
} else if (lineNumber == numberOfLines() - 1) {
return subSequence(lineBreaks.get(lineBreaks.size() - 1) + textRange.start(), textRange.end());
@@ -195,9 +206,9 @@ public class AtomicTextBlock implements TextBlock {
public int getNextLinebreak(int fromIndex) {
return lineBreaks.stream()//
.filter(linebreak -> linebreak > fromIndex - textRange.start()) //
.findFirst() //
.orElse(searchText.length()) + textRange.start();
.filter(linebreak -> linebreak > fromIndex - textRange.start()) //
.findFirst() //
.orElse(searchText.length()) + textRange.start();
}
@@ -205,9 +216,9 @@ public class AtomicTextBlock implements TextBlock {
public int getPreviousLinebreak(int fromIndex) {
return lineBreaks.stream()//
.filter(linebreak -> linebreak <= fromIndex - textRange.start())//
.reduce((a, b) -> b)//
.orElse(0) + textRange.start();
.filter(linebreak -> linebreak <= fromIndex - textRange.start())//
.reduce((a, b) -> b)//
.orElse(0) + textRange.start();
}
@@ -255,7 +266,10 @@ public class AtomicTextBlock implements TextBlock {
protected List<Integer> getAllLineBreaksInBoundary(TextRange textRange) {
return getLineBreaks().stream().map(linebreak -> linebreak + this.textRange.start()).filter(textRange::contains).toList();
return getLineBreaks().stream()
.map(linebreak -> linebreak + this.textRange.start())
.filter(textRange::contains)
.toList();
}
@@ -10,8 +10,8 @@ import lombok.NonNull;
import lombok.RequiredArgsConstructor;
@Data
@RequiredArgsConstructor
@AllArgsConstructor
@RequiredArgsConstructor
public class ClassifiedImage {
@NonNull
@@ -20,11 +20,18 @@ public class ClassifiedImage {
private ImageType imageType;
private boolean sourceByAi;
private boolean isAppendedToSection;
@NonNull
private boolean hasTransparency;
@NonNull
private int page;
@NonNull
private String representation;
public ClassifiedImage(@NonNull Rectangle2D position, @NonNull ImageType imageType, boolean hasTransparency, int page, String representation) {
this.position = position;
this.imageType = imageType;
this.hasTransparency = hasTransparency;
this.page = page;
this.representation = representation;
}
}
@@ -1,5 +1,6 @@
package com.knecon.fforesight.service.layoutparser.processor.model.outline;
import java.awt.geom.AffineTransform;
import java.awt.geom.Point2D;
import java.io.IOException;
import java.util.ArrayList;
@@ -26,6 +27,9 @@ import org.apache.pdfbox.pdmodel.interactive.documentnavigation.outline.PDDocume
import org.apache.pdfbox.pdmodel.interactive.documentnavigation.outline.PDOutlineItem;
import org.springframework.stereotype.Service;
import com.knecon.fforesight.service.layoutparser.processor.utils.CoordinateTransforms;
import com.knecon.fforesight.service.layoutparser.processor.utils.PageInformation;
import lombok.SneakyThrows;
import lombok.extern.slf4j.Slf4j;
@@ -82,11 +86,20 @@ public class OutlineExtractorService {
String title = item.getTitle();
PDPage page = item.findDestinationPage(document);
if (page == null) {
PDPage page;
try {
// Can throw: "Error: can't convert to Destination COSArray" for some OCR'd PDFs
page = item.findDestinationPage(document);
if (page == null) {
return Optional.empty();
}
} catch (IOException e) {
log.info(String.format("Error occurred during position resolution for outline item with title %s: " + e, title));
return Optional.empty();
}
int pageNumber = document.getPages().indexOf(page);
int pageNumber = document.getPages().indexOf(page) + 1;
AffineTransform userSpaceToPageCoords = CoordinateTransforms.calculateInitialUserSpaceCoordsToPageCoords(PageInformation.fromPDPage(pageNumber, page));
Optional<Point2D> outlinePosition = Optional.empty();
@@ -115,8 +128,15 @@ public class OutlineExtractorService {
log.info(String.format("Error occurred during position resolution for outline item on page %s with title %s: " + e, pageNumber, title));
}
return Optional.of(new OutlineObjectTreeNode(new OutlineObject(title, pageNumber, outlinePosition.orElse(new Point2D.Float(0, 0)), depth)));
return Optional.of(new OutlineObjectTreeNode(new OutlineObject(title,
pageNumber,
transformPointToPageCoords(outlinePosition, userSpaceToPageCoords), depth)));
}
private static Point2D transformPointToPageCoords(Optional<Point2D> outlinePosition, AffineTransform userSpaceToPageCoords) {
return outlinePosition.map(point -> userSpaceToPageCoords.transform(point, null)).orElse(null);
}
@@ -1,27 +1,34 @@
package com.knecon.fforesight.service.layoutparser.processor.model.outline;
import java.awt.geom.Point2D;
import java.util.Optional;
import lombok.AllArgsConstructor;
import lombok.Data;
import lombok.RequiredArgsConstructor;
import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.BoundingBox;
import lombok.Getter;
import lombok.Setter;
@Data
@RequiredArgsConstructor
@AllArgsConstructor
public class OutlineObject {
@Getter
private final String title;
@Getter
private final int pageNumber;
private Point2D point;
@Getter
private final int treeDepth;
private Point2D point; // java coordinates, (0, 0) is always top left
@Getter
@Setter
private boolean found;
public OutlineObject(String title, int pageNumber, Point2D point2D, int depth) {
this(title, pageNumber, depth);
this.title = title;
this.pageNumber = pageNumber;
this.treeDepth = depth;
this.point = point2D;
}
@@ -32,4 +39,39 @@ public class OutlineObject {
return "OutlineObject{" + "title='" + title + '\'' + '}';
}
public Optional<Point2D> getPoint() {
return Optional.ofNullable(point);
}
public boolean isAbove(BoundingBox boundingBox) {
if (point == null) {
return true;
}
return point.getY() <= boundingBox.getMaxY();
}
public double distance(BoundingBox boundingBox) {
if (point == null) {
return 0;
}
if (boundingBox.getBBox().contains(point)) {
return 0;
}
double deltaX = Math.min(Math.abs(boundingBox.getMinX() - point.getX()), Math.abs(boundingBox.getMaxX() - point.getX()));
double deltaY = Math.min(Math.abs(boundingBox.getMinY() - point.getY()), Math.abs(boundingBox.getMaxY() - point.getY()));
return Math.sqrt(deltaX * deltaX + deltaY * deltaY);
}
public void resetPoint() {
this.point = null;
}
}
@@ -39,4 +39,28 @@ public class OutlineObjectTree {
}
}
@Override
public String toString() {
StringBuilder sb = new StringBuilder();
sb.append("OutlineObjectTree(\n");
for (OutlineObjectTreeNode node : rootNodes) {
buildString(node, sb, 1);
}
sb.append(")");
return sb.toString();
}
private void buildString(OutlineObjectTreeNode node, StringBuilder sb, int depth) {
for (int i = 0; i < depth; i++) {
sb.append(" ");
}
sb.append(node.getOutlineObject().getTitle()).append("\n");
for (OutlineObjectTreeNode child : node.getChildren()) {
buildString(child, sb, depth + 1);
}
}
}
@@ -10,6 +10,7 @@ import java.util.TreeSet;
import org.springframework.stereotype.Service;
import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationDocument;
import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
import io.micrometer.observation.annotation.Observed;
@@ -20,7 +21,9 @@ import lombok.extern.slf4j.Slf4j;
public class OutlineValidationService {
@Observed(name = "OutlineValidationService", contextualName = "create-toc")
