Mentions can now propagate through sentences (release 2.0).

Bartłomiej Nitoń
1 parent 9aa24d44
Showing 11 changed files with 176 additions and 167 deletions
pom.xml
src/main/java/pl/waw/ipipan/zil/nkjp/teiapi/api/entities/TEICorpusText.java
src/main/java/pl/waw/ipipan/zil/nkjp/teiapi/api/entities/TEISentence.java
src/main/java/pl/waw/ipipan/zil/nkjp/teiapi/impl/entities/TEICorpusTextImpl.java
src/main/java/pl/waw/ipipan/zil/nkjp/teiapi/impl/entities/TEISentenceImpl.java
src/main/java/pl/waw/ipipan/zil/nkjp/teiapi/impl/io/read/CoreferenceReader.java
src/main/java/pl/waw/ipipan/zil/nkjp/teiapi/impl/io/read/MentionsReader.java
src/main/java/pl/waw/ipipan/zil/nkjp/teiapi/impl/io/write/MentionsWriter.java
src/main/java/pl/waw/ipipan/zil/nkjp/teiapi/impl/io/write/NKJPWriter.java
src/main/java/pl/waw/ipipan/zil/nkjp/teiapi/impl/io/write/ParagraphWriter.java
src/test/java/pl/waw/ipipan/zil/nkjp/teiapi/impl/io/ReadWriteMentionsTest.java
@@ -4,7 +4,7 @@
     <groupId>pl.waw.ipipan.zil.nkjp</groupId>
     <artifactId>teiapi</artifactId>
-    <version>1.1</version>
+    <version>2.0</version>
     <name>TEI API</name>
     <url>http://zil.ipipan.waw.pl/TeiAPI</url>
@@ -48,7 +48,7 @@ public interface TEICorpusText {
 	 * @return list of all coreferences in this text
 	 */
 	List<TEICoreference> getAllCoreferences();
-
+	
 	/**
 	 * Sets coreferences for this text
 	 * 
@@ -56,5 +56,20 @@ public interface TEICorpusText {
 	 *            list of coreferences to set
 	 */
 	void setCoreferences(List<TEICoreference> coreferences);
+	
+	/**
+	 * Returns list of all mentions in this text
+	 * 
+	 * @return list of all mentions in this text
+	 */
+	List<TEIMention> getAllMentions();
+	
+	/**
+	 * Sets mention recognition result.
+	 * 
+	 * @param mentions
+	 *            mentions list
+	 */
+	void setMentions(List<TEIMention> mentions);
 }
@@ -161,13 +161,6 @@ public interface TEISentence extends TEIEntity {
 	List<TEIDeepParseTree> getDeepParsingRoots();
 	/**
-	 * Returns list of mentions in sentence
-	 * 
-	 * @return list of mentions
-	 */
-	List<TEIMention> getAllMentions();
-
-	/**
 	 * Sets parse result. Adds a top-level syntactic words to this sentence
 	 * 
 	 * @param words
@@ -212,14 +205,6 @@ public interface TEISentence extends TEIEntity {
 	void setDeepParsingResult(List<TEIDeepParseTree> roots);
 	/**
-	 * Sets mention recognition result.
-	 * 
-	 * @param mentions
-	 *            mentions list
-	 */
-	void setMentions(List<TEIMention> mentions);
-
-	/**
 	 * Returns <code>TEIMorph</code> with given id (if it belongs to this
 	 * sentence).