public TableOfContents createToC(List<TextPageBlock> headlines) {
public TableOfContents createToC(ClassificationDocument classificationDocument) {
List<TextPageBlock> headlines = extractHeadlines(classificationDocument);
List<TableOfContentItem> mainSections = new ArrayList<>();
Map<Integer, TableOfContentItem> lastItemsPerDepth = new HashMap<>();
@@ -60,4 +63,16 @@ public class OutlineValidationService {
return new TableOfContents(mainSections);
}
private static List<TextPageBlock> extractHeadlines(ClassificationDocument classificationDocument) {
return classificationDocument.getPages()
.stream()
.flatMap(classificationPage -> classificationPage.getTextBlocks()
.stream()
.filter(tb -> tb instanceof TextPageBlock && tb.getClassification() != null && tb.getClassification().isHeadline())
.map(tb -> (TextPageBlock) tb))
.toList();
}
}
@@ -1,6 +1,7 @@
package com.knecon.fforesight.service.layoutparser.processor.model.outline;
import java.util.ArrayList;
import java.util.Collection;
import java.util.Collections;
import java.util.Iterator;
import java.util.List;
@@ -167,7 +168,7 @@ public class TOCEnrichmentService {
}
}
if (!startBlocks.isEmpty()) {
if (!startBlocks.isEmpty() || !startImages.isEmpty()) {
TableOfContentItem unassigned = new TableOfContentItem(null);
unassigned.setSectionBlocks(startBlocks);
unassigned.setImages(startImages);
@@ -185,12 +186,8 @@ public class TOCEnrichmentService {
List<Cell> previousTableNonHeaderRow = getRowWithNonHeaderCells(previousTable);
List<Cell> tableNonHeaderRow = getRowWithNonHeaderCells(currentTable);
// Allow merging of tables if header row is separated from first logical non-header row
if (previousTableNonHeaderRow.isEmpty()
&& previousTable.getRowCount() == 1
&& previousTable.getRows()
.get(0).size() == tableNonHeaderRow.size()) {
previousTableNonHeaderRow = previousTable.getRows()
.get(0)
if (previousTableNonHeaderRow.isEmpty() && previousTable.getRowCount() == 1 && previousTable.getRows().get(0).size() == tableNonHeaderRow.size()) {
previousTableNonHeaderRow = previousTable.getRows().get(0)
.stream()
.map(cell -> {
Cell fakeCell = Cell.copy(cell);
@@ -201,8 +198,7 @@ public class TOCEnrichmentService {
}
if (previousTableNonHeaderRow.size() == tableNonHeaderRow.size()) {
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);
List<Cell> row = currentTable.getRows().get(i);
if (row.size() == tableNonHeaderRow.size() && row.stream()
.allMatch(cell -> cell.getHeaderCells().isEmpty())) {
for (int j = 0; j < row.size(); j++) {
@@ -225,18 +221,15 @@ public class TOCEnrichmentService {
return table.getRows()
.stream()
.flatMap(row -> row.stream()
.filter(cell -> !cell.getHeaderCells().isEmpty()))
.findAny().isEmpty();
.flatMap(Collection::stream)
.allMatch(cell -> cell.getHeaderCells().isEmpty());
}
private List<Cell> getRowWithNonHeaderCells(TablePageBlock 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);
List<Cell> row = table.getRows().get(i);
if (row.size() == 1) {
continue;
}
@@ -5,7 +5,7 @@ import java.util.List;
import java.util.stream.Collectors;
import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.AbstractSemanticNode;
import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.GenericSemanticNode;
import com.knecon.fforesight.service.layoutparser.processor.model.image.ClassifiedImage;
import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
@@ -24,7 +24,7 @@ public class TableOfContentItem {
private List<AbstractPageBlock> sectionBlocks = new ArrayList<>();
private List<ClassifiedImage> images = new ArrayList<>();
private AbstractSemanticNode section;
private GenericSemanticNode section;
public TableOfContentItem(TextPageBlock headline) {
@@ -45,8 +45,7 @@ public class TableOfContentItem {
if (parent != null) {
int index = parent.getChildren().indexOf(this);
if (index > 0) {
return parent.getChildren()
.get(index - 1);
return parent.getChildren().get(index - 1);
}
}
return null;
@@ -58,8 +57,7 @@ public class TableOfContentItem {
if (parent != null) {
int index = parent.getChildren().indexOf(this);
if (index >= 0 && index < parent.getChildren().size() - 1) {
return parent.getChildren()
.get(index + 1);
return parent.getChildren().get(index + 1);
}
}
return null;
@@ -93,17 +91,19 @@ public class TableOfContentItem {
return false;
}
public List<AbstractPageBlock> getNonEmptySectionBlocks() {
return sectionBlocks.stream().filter(pageBlock -> !pageBlock.isEmpty()).collect(Collectors.toList());
return sectionBlocks.stream()
.filter(pageBlock -> !pageBlock.isEmpty())
.collect(Collectors.toList());
}
@Override
public String toString() {
return "OutlineObjectTreeNode{" + "textPageBlock=" + headline + '}';
}
}
@@ -9,7 +9,7 @@ import java.util.List;
import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.BoundingBox;
import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;
import com.knecon.fforesight.service.layoutparser.processor.utils.TextNormalizationUtilities;
import lombok.Data;
@@ -35,14 +35,14 @@ public class Cell extends BoundingBox {
public Cell(Point2D topLeft, Point2D bottomRight) {
this.bBoxInitialUserSpace = new Rectangle2D.Double(topLeft.getX(), topLeft.getY(), (bottomRight.getX() - topLeft.getX()), (bottomRight.getY() - topLeft.getY()));
this.bBox = bBoxInitialUserSpace;
this.bBoxPdf = new Rectangle2D.Double(topLeft.getX(), topLeft.getY(), (bottomRight.getX() - topLeft.getX()), (bottomRight.getY() - topLeft.getY()));
this.bBox = bBoxPdf;
}
public Cell(Rectangle2D bBoxInitialUserSpace, AffineTransform initialUserSpaceToJava) {
this.bBoxInitialUserSpace = bBoxInitialUserSpace;
this.bBoxPdf = bBoxInitialUserSpace;
this.bBox = initialUserSpaceToJava.createTransformedShape(bBoxInitialUserSpace).getBounds2D();
}
@@ -50,7 +50,7 @@ public class Cell extends BoundingBox {
public static Cell copy(Cell cell) {
Cell copy = new Cell();
copy.bBoxInitialUserSpace = cell.bBoxInitialUserSpace;
copy.bBoxPdf = cell.bBoxPdf;
copy.bBox = cell.bBox;
return copy;
}
@@ -68,12 +68,12 @@ public class Cell extends BoundingBox {
StringBuilder sb = new StringBuilder();
Iterator<TextPageBlock> itty = textBlocks.iterator();
TextPositionSequence previous = null;
Word previous = null;
while (itty.hasNext()) {
TextPageBlock textBlock = itty.next();
for (TextPositionSequence word : textBlock.getSequences()) {
for (Word word : textBlock.getWords()) {
if (previous != null) {
if (Math.abs(previous.getMaxYDirAdj() - word.getMaxYDirAdj()) > word.getTextHeight()) {
sb.append('\n');
@@ -87,7 +87,7 @@ public class Cell extends BoundingBox {
}
return TextNormalizationUtilities.removeHyphenLineBreaks(sb.toString()).replaceAll("\n", " ").replaceAll(" {2}", " ");
return TextNormalizationUtilities.cleanString(sb.toString());
}
@@ -70,7 +70,7 @@ public class CleanRulings {
public boolean lineBetween(BoundingBox a, BoundingBox b) {
return lineBetween(a.getBBoxInitialUserSpace(), b.getBBoxInitialUserSpace());
return lineBetween(a.getBBoxPdf(), b.getBBoxPdf());
}
@@ -263,8 +263,8 @@ public class TablePageBlock extends AbstractPageBlock {
cells.stream()
.map(originalCell -> new CellWithIntersection(originalCell,
RectangleTransformations.calculateIntersectedArea(cellFromGridStructure.getBBoxInitialUserSpace(),
originalCell.getBBoxInitialUserSpace())))
RectangleTransformations.calculateIntersectedArea(cellFromGridStructure.getBBoxPdf(),
originalCell.getBBoxPdf())))
.filter(cellWithIntersection -> cellWithIntersection.intersectedArea > 0)
.filter(cellWithIntersection -> cellWithIntersection.originalCell.getArea() > cellWithIntersection.intersectedArea * CELL_AREA_CONTAINED_THRESHOLD)
.max(Comparator.comparing(CellWithIntersection::intersectedArea))
@@ -0,0 +1,8 @@
package com.knecon.fforesight.service.layoutparser.processor.model.text;
import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
public record AbstractBlockOnPage(AbstractPageBlock block, ClassificationPage page) {
}
@@ -0,0 +1,9 @@
package com.knecon.fforesight.service.layoutparser.processor.model.text;
public enum FontStyle {
REGULAR,
BOLD,
ITALIC,
BOLD_ITALIC;
}
@@ -0,0 +1,21 @@
package com.knecon.fforesight.service.layoutparser.processor.model.text;
import com.knecon.fforesight.service.layoutparser.processor.model.FloatFrequencyCounter;
import lombok.AccessLevel;
import lombok.Getter;
import lombok.NoArgsConstructor;
import lombok.experimental.FieldDefaults;
@Getter
@NoArgsConstructor
@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
public class FrequencyCounters {
FloatFrequencyCounter lineHeightFrequencyCounter = new FloatFrequencyCounter();
FloatFrequencyCounter fontSizeFrequencyCounter = new FloatFrequencyCounter();
FloatFrequencyCounter spaceFrequencyCounter = new FloatFrequencyCounter();
StringFrequencyCounter fontFrequencyCounter = new StringFrequencyCounter();
StringFrequencyCounter styleFrequencyCounter = new StringFrequencyCounter();
}
@@ -0,0 +1,107 @@
package com.knecon.fforesight.service.layoutparser.processor.model.text;
import java.util.List;
import java.util.Optional;
import java.util.regex.Matcher;
import java.util.regex.Pattern;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Getter;
import lombok.experimental.FieldDefaults;
@AllArgsConstructor
@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
public class ListIdentifier {
public static final Pattern NUMBER_WITH_DOT = Pattern.compile("^\\s*([1-9]{1,4})\\.\\s+");
public static final Pattern NUMBER_IN_PARENTHESES = Pattern.compile("^\\s*\\(([1-9]{1,4})\\)\\s+");
enum Format {
NUMBER_WITH_DOT,
NUMBER_IN_PARENTHESES
}
Format format;
@Getter
Word word;
@Getter
int page;
int representation;
public static Optional<ListIdentifier> parse(TextPageBlock textPageBlock, int page) {
return parse(textPageBlock.getWords().subList(0, Math.min(5, textPageBlock.getWords().size())), page);
}
public static Optional<ListIdentifier> parse(List<Word> sequences, int page) {
StringBuilder sb = new StringBuilder();
for (Word sequence : sequences) {
sb.append(sequence.toString());
sb.append(" ");
}
sb.replace(sb.length() - 1, sb.length(), "");
String text = sb.toString();
Matcher numberMatcher = NUMBER_WITH_DOT.matcher(text);
if (numberMatcher.find()) {
Optional<Integer> representation = parseInteger(numberMatcher.group(1));
if (representation.isPresent()) {
return Optional.of(new ListIdentifier(Format.NUMBER_WITH_DOT, sequences.get(0), page, representation.get()));
}
}
Matcher parenthesisMatcher = NUMBER_IN_PARENTHESES.matcher(text);
if (parenthesisMatcher.find()) {
Optional<Integer> representation = parseInteger(parenthesisMatcher.group(1));
if (representation.isPresent()) {
return Optional.of(new ListIdentifier(Format.NUMBER_IN_PARENTHESES, sequences.get(0), page, representation.get()));
}
}
return Optional.empty();
}
private static Optional<Integer> parseInteger(String text) {
try {
return Optional.of(Integer.parseInt(text));
} catch (NumberFormatException e) {
return Optional.empty();
}
}
public static boolean isInOrder(List<ListIdentifier> listIdentifiers) {
if (listIdentifiers.size() <= 1) {
return true;
}
for (int i = 1; i < listIdentifiers.size(); i++) {
ListIdentifier current = listIdentifiers.get(i);
ListIdentifier previous = listIdentifiers.get(i - 1);
if (current.format != previous.format) {
return false;
}
if (current.representation <= previous.representation) {
return false;
}
if (!current.word.intersectsXDirAdj(previous.word, 2)) {
return false;
}
if (current.page == previous.page && !current.word.isBelowDirAdj(previous.word)) {
return false;
}
if (current.page < previous.page) {
return false;
}
}
return true;
}
}
@@ -5,64 +5,52 @@ import java.awt.geom.Rectangle2D;
import org.apache.pdfbox.text.TextPosition;
import com.fasterxml.jackson.annotation.JsonIgnore;
import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.BoundingBox;
import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.TextBoundingBox;
import com.knecon.fforesight.service.layoutparser.processor.docstrum.utils.FastAtan2;
import com.knecon.fforesight.service.layoutparser.processor.utils.CoordinateTransforms;
import com.knecon.fforesight.service.layoutparser.processor.utils.PageInformation;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
import lombok.EqualsAndHashCode;
import lombok.NoArgsConstructor;
import lombok.SneakyThrows;
import lombok.experimental.FieldDefaults;
import lombok.experimental.SuperBuilder;
@Data
@Builder
@SuperBuilder
@NoArgsConstructor
@AllArgsConstructor
public class RedTextPosition extends BoundingBox {
@EqualsAndHashCode(callSuper = true)
@FieldDefaults(level = AccessLevel.PRIVATE)
public class RedTextPosition extends TextBoundingBox {
public final static int HEIGHT_PADDING = 2;
private Rectangle2D.Float bBoxDirAdj; // adjusted to text rotation
String unicode;
@JsonIgnore
private int rotation;
// estimated using the TextMatrix in radians
float exactDir;
@JsonIgnore
private float pageHeight;
float widthOfSpace;
@JsonIgnore
private float pageWidth;
float fontSizeInPt;
private String unicode;
@JsonIgnore
private float dir;
// not used in reanalysis
@JsonIgnore
private float widthOfSpace;
// not used in reanalysis
@JsonIgnore
private float fontSizeInPt;
// not used in reanalysis
@JsonIgnore
private String fontName;
String fontName;
@SneakyThrows
public static RedTextPosition fromTextPosition(TextPosition textPosition) {
var pos = new RedTextPosition();
pos.setRotation(textPosition.getRotation());
pos.setPageHeight(textPosition.getPageHeight());
pos.setPageWidth(textPosition.getPageWidth());
pos.setUnicode(textPosition.getUnicode());
pos.setDir(textPosition.getDir());
pos.setWidthOfSpace(textPosition.getWidthOfSpace());
pos.setFontSizeInPt(textPosition.getFontSizeInPt());
pos.setFontName(textPosition.getFont().getName());
pos.setExactDir((float) FastAtan2.fastAtan2(textPosition.getTextMatrix().getShearY(), textPosition.getTextMatrix().getScaleX()));
pos.setDir(TextDirection.fromDegrees(textPosition.getDir()));
//TODO: There is a mismatch in the java coords of the text and the rulings,
// I guess if we start with the initial user space positions and transform them the same way we do the rulings it would work.