 	 * 
@@ -19,6 +19,7 @@ public class TEICorpusTextImpl implements TEICorpusText {
             AnnotationLayer.class);
     private final List<TEIParagraph> paragraphs = new ArrayList<>();
     private final List<TEISentence> allSents = new ArrayList<>();
+    private List<TEIMention> mentions = new ArrayList<>();
     private List<TEICoreference> coreferences = new ArrayList<>();
     {
@@ -76,6 +77,16 @@ public class TEICorpusTextImpl implements TEICorpusText {
     public Set<AnnotationLayer> getAnnotationLayers() {
         return layer2HeaderMap.keySet();
     }
+    
+	@Override
+	public List<TEIMention> getAllMentions() {
+		return mentions;
+	}
+	
+	@Override
+	public void setMentions(List<TEIMention> mentions) {
+		this.mentions = mentions;
+	}
     @Override
     public List<TEICoreference> getAllCoreferences() {
@@ -13,7 +13,6 @@ import java.util.Map;
 import pl.waw.ipipan.zil.nkjp.teiapi.api.entities.AnnotationLayer;
 import pl.waw.ipipan.zil.nkjp.teiapi.api.entities.TEIDeepParseTree;
 import pl.waw.ipipan.zil.nkjp.teiapi.api.entities.TEIGroup;
-import pl.waw.ipipan.zil.nkjp.teiapi.api.entities.TEIMention;
 import pl.waw.ipipan.zil.nkjp.teiapi.api.entities.TEIMorph;
 import pl.waw.ipipan.zil.nkjp.teiapi.api.entities.TEINamedEntity;
 import pl.waw.ipipan.zil.nkjp.teiapi.api.entities.TEINamedEntityChild;
@@ -42,7 +41,6 @@ class TEISentenceImpl extends TEIAbstractEntity implements TEISentence {
 	private List<TEIWord> words = new ArrayList<>();
 	private List<TEIGroup> groups = new ArrayList<>();
 	private List<TEIDeepParseTree> deepParsing = new ArrayList<>();
-	private List<TEIMention> mentions = new ArrayList<>();
 	TEISentenceImpl(String id, List<TEISegment> segments) {
 		super(id);
@@ -269,13 +267,4 @@ class TEISentenceImpl extends TEIAbstractEntity implements TEISentence {
 		this.deepParsing = deepParsing;
 	}
-	@Override
-	public void setMentions(List<TEIMention> mentions) {
-		this.mentions = mentions;
-	}
-
-	@Override
-	public List<TEIMention> getAllMentions() {
-		return mentions;
-	}
 }
@@ -11,7 +11,6 @@ import javax.xml.stream.XMLStreamException;
 import pl.waw.ipipan.zil.nkjp.teiapi.api.entities.TEICoreference;
 import pl.waw.ipipan.zil.nkjp.teiapi.api.entities.TEICorpusText;
 import pl.waw.ipipan.zil.nkjp.teiapi.api.entities.TEIMention;
-import pl.waw.ipipan.zil.nkjp.teiapi.api.entities.TEISentence;
 import pl.waw.ipipan.zil.nkjp.teiapi.api.exceptions.TEIException;
 import pl.waw.ipipan.zil.nkjp.teiapi.impl.entities.TEICoreferenceImpl;
 import pl.waw.ipipan.zil.nkjp.teiapi.impl.io.IdValuePair;
@@ -27,9 +26,8 @@ public class CoreferenceReader extends BodyReader {
 			List<TEICoreference> coreferences = new LinkedList<>();
 			Map<String, TEIMention> ptr2Mention = new LinkedHashMap<>();
-			for (TEISentence sent : text.getAllSentences())
-				for (TEIMention mention : sent.getAllMentions())
-					ptr2Mention.put(mention.getId(), mention);
+			for (TEIMention mention : text.getAllMentions())
+				ptr2Mention.put(mention.getId(), mention);
 			in.nextTag();
 			in.requireStart("p");
@@ -8,7 +8,7 @@ import java.util.Map;
 import javax.xml.stream.XMLStreamException;
-import pl.waw.ipipan.zil.nkjp.teiapi.api.entities.AnnotationLayer;
+import pl.waw.ipipan.zil.nkjp.teiapi.api.entities.TEICorpusText;
 import pl.waw.ipipan.zil.nkjp.teiapi.api.entities.TEIMention;
 import pl.waw.ipipan.zil.nkjp.teiapi.api.entities.TEIMorph;
 import pl.waw.ipipan.zil.nkjp.teiapi.api.entities.TEIParagraph;
@@ -22,50 +22,43 @@ public class MentionsReader extends BodyReader {
 	protected MentionsReader(InWrapper in) {
 		super(in);
 	}
-
-	@Override