@@ -73,18 +61,18 @@ public class RedTextPosition extends BoundingBox {
textPosition.getYDirAdj() - textHeight,
textPosition.getWidthDirAdj(),
textHeight + HEIGHT_PADDING);
pos.setBBoxDirAdj(dirAdjPosition);
AffineTransform affineTransform = getRotationMatrix(TextDirection.fromDegrees(textPosition.getDir()), textPosition.getPageWidth(), textPosition.getPageHeight());
Rectangle2D bBoxInitialUserSpace = affineTransform.createTransformedShape(dirAdjPosition).getBounds2D();
pos.setBBoxInitialUserSpace(bBoxInitialUserSpace); // These are definitely correct
pos.setBBoxPdf(bBoxInitialUserSpace); // These are definitely correct
return pos;
}
private static AffineTransform getRotationMatrix(TextDirection textDirection, float pageWidth, float pageHeight) {
AffineTransform transform = new AffineTransform();
@@ -103,32 +91,4 @@ public class RedTextPosition extends BoundingBox {
return transform;
}
@JsonIgnore
public float getXDirAdj() {
return this.bBoxDirAdj.x;
}
@JsonIgnore
public float getYDirAdj() {
return this.bBoxDirAdj.y;
}
@JsonIgnore
public float getWidthDirAdj() {
return this.bBoxDirAdj.width;
}
@JsonIgnore
public float getHeightDir() {
return this.bBoxDirAdj.height;
}
}
@@ -2,6 +2,7 @@ package com.knecon.fforesight.service.layoutparser.processor.model.text;
import java.util.ArrayList;
import java.util.List;
import com.knecon.fforesight.service.layoutparser.processor.utils.TextNormalizationUtilities;
import lombok.Getter;
@@ -9,18 +10,18 @@ import lombok.Getter;
@Getter
public class SearchableText {
private final List<TextPositionSequence> sequences = new ArrayList<>();
private final List<Word> sequences = new ArrayList<>();
public void add(TextPositionSequence textPositionSequence) {
public void add(Word word) {
sequences.add(textPositionSequence);
sequences.add(word);
}
public void addAll(List<TextPositionSequence> textPositionSequences) {
public void addAll(List<Word> words) {
sequences.addAll(textPositionSequences);
sequences.addAll(words);
}
@@ -31,18 +32,14 @@ public class SearchableText {
}
public static String buildString(List<TextPositionSequence> sequences) {
public static String buildString(List<Word> sequences) {
StringBuilder sb = new StringBuilder();
for (TextPositionSequence word : sequences) {
for (Word word : sequences) {
sb.append(word);
sb.append(' ');
}
String text = sb.toString();
text = TextNormalizationUtilities.removeHyphenLineBreaks(text);
text = TextNormalizationUtilities.removeLineBreaks(text);
text = TextNormalizationUtilities.removeRepeatingWhitespaces(text);
return text;
return TextNormalizationUtilities.cleanString(sb.toString());
}
}
@@ -9,10 +9,14 @@ public class StringFrequencyCounter {
@Getter
private final Map<String, Integer> countPerValue = new HashMap<>();
boolean changed;
String mostPopularCache;
public void add(String value) {
changed = true;
if (!countPerValue.containsKey(value)) {
countPerValue.put(value, 1);
} else {
@@ -23,6 +27,8 @@ public class StringFrequencyCounter {
public void addAll(Map<String, Integer> otherCounter) {
changed = true;
for (Map.Entry<String, Integer> entry : otherCounter.entrySet()) {
if (countPerValue.containsKey(entry.getKey())) {
countPerValue.put(entry.getKey(), countPerValue.get(entry.getKey()) + entry.getValue());
@@ -35,13 +41,18 @@ public class StringFrequencyCounter {
public String getMostPopular() {
Map.Entry<String, Integer> mostPopular = null;
for (Map.Entry<String, Integer> entry : countPerValue.entrySet()) {
if (mostPopular == null || entry.getValue() > mostPopular.getValue()) {
mostPopular = entry;
if (changed || mostPopularCache == null) {
Map.Entry<String, Integer> mostPopular = null;
for (Map.Entry<String, Integer> entry : countPerValue.entrySet()) {
if (mostPopular == null || entry.getValue() >= mostPopular.getValue()) {
mostPopular = entry;
}
}
mostPopularCache = mostPopular != null ? mostPopular.getKey() : null;
changed = false;
}
return mostPopular != null ? mostPopular.getKey() : null;
return mostPopularCache;
}
}
@@ -0,0 +1,7 @@
package com.knecon.fforesight.service.layoutparser.processor.model.text;
import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
public record TextBlockOnPage(TextPageBlock textBlock, ClassificationPage page) {
}
@@ -44,4 +44,15 @@ public enum TextDirection {
throw new IllegalArgumentException(String.format("A value of %f is not supported by TextDirection", degrees));
}
public int getRotation() {
return switch (this) {
case ZERO -> 0;
case QUARTER_CIRCLE -> 1;
case HALF_CIRCLE -> 2;
case THREE_QUARTER_CIRCLE -> 3;
};
}
}
@@ -2,12 +2,13 @@ package com.knecon.fforesight.service.layoutparser.processor.model.text;
import java.awt.geom.Rectangle2D;
import java.util.ArrayList;
import java.util.Collections;
import java.util.List;
import com.fasterxml.jackson.annotation.JsonIgnore;
import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
import com.knecon.fforesight.service.layoutparser.processor.model.FloatFrequencyCounter;
import com.knecon.fforesight.service.layoutparser.processor.model.PageBlockType;
import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
import com.knecon.fforesight.service.layoutparser.processor.utils.TextNormalizationUtilities;
import lombok.AllArgsConstructor;
@@ -24,62 +25,59 @@ import lombok.NoArgsConstructor;
public class TextPageBlock extends AbstractPageBlock {
@Builder.Default
private List<TextPositionSequence> sequences = new ArrayList<>();
private List<Word> words = new ArrayList<>();
@Builder.Default
private FrequencyCounters frequencyCounters = new FrequencyCounters();
private String mostPopularWordFont;
private Rectangle2D bBoxDirAdj;
private String mostPopularWordStyle;
private float mostPopularWordFontSize;
private float mostPopularWordHeight;
private float mostPopularWordSpaceWidth;
private float highestFontSize;
private boolean underlined;
private PageBlockType classification;
private boolean toDuplicate;
private String text;
private boolean changed;
public TextPageBlock(List<TextPositionSequence> sequences) {
this.sequences = sequences;
calculateFrequencyCounters();
public TextPageBlock(List<Word> words) {
this.words = new ArrayList<>(words);
this.frequencyCounters = new FrequencyCounters();
if (!words.isEmpty()) {
addToFrequencyCounters(words);
}
calculateBBox();
}
@JsonIgnore
public List<Word> getWords() {
return Collections.unmodifiableList(words);
}
public TextDirection getDir() {
return sequences.get(0).getDir();
}
@JsonIgnore
public float getPageHeight() {
return sequences.get(0).getPageHeight();
}
@JsonIgnore
public float getPageWidth() {
return sequences.get(0).getPageWidth();
return words.get(0).getDir();
}
private void calculateBBox() {
if (sequences == null) {
if (words == null) {
this.bBox = new Rectangle2D.Double();
this.bBoxInitialUserSpace = new Rectangle2D.Double();
this.bBoxPdf = new Rectangle2D.Double();
this.bBoxDirAdj = new Rectangle2D.Double();
return;
}
setToBBoxOfComponents(sequences);
this.bBoxDirAdj = words.stream()
.map(Word::getBBoxDirAdj)
.collect(RectangleTransformations.collectBBox());
setToBBoxOfComponents(words);
}
@@ -101,8 +99,8 @@ public class TextPageBlock extends AbstractPageBlock {
throw new IllegalArgumentException("Cannot merge textBlocks on different pages.");
}
List<TextPositionSequence> sequences = textBlocksToMerge.stream()
.map(TextPageBlock::getSequences)
List<Word> sequences = textBlocksToMerge.stream()
.map(TextPageBlock::getWords)
.flatMap(java.util.Collection::stream)
.toList();
sequences = new ArrayList<>(sequences);
@@ -111,38 +109,27 @@ public class TextPageBlock extends AbstractPageBlock {
}
private void calculateFrequencyCounters() {
private void addToFrequencyCounters(List<Word> sequences) {
FloatFrequencyCounter lineHeightFrequencyCounter = new FloatFrequencyCounter();
FloatFrequencyCounter fontSizeFrequencyCounter = new FloatFrequencyCounter();
FloatFrequencyCounter spaceFrequencyCounter = new FloatFrequencyCounter();
StringFrequencyCounter fontFrequencyCounter = new StringFrequencyCounter();
StringFrequencyCounter styleFrequencyCounter = new StringFrequencyCounter();
for (TextPositionSequence wordBlock : sequences) {
lineHeightFrequencyCounter.add(wordBlock.getTextHeight());
fontSizeFrequencyCounter.add(wordBlock.getFontSize());
spaceFrequencyCounter.add(wordBlock.getSpaceWidth());
fontFrequencyCounter.add(wordBlock.getFont());
styleFrequencyCounter.add(wordBlock.getFontStyle());
for (Word wordBlock : sequences) {
frequencyCounters.getLineHeightFrequencyCounter().add(wordBlock.getTextHeight());
frequencyCounters.getFontSizeFrequencyCounter().add(wordBlock.getFontSize());
frequencyCounters.getSpaceFrequencyCounter().add(wordBlock.getSpaceWidth());
frequencyCounters.getFontFrequencyCounter().add(wordBlock.getFont());
frequencyCounters.getStyleFrequencyCounter().add(wordBlock.getFontStyle());
}
setMostPopularWordFont(fontFrequencyCounter.getMostPopular());
setMostPopularWordStyle(styleFrequencyCounter.getMostPopular());
setMostPopularWordFontSize(fontSizeFrequencyCounter.getMostPopular());
setMostPopularWordHeight(lineHeightFrequencyCounter.getMostPopular());
setMostPopularWordSpaceWidth(spaceFrequencyCounter.getMostPopular());
setHighestFontSize(fontSizeFrequencyCounter.getHighest());
setUnderlined(this.words.stream()
.allMatch(Word::isUnderline));
}
public TextPageBlock union(TextPositionSequence r) {
public TextPageBlock union(Word r) {
TextPageBlock union = this.copy();
union.getSequences().add(r);
calculateFrequencyCounters();
union.add(r);
addToFrequencyCounters(List.of(r));
calculateBBox();
return union;
}
@@ -151,51 +138,50 @@ public class TextPageBlock extends AbstractPageBlock {
public TextPageBlock union(TextPageBlock r) {
TextPageBlock union = this.copy();
union.getSequences().addAll(r.getSequences());
calculateFrequencyCounters();
union.addAll(r.getWords());
addToFrequencyCounters(r.getWords());
calculateBBox();
return union;
}
public void add(TextPageBlock r) {
public void add(TextPageBlock textPageBlock) {
sequences.addAll(r.getSequences());
calculateFrequencyCounters();
changed = true;
words.addAll(textPageBlock.getWords());
addToFrequencyCounters(textPageBlock.getWords());
calculateBBox();
}
public void add(TextPositionSequence r) {
public void add(Word word) {
sequences.add(r);
calculateFrequencyCounters();
changed = true;
words.add(word);
addToFrequencyCounters(List.of(word));
calculateBBox();
}
public void addAll(List<Word> words) {
changed = true;
this.words.addAll(words);
addToFrequencyCounters(words);
calculateBBox();
}
public TextPageBlock copy() {
return new TextPageBlock(new ArrayList<>(sequences));
return new TextPageBlock(new ArrayList<>(words));
}
@Override
public String toString() {
StringBuilder builder = new StringBuilder();
for (int i = 0; i < sequences.size(); i++) {
String sequenceAsString = sequences.get(i).toString();
// Fix for missing Whitespace. This is recognized in getSequences method. See PDFTextStripper Line 1730.