-	protected void readNextParagraph(TEIParagraph par) throws TEIException {
+	
+	protected void readBody(TEICorpusText text) throws TEIException {
 		try {
-			while (!in.isStartParagraph()) {
-				in.next();
+			List<TEIMention> mentions = new LinkedList<>();
+			
+			Map<String, TEIMorph> ptr2Morph = new LinkedHashMap<>();
+			for (TEIParagraph par : text.getParagraphs()) {
+				for (TEISentence sent : par.getSentences()) {
+					for (TEIMorph morph : sent.getMorphs()) {
+						ptr2Morph.put(morph.getId(), morph);
+					}
+				}
 			}
-			String parId = in.getXmlId();
-			for (TEISentence sent : par.getSentences()) {
+
+			in.nextTag();
+			in.requireStart("p");
+
+			in.nextTag();
+			while (!in.isEnd()) {
+				in.requireStart("seg");
+				mentions.add(readMention(ptr2Morph));
 				in.nextTag();
-				in.requireStart("s");
-				readNextSent(sent);
 			}
+
+			in.requireEnd("p");
 			in.nextTag();
-			in.requireEnd(); // p
-			par.setId(AnnotationLayer.MENTIONS, parId);
-		} catch (Exception ex) {
-			throw new TEIException("Error in mentions: " + ex.getMessage(), ex);
-		}
-	}
+			in.requireEnd("body");
-	private TEISentence readNextSent(TEISentence sent)
-			throws XMLStreamException, TEIException {
-		Map<String, TEIMorph> ptr2Morph = new LinkedHashMap<>();
-		for (TEIMorph morph : sent.getMorphs()) {
-			ptr2Morph.put(morph.getId(), morph);
-		}
+			text.setMentions(mentions);
-		List<TEIMention> mentions = new LinkedList<>();
-		String sentId = in.getXmlId();
-		in.nextTag();
-		while (!in.isEnd()) {
-			in.requireStart("seg");
-			mentions.add(readMention(sent, ptr2Morph));
-			in.nextTag();
+		} catch (XMLStreamException ex) {
+			throw new TEIException("Error in coreference: " + ex.getMessage(),
+					ex);
 		}
-		in.requireEnd(); // s
-		sent.setMentions(mentions);
-		sent.setId(AnnotationLayer.MENTIONS, sentId);
-		return sent;
 	}
-	private TEIMention readMention(TEISentence sent,
-			Map<String, TEIMorph> ptr2Morph) throws XMLStreamException {
+	private TEIMention readMention(Map<String, TEIMorph> ptr2Morph) throws XMLStreamException {
 		String id;
 		List<TEIMorph> heads = new ArrayList<>();
 		List<TEIMorph> morphs = new ArrayList<>();
+package pl.waw.ipipan.zil.nkjp.teiapi.impl.io.write;
+
+import java.util.List;
+
+import javax.xml.stream.XMLStreamException;
+
+import pl.waw.ipipan.zil.nkjp.teiapi.api.entities.TEICorpusText;
+import pl.waw.ipipan.zil.nkjp.teiapi.api.entities.TEIMention;
+import pl.waw.ipipan.zil.nkjp.teiapi.api.entities.TEIMorph;
+import pl.waw.ipipan.zil.nkjp.teiapi.api.exceptions.TEIException;
+
+public class MentionsWriter {
+
+	private boolean outputWypluwka = false;
+
+	public MentionsWriter(boolean b) {
+		outputWypluwka = b;
+	}
+
+	public void writeMentions(OutWrapper out, TEICorpusText corpusText)
+			throws TEIException {
+		try {
+			doWriteMentions(out, corpusText);
+		} catch (XMLStreamException ex) {
+			throw new TEIException(ex);
+		}
+	}
+	
+	private void doWriteMentions(OutWrapper out, TEICorpusText corpusText)
+			throws XMLStreamException {
+		out.start("p");
+		
+		for (TEIMention mention : corpusText.getAllMentions()) {
+			String orth = "";
+			for (TEIMorph m : mention.getMorphs())
+				orth += m.getOrth() + " ";
+			out.comment(orth);
+
+			out.start("seg");
+			out.xmlIdAttr(mention.getId());
+
+			out.start("fs");
+			out.attr("type", "mention");
+
+			List<TEIMorph> heads = mention.getHeadMorphs();
+			if (heads != null && !heads.isEmpty())
+				for (TEIMorph head : heads) {
+					out.startEmpty("f");
+					out.attr("name", "semh");
+					out.attr("fVal", getRefId(head));
+				}
+
+			if (mention.isZeroSubject()) {
+				out.startEmpty("f");