if (i != 0 && sequences.get(i - 1).charAt(sequences.get(i - 1).length() - 1) != ' ' && sequenceAsString.charAt(0) != ' ') {
builder.append(' ');
}
builder.append(sequenceAsString);
}
return builder.toString();
return getText();
}
@@ -203,30 +189,36 @@ public class TextPageBlock extends AbstractPageBlock {
@JsonIgnore
public String getText() {
StringBuilder sb = new StringBuilder();
if (text == null || changed) {
TextPositionSequence previous = null;
for (TextPositionSequence word : sequences) {
if (previous != null) {
if (Math.abs(previous.getMaxYDirAdj() - word.getMaxYDirAdj()) > word.getTextHeight()) {
sb.append('\n');
} else {
sb.append(' ');
StringBuilder sb = new StringBuilder();
Word previous = null;
for (Word word : words) {
if (previous != null) {
if (Math.abs(previous.getMaxYDirAdj() - word.getMaxYDirAdj()) > word.getTextHeight()) {
sb.append('\n');
} else {
sb.append(' ');
}
}
sb.append(word.toString());
previous = word;
}
sb.append(word.toString());
previous = word;
text = TextNormalizationUtilities.removeHyphenLinebreaks(sb.toString());
changed = false;
}
return TextNormalizationUtilities.removeHyphenLineBreaks(sb.toString());
return text;
}
public int getNumberOfLines() {
int numberOfLines = 1;
TextPositionSequence previous = null;
for (TextPositionSequence word : sequences) {
Word previous = null;
for (Word word : words) {
if (previous != null) {
if (word.getMaxYDirAdj() - previous.getMaxYDirAdj() > word.getTextHeight()) {
numberOfLines++;
@@ -238,10 +230,47 @@ public class TextPageBlock extends AbstractPageBlock {
}
public String getMostPopularWordFont() {
return frequencyCounters.getFontFrequencyCounter().getMostPopular();
}
public String getMostPopularWordStyle() {
return frequencyCounters.getStyleFrequencyCounter().getMostPopular();
}
public double getMostPopularWordFontSize() {
return frequencyCounters.getFontSizeFrequencyCounter().getMostPopular();
}
public double getMostPopularWordHeight() {
return frequencyCounters.getLineHeightFrequencyCounter().getMostPopular();
}
public double getMostPopularWordSpaceWidth() {
return frequencyCounters.getSpaceFrequencyCounter().getMostPopular();
}
public double getHighestFontSize() {
Double highest = frequencyCounters.getFontSizeFrequencyCounter().getHighest();
return highest == null ? 0 : highest;
}
@Override
public boolean isEmpty() {
return sequences.isEmpty();
return words.isEmpty();
}
}
@@ -1,271 +0,0 @@
package com.knecon.fforesight.service.layoutparser.processor.model.text;
import java.awt.geom.Rectangle2D;
import java.util.ArrayList;
import java.util.List;
import java.util.Locale;
import java.util.stream.Collectors;
import org.apache.pdfbox.text.TextPosition;
import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.BoundingBox;
import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
import lombok.EqualsAndHashCode;
import lombok.NoArgsConstructor;
import lombok.extern.slf4j.Slf4j;
@Slf4j
@Data
@Builder
@NoArgsConstructor
@AllArgsConstructor
@EqualsAndHashCode(onlyExplicitlyIncluded = true, callSuper = false)
public class TextPositionSequence extends BoundingBox implements CharSequence {
public static final int HEIGHT_PADDING = 2;
@EqualsAndHashCode.Include
private int page;
@EqualsAndHashCode.Include
private List<RedTextPosition> textPositions = new ArrayList<>();
private Rectangle2D bBoxDirAdj;
@EqualsAndHashCode.Include
private TextDirection dir;
private int rotation;
private float pageHeight;
private float pageWidth;
private boolean isParagraphStart;
private boolean strikethrough;
private boolean underline;
public TextPositionSequence(List<TextPosition> textPositions, int pageNumber, boolean isParagraphStart) {
this.textPositions = textPositions.stream()
.map(RedTextPosition::fromTextPosition)
.collect(Collectors.toList());
this.page = pageNumber;
this.dir = TextDirection.fromDegrees(textPositions.get(0).getDir());
this.rotation = textPositions.get(0).getRotation();
this.pageHeight = textPositions.get(0).getPageHeight();
this.pageWidth = textPositions.get(0).getPageWidth();
this.isParagraphStart = isParagraphStart;
calculateBBox();
}
private void calculateBBox() {
this.bBoxDirAdj = textPositions.stream()
.map(RedTextPosition::getBBoxDirAdj)
.collect(RectangleTransformations.collectBBox());
setToBBoxOfComponents(getTextPositions());
}
public TextPositionSequence(List<RedTextPosition> textPositions, int page) {
this.textPositions = textPositions;
this.page = page;
this.dir = TextDirection.fromDegrees(textPositions.get(0).getDir());
this.rotation = textPositions.get(0).getRotation();
this.pageHeight = textPositions.get(0).getPageHeight();
this.pageWidth = textPositions.get(0).getPageWidth();
calculateBBox();
}
@Override
public int length() {
return textPositions.size();
}
@Override
public char charAt(int index) {
RedTextPosition textPosition = textPositionAt(index);
String text = textPosition.getUnicode();
return text.charAt(0);
}
public char charAt(int index, boolean caseInSensitive) {
RedTextPosition textPosition = textPositionAt(index);
String text = textPosition.getUnicode();
return caseInSensitive ? text.toLowerCase(Locale.ROOT).charAt(0) : text.charAt(0);
}
@Override
public TextPositionSequence subSequence(int start, int end) {
var textPositionSequence = new TextPositionSequence();
textPositionSequence.textPositions = textPositions.subList(start, end);
textPositionSequence.page = page;
textPositionSequence.dir = dir;
textPositionSequence.rotation = rotation;
textPositionSequence.pageHeight = pageHeight;
textPositionSequence.pageWidth = pageWidth;
textPositionSequence.setToBBoxOfComponents(getTextPositions());
return textPositionSequence;
}
@Override
public String toString() {
StringBuilder builder = new StringBuilder(length());
for (int i = 0; i < length(); i++) {
builder.append(charAt(i));
}
return builder.toString();
}
public RedTextPosition textPositionAt(int index) {
return textPositions.get(index);
}
public void add(TextPositionSequence textPositionSequence, RedTextPosition textPosition) {
this.textPositions.add(textPosition);
this.page = textPositionSequence.getPage();
this.dir = textPositionSequence.getDir();
this.rotation = textPositionSequence.getRotation();
this.pageHeight = textPositionSequence.getPageHeight();
this.pageWidth = textPositionSequence.getPageWidth();
calculateBBox();
}
public void add(TextPosition textPosition) {
this.textPositions.add(RedTextPosition.fromTextPosition(textPosition));
this.dir = TextDirection.fromDegrees(textPositions.get(0).getDir());
this.rotation = textPositions.get(0).getRotation();
this.pageHeight = textPositions.get(0).getPageHeight();
this.pageWidth = textPositions.get(0).getPageWidth();
calculateBBox();
}
/**
* This value is adjusted so that 0,0 is upper left and it is adjusted based on the text direction.
* This method ignores the page rotation but takes the text rotation and adjusts the coordinates to awt.
*
* @return the text direction adjusted minX value
*/
public float getMinXDirAdj() {
return textPositions.get(0).getXDirAdj();
}
/**
* This value is adjusted so that 0,0 is upper left and it is adjusted based on the text direction.
* This method ignores the page rotation but takes the text rotation and adjusts the coordinates to awt.
*
* @return the text direction adjusted maxX value
*/
public float getMaxXDirAdj() {
return textPositions.get(textPositions.size() - 1).getXDirAdj() + textPositions.get(textPositions.size() - 1).getWidthDirAdj() + HEIGHT_PADDING;
}
/**
* This value is adjusted so that 0,0 is upper left and it is adjusted based on the text direction.
* This method ignores the page rotation but takes the text rotation and adjusts the coordinates to awt.
*
* @return the text direction adjusted minY value. The upper border of the bounding box of the word.
*/
public float getMinYDirAdj() {
return textPositions.get(0).getYDirAdj() - getTextHeight();
}
/**
* This value is adjusted so that 0,0 is upper left and it is adjusted based on the text direction.
* This method ignores the page rotation but takes the text rotation and adjusts the coordinates to awt.
*
* @return the text direction adjusted maxY value. The lower border of the bounding box of the word.
*/
public float getMaxYDirAdj() {
return textPositions.get(0).getYDirAdj();
}
public float getTextHeightNoPadding() {
return textPositions.get(0).getHeightDir();
}
public float getTextHeight() {
return textPositions.get(0).getHeightDir() + HEIGHT_PADDING;
}
public String getFont() {
if (textPositions.get(0).getFontName() == null) {
return "none";
}
return textPositions.get(0).getFontName().toLowerCase(Locale.ROOT).replaceAll(",bold", "").replaceAll(",italic", "");
}
public String getFontStyle() {
if (textPositions.get(0).getFontName() == null) {
return "standard";
}
String lowercaseFontName = textPositions.get(0).getFontName().toLowerCase(Locale.ROOT);
if (lowercaseFontName.contains("bold") && lowercaseFontName.contains("italic")) {
return "bold, italic";
} else if (lowercaseFontName.contains("bold")) {
return "bold";
} else if (lowercaseFontName.contains("italic")) {
return "italic";
} else {
return "standard";
}
}
public float getFontSize() {
return textPositions.get(0).getFontSizeInPt();
}
public float getSpaceWidth() {
return textPositions.get(0).getWidthOfSpace();
}
}
@@ -0,0 +1,34 @@
package com.knecon.fforesight.service.layoutparser.processor.model.text;
import java.util.Comparator;
import java.util.HashMap;
public class TextPositionSequenceComparator implements Comparator<Word> {
private HashMap<Word, TextBlockOnPage> lookup;
public TextPositionSequenceComparator(HashMap<Word, TextBlockOnPage> lookup) {
this.lookup = lookup;
}
@Override
public int compare(Word number1, Word number2) {
int page1 = lookup.get(number1).page().getPageNumber();
int page2 = lookup.get(number2).page().getPageNumber();
if (page1 != page2) {
return Integer.compare(page1, page2);
}
if (number1.getY() != number2.getY()) {
return Double.compare(number1.getY(), number2.getY());
}
return Integer.compare(Integer.parseInt(number1.toString()), Integer.parseInt(number2.toString()));
}
}
@@ -0,0 +1,272 @@
package com.knecon.fforesight.service.layoutparser.processor.model.text;
import static com.knecon.fforesight.service.layoutparser.processor.model.text.RedTextPosition.HEIGHT_PADDING;
import java.awt.geom.AffineTransform;
import java.awt.geom.Rectangle2D;
import java.util.ArrayList;
import java.util.List;
import java.util.Locale;
import java.util.Objects;
import java.util.regex.Pattern;
import java.util.stream.Collectors;
import org.apache.pdfbox.text.TextPosition;
import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Character;
import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.TextBoundingBox;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
import lombok.NoArgsConstructor;
import lombok.extern.slf4j.Slf4j;
@Slf4j
@Data
@Builder
@NoArgsConstructor
@AllArgsConstructor
@SuppressWarnings("pmd")
public class Word extends TextBoundingBox implements CharSequence {
public static final String STANDARD = "standard";
public static final String BOLD_ITALIC = "bold, italic";
public static final String BOLD = "bold";
public static final String ITALIC = "italic";
public static final Pattern FONT_CLEANER = Pattern.compile(",bold|,italic");
private int page;
@Builder.Default
private List<Character> characters = new ArrayList<>();
private boolean isParagraphStart;
private boolean strikethrough;
private boolean underline;
private Integer hashcodeCache;
public Word(List<TextPosition> textPositions, int pageNumber, boolean isParagraphStart) {
this.characters = textPositions.stream()
.map(RedTextPosition::fromTextPosition)
.map(Character::new)
.collect(Collectors.toList());
this.page = pageNumber;
this.isParagraphStart = isParagraphStart;
calculateBBoxAndHashcode();
}
private void calculateBBoxAndHashcode() {
setToBBoxOfComponents(getTextPositions());
hashcodeCache = null;
}
public Word(List<Character> textPositions, int page) {
this.characters = new ArrayList<>(textPositions);
this.page = page;
calculateBBoxAndHashcode();
}
@Override
public int length() {
return characters.size();
}
@Override
public char charAt(int index) {
RedTextPosition textPosition = textPositionAt(index);
String text = textPosition.getUnicode();
return text.charAt(0);
}
public char charAt(int index, boolean caseInSensitive) {
RedTextPosition textPosition = textPositionAt(index);
String text = textPosition.getUnicode();
return caseInSensitive ? text.toLowerCase(Locale.ROOT).charAt(0) : text.charAt(0);
}
@Override
public Word subSequence(int start, int end) {
var textPositionSequence = new Word();
textPositionSequence.characters = characters.subList(start, end);
textPositionSequence.page = page;
textPositionSequence.dir = dir;
textPositionSequence.setToBBoxOfComponents(getTextPositions());
return textPositionSequence;
}
@Override
public String toString() {
StringBuilder builder = new StringBuilder(length());
for (int i = 0; i < length(); i++) {
builder.append(charAt(i));
}
return builder.toString();
}
public RedTextPosition textPositionAt(int index) {
return characters.get(index).getTextPosition();
}
public void add(Word word, RedTextPosition textPosition) {
this.characters.add(new Character(textPosition));
this.page = word.getPage();
calculateBBoxAndHashcode();
}
public void add(Character current) {
characters.add(current);
calculateBBoxAndHashcode();
}
public void add(TextPosition textPosition) {
add(new Character(RedTextPosition.fromTextPosition(textPosition)));
}
public double getTextHeightNoPadding() {
return characters.get(0).getTextPosition().getHeightDirAdj();
}
public double getTextHeight() {
return characters.get(0).getTextPosition().getHeightDirAdj() + HEIGHT_PADDING;
}
public String getFont() {
if (characters.get(0).getTextPosition().getFontName() == null) {
return "none";
}
return FONT_CLEANER.matcher(characters.get(0).getTextPosition().getFontName().toLowerCase(Locale.ROOT)).replaceAll("");
}
public String getFontStyle() {
if (characters.get(0).getTextPosition().getFontName() == null) {
return STANDARD;
}
String lowercaseFontName = characters.get(0).getTextPosition().getFontName().toLowerCase(Locale.ROOT);
if (lowercaseFontName.contains(BOLD) && lowercaseFontName.contains(ITALIC)) {
return BOLD_ITALIC;
} else if (lowercaseFontName.contains(BOLD)) {
return BOLD;
} else if (lowercaseFontName.contains(ITALIC)) {
return ITALIC;
} else {
return STANDARD;
}
}
public float getFontSize() {
return characters.get(0).getTextPosition().getFontSizeInPt();
}
public float getSpaceWidth() {
return characters.get(0).getTextPosition().getWidthOfSpace();
}
public boolean equals(final Object o) {
// auto-generated with lombok
if (o == this) {
return true;
}
if (!(o instanceof Word other)) {
return false;
}
if (!other.canEqual((Object) this)) {