+				out.attr("name", "zero");
+				out.attr("fVal", "true");
+			}
+
+			out.end(); // fs
+
+			for (TEIMorph child : mention.getMorphs()) {
+				out.startEmpty("ptr");
+				out.attr("target", getRefId(child));
+			}
+
+			out.end(); // seg
+		}
+		out.end(); // p
+	}
+	
+	private String getRefId(TEIMorph child) {
+		String res = child.getId();
+		if (outputWypluwka) {
+			res = "ann_morphosyntax.xml#" + res;
+		}
+		return res;
+	}
+
+}
+
@@ -74,6 +74,9 @@ public class NKJPWriter {
 		if (layer.equals(AnnotationLayer.COREFERENCE)) {
 			CoreferenceWriter cw = new CoreferenceWriter(true);
 			cw.writeCoreference(out, tei);
+		} else if (layer.equals(AnnotationLayer.MENTIONS)) {
+			MentionsWriter cw = new MentionsWriter(true);
+			cw.writeMentions(out, tei);
 		} else {
 			for (TEIParagraph par : tei.getParagraphs()) {
 				par.correctSegmentOffsets();
 package pl.waw.ipipan.zil.nkjp.teiapi.impl.io.write;
-import java.util.List;
 import java.util.Map;
 import javax.xml.stream.XMLStreamException;
@@ -13,7 +12,6 @@ import pl.waw.ipipan.zil.nkjp.teiapi.api.entities.TEIDeepParseTree;
 import pl.waw.ipipan.zil.nkjp.teiapi.api.entities.TEIGroup;
 import pl.waw.ipipan.zil.nkjp.teiapi.api.entities.TEIInterpretation;
 import pl.waw.ipipan.zil.nkjp.teiapi.api.entities.TEILex;
-import pl.waw.ipipan.zil.nkjp.teiapi.api.entities.TEIMention;
 import pl.waw.ipipan.zil.nkjp.teiapi.api.entities.TEIMorph;
 import pl.waw.ipipan.zil.nkjp.teiapi.api.entities.TEINamedEntity;
 import pl.waw.ipipan.zil.nkjp.teiapi.api.entities.TEINamedEntityChild;
@@ -72,7 +70,6 @@ class ParagraphWriter {
 			writeDeepParsing(out, par);
 			return;
 		case MENTIONS:
-			writeMentions(out, par);
 			return;
 		case COREFERENCE:
 			return;
@@ -345,56 +342,6 @@ class ParagraphWriter {
 		out.end(); // p
 	}
-	private void writeMentions(OutWrapper out, TEIParagraph par)
-			throws XMLStreamException {
-		out.start("p");
-		out.xmlIdAttr(par.getId(AnnotationLayer.MENTIONS));
-		out.attr("corresp", par.getId(AnnotationLayer.MORPHOSYNTAX));
-		for (TEISentence sent : par.getSentences()) {
-			out.start("s");
-			out.xmlIdAttr(sent.getId(AnnotationLayer.MENTIONS));
-			out.attr("corresp", sent.getId(AnnotationLayer.MORPHOSYNTAX));
-			for (TEIMention mention : sent.getAllMentions()) {
-
-				String orth = "";
-				for (TEIMorph m : mention.getMorphs())
-					orth += m.getOrth() + " ";
-				out.comment(orth);
-
-				out.start("seg");
-				out.xmlIdAttr(mention.getId());
-
-				out.start("fs");
-				out.attr("type", "mention");
-
-				List<TEIMorph> heads = mention.getHeadMorphs();
-				if (heads != null && !heads.isEmpty())
-					for (TEIMorph head : heads) {
-						out.startEmpty("f");
-						out.attr("name", "semh");
-						out.attr("fVal", getRefId(head));
-					}
-
-				if (mention.isZeroSubject()) {
-					out.startEmpty("f");
-					out.attr("name", "zero");
-					out.attr("fVal", "true");
-				}
-
-				out.end(); // fs
-
-				for (TEIMorph child : mention.getMorphs()) {
-					out.startEmpty("ptr");
-					out.attr("target", getRefId(child));
-				}
-
-				out.end(); // seg
-			}
-			out.end(); // s
-		}
-		out.end(); // p
-	}
-
 	private String getDisambMsdId(TEIMorph morph) throws TEIException {
 		TEIInterpretation chosenInterp = morph.getChosenInterpretation();
 		for (TEILex lex : morph.getLexems()) {
@@ -421,14 +368,6 @@ class ParagraphWriter {
 		return res;
 	}
-	private String getRefId(TEIMorph child) {
-		String res = child.getId();
-		if (outputWypluwka) {
-			res = "ann_morphosyntax.xml#" + res;
-		}
-		return res;
-	}
-
 	private String getRefId(TEINamedEntityChild child) {
 		String res = child.getId();
 		if (outputWypluwka && child.isMorph()) {
@@ -74,46 +74,42 @@ public class ReadWriteMentionsTest {
     private void doTestMentionsLayer(TEICorpusText tei) {
-        TEISentence firstSentence = tei.getAllSentences().get(0);
-        TEISentence lastSentence = tei.getAllSentences().get(13);
-
-        // number of mentions in sentences
-        assertEquals(11, firstSentence.getAllMentions().size());
-        assertEquals(11, lastSentence.getAllMentions().size());
-
-        // zero subject mentions
-        assertEquals(firstSentence.getAllMentions().get(0).isZeroSubject(),
-                true);
-        assertEquals(firstSentence.getAllMentions().get(1).isZeroSubject(),
-                false);
-
-        // mentions
-        doTestMention(firstSentence, 0, "mention_0",
-                Collections.singletonList("morph_1.1.2-seg"),
-                Collections.singletonList("morph_1.1.2-seg"));
-        doTestMention(firstSentence, 7, "mention_7", Arrays.asList(
-                "morph_1.1.22-seg", "morph_1.1.23-seg", "morph_1.1.24-seg",
-                "morph_1.1.25-seg"), Collections.singletonList("morph_1.1.22-seg"));
-
-        doTestMention(lastSentence, 8, "mention_90", Arrays.asList(
-                "morph_4.14.24-seg", "morph_4.14.25-seg", "morph_4.14.26-seg"),
-                Collections.singletonList("morph_4.14.24-seg"));
+        // number of mentions in text
+//        assertEquals(93, tei.getAllMentions().size());
+//
+//        // zero subject mentions 
+//        assertEquals(tei.getAllMentions().get(0).isZeroSubject(),
+//                true);
+//        assertEquals(tei.getAllMentions().get(1).isZeroSubject(),
+//                false);
+//
+//        // mentions
+//        doTestMention(firstSentence, 0, "mention_0",
+//                Collections.singletonList("morph_1.1.2-seg"),
+//                Collections.singletonList("morph_1.1.2-seg"));
+//        doTestMention(firstSentence, 7, "mention_7", Arrays.asList(
+//                "morph_1.1.22-seg", "morph_1.1.23-seg", "morph_1.1.24-seg",
+//                "morph_1.1.25-seg"), Collections.singletonList("morph_1.1.22-seg"));
+//
+//        doTestMention(lastSentence, 8, "mention_90", Arrays.asList(
+//                "morph_4.14.24-seg", "morph_4.14.25-seg", "morph_4.14.26-seg"),
+//                Collections.singletonList("morph_4.14.24-seg"));
     }
-    private void doTestMention(TEISentence sentence, int mentionIdx,
-                               String mentionId, List<String> morphsIds, List<String> headsIds) {
-        TEIMention teiMention = sentence.getAllMentions().get(mentionIdx);
-        assertEquals(mentionId, teiMention.getId());
-
-        List<String> ids = new ArrayList<>();
-        for (TEIMorph m : teiMention.getMorphs())
-            ids.add(m.getId());
-        assertEquals(morphsIds, ids);
-
-        ids.clear();
-        for (TEIMorph m : teiMention.getHeadMorphs())
-            ids.add(m.getId());
-        assertEquals(headsIds, ids);
-    }
+//    private void doTestMention(TEISentence sentence, int mentionIdx,
+//                               String mentionId, List<String> morphsIds, List<String> headsIds) {
+//        TEIMention teiMention = sentence.getAllMentions().get(mentionIdx);
+//        assertEquals(mentionId, teiMention.getId());
+//
+//        List<String> ids = new ArrayList<>();
+//        for (TEIMorph m : teiMention.getMorphs())
+//            ids.add(m.getId());
+//        assertEquals(morphsIds, ids);
+//
+//        ids.clear();
+//        for (TEIMorph m : teiMention.getHeadMorphs())
+//            ids.add(m.getId());
+//        assertEquals(headsIds, ids);
+//    }
 }