return false;
}
if (!super.equals(o)) {
return false;
}
if (this.getPage() != other.getPage()) {
return false;
}
final Object this$textPositions = this.getTextPositions();
final Object other$textPositions = other.getTextPositions();
if (!Objects.equals(this$textPositions, other$textPositions)) {
return false;
}
return Objects.equals(this.getHashcodeCache(), other.getHashcodeCache());
}
protected boolean canEqual(final Object other) {return other instanceof Word;}
public int hashCode() {
if (hashcodeCache == null) {
hashcodeCache = hashcodeCalculation();
}
return hashcodeCache;
}
private int hashcodeCalculation() {
final int PRIME = 59;
int result = super.hashCode();
result = result * PRIME + this.getPage();
final Object $textPositions = this.getTextPositions();
result = result * PRIME + ($textPositions == null ? 43 : $textPositions.hashCode());
return result;
}
private List<RedTextPosition> getTextPositions() {
return characters.stream()
.map(Character::getTextPosition)
.toList();
}
public void transform(AffineTransform rotateInstance) {
for (RedTextPosition textPosition : getTextPositions()) {
Rectangle2D exactDirAdjCoordinates = rotateInstance.createTransformedShape(textPosition.getBBoxDirAdj()).getBounds2D();
textPosition.setBBoxDirAdj(exactDirAdjCoordinates);
}
calculateBBoxAndHashcode();
}
}
@@ -56,7 +56,7 @@ public class ImageServiceResponseAdapter {
classificationPage.getImages().forEach(image -> {
if (image.getImageType().equals(ImageType.OTHER)) {
for (AbstractPageBlock textblock : classificationPage.getTextBlocks()) {
if (image.getPosition().contains(textblock.getBBoxInitialUserSpace())) {
if (image.getPosition().contains(textblock.getBBoxPdf())) {
image.setImageType(ImageType.OCR);
return;
}
@@ -14,6 +14,7 @@ import lombok.NoArgsConstructor;
@AllArgsConstructor
public class Classification {
@Builder.Default
private Map<String, Float> probabilities = new HashMap<>();
private String label;
@@ -22,8 +22,10 @@ public class ImageServiceResponse {
@JsonProperty(value = "imageMetadata")
@JsonAlias("data")
@Builder.Default
private List<ImageMetadata> data = new ArrayList<>();
@Builder.Default
private List<ImageMetadata> dataCV = new ArrayList<>();
@@ -15,6 +15,7 @@ import lombok.NoArgsConstructor;
public class TableData {
private PageInfo pageInfo;
@Builder.Default
private List<TableCells> tableCells = new ArrayList<>();
}
@@ -19,7 +19,7 @@ public class TableServiceResponse {
private String operation;
private String targetFileExtension;
private String responseFileExtension;
@Builder.Default
private List<TableData> data = new ArrayList<>();
}
@@ -23,11 +23,11 @@ public class DividingColumnDetectionService {
public List<Rectangle2D> detectColumns(PageContents pageContents) {
if (pageContents.getSortedTextPositionSequences().size() < 2) {
if (pageContents.getSortedWords().size() < 2) {
return List.of(pageContents.getCropBox());
}
GapInformation linesWithGapInformation = GapDetectionService.findGapsInLines(pageContents.getSortedTextPositionSequences(), pageContents.getCropBox());
GapInformation linesWithGapInformation = GapDetectionService.findGapsInLines(pageContents.getSortedWords(), pageContents.getCropBox());
return detectColumnsFromLines(linesWithGapInformation.getXGaps(), pageContents.getCropBox());
}
@@ -5,8 +5,7 @@ import java.util.LinkedList;
import java.util.List;
import com.knecon.fforesight.service.layoutparser.processor.model.GapInformation;
import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
import com.knecon.fforesight.service.layoutparser.processor.model.text.Word;
import lombok.AllArgsConstructor;
import lombok.experimental.UtilityClass;
@@ -19,26 +18,26 @@ public class GapDetectionService {
private static final double NEW_LINE_FACTOR = 0.2;
public static GapInformation findGapsInLines(List<TextPositionSequence> sortedTextPositionSequences, Rectangle2D mainBodyTextFrame) {
public static GapInformation findGapsInLines(List<Word> sortedWords, Rectangle2D mainBodyTextFrame) {
if (sortedTextPositionSequences.isEmpty()) {
if (sortedWords.isEmpty()) {
return new GapInformation();
}
final double avgTextPositionHeight = getAvgTextPositionHeight(sortedTextPositionSequences);
final double avgTextPositionHeight = getAvgTextPositionHeight(sortedWords);
XGapsContext xGapContext = XGapsContext.init(mainBodyTextFrame);
YGapsContext yGapContext = YGapsContext.init(mainBodyTextFrame);
var previousTextPosition = sortedTextPositionSequences.get(0);
var previousTextPosition = sortedWords.get(0);
Rectangle2D rectangle = toRectangle2D(previousTextPosition);
xGapContext.addGapFromLeftEdgeOfMainBody(rectangle);
for (TextPositionSequence currentTextPosition : sortedTextPositionSequences.subList(1, sortedTextPositionSequences.size())) {
for (Word currentTextPosition : sortedWords.subList(1, sortedWords.size())) {
double yDifference = Math.abs(currentTextPosition.getMaxYDirAdj() - previousTextPosition.getMaxYDirAdj());
double xGap = Math.abs(previousTextPosition.getMaxXDirAdj() - currentTextPosition.getMinXDirAdj());
double xGap = Math.abs(previousTextPosition.getMaxXDirAdj() - currentTextPosition.getXDirAdj());
Rectangle2D previousTextPositionBBox = toRectangle2D(previousTextPosition);
Rectangle2D currentTextPositionBBox = toRectangle2D(currentTextPosition);
@@ -60,14 +59,14 @@ public class GapDetectionService {
}
previousTextPosition = currentTextPosition;
}
xGapContext.addGapToRightEdgeOfMainBody(toRectangle2D(sortedTextPositionSequences.get(sortedTextPositionSequences.size() - 1)));
xGapContext.addGapToRightEdgeOfMainBody(toRectangle2D(sortedWords.get(sortedWords.size() - 1)));
xGapContext.gapsPerLine.add(xGapContext.gapsInCurrentLine);
return new GapInformation(xGapContext.gapsPerLine, yGapContext.gapsPerLine);
}
private static Rectangle2D toRectangle2D(TextPositionSequence textPosition) {
private static Rectangle2D toRectangle2D(Word textPosition) {
return mirrorY(textPosition.getBBox());
}
@@ -88,18 +87,18 @@ public class GapDetectionService {
}
private static void assertAllTextPositionsHaveSameDir(List<TextPositionSequence> textPositionSequences) {
private static void assertAllTextPositionsHaveSameDir(List<Word> words) {
assert textPositionSequences.stream()
.map(TextPositionSequence::getDir)
.allMatch(a -> a.equals(textPositionSequences.get(0).getDir()));
assert words.stream()
.map(Word::getDir)
.allMatch(a -> a.equals(words.get(0).getDir()));
}
private static double getAvgTextPositionHeight(List<TextPositionSequence> textPositionSequences) {
private static double getAvgTextPositionHeight(List<Word> words) {
return textPositionSequences.stream()
.mapToDouble(TextPositionSequence::getHeight).average().orElseThrow();
return words.stream()
.mapToDouble(Word::getHeight).average().orElseThrow();
}

Some files were not shown because too many files have changed in this diff Show More