Examples with Mention edu.stanford.nlp.coref.data.Mention used on opensource projects

Example 16 with Mention

use of edu.stanford.nlp.coref.data.Mention in project CoreNLP by stanfordnlp.

the class HybridCorefPrinter method sentenceStringWithMention.

public static String sentenceStringWithMention(int i, Document document, boolean gold, boolean printClusterID) {
    StringBuilder sentStr = new StringBuilder();
    List<CoreMap> sentences = document.annotation.get(CoreAnnotations.SentencesAnnotation.class);
    List<List<Mention>> allMentions;
    if (gold) {
        allMentions = document.goldMentions;
    } else {
        allMentions = document.predictedMentions;
    }
    //    String filename = document.annotation.get()
    int previousOffset = 0;
    CoreMap sentence = sentences.get(i);
    List<Mention> mentions = allMentions.get(i);
    List<CoreLabel> t = sentence.get(CoreAnnotations.TokensAnnotation.class);
    String speaker = t.get(0).get(SpeakerAnnotation.class);
    if (NumberMatchingRegex.isDecimalInteger(speaker))
        speaker = speaker + ": " + document.predictedMentionsByID.get(Integer.parseInt(speaker)).spanToString();
    sentStr.append("\tspeaker: " + speaker + " (" + t.get(0).get(UtteranceAnnotation.class) + ") ");
    String[] tokens = new String[t.size()];
    for (CoreLabel c : t) {
        tokens[c.index() - 1] = c.word();
    }
    //    if(previousOffset+2 < t.get(0).get(CoreAnnotations.CharacterOffsetBeginAnnotation.class) && printClusterID) {
    //      sentStr.append("\n");
    //    }
    previousOffset = t.get(t.size() - 1).get(CoreAnnotations.CharacterOffsetEndAnnotation.class);
    Counter<Integer> startCounts = new ClassicCounter<>();
    Counter<Integer> endCounts = new ClassicCounter<>();
    Map<Integer, Deque<Mention>> endMentions = Generics.newHashMap();
    for (Mention m : mentions) {
        //      if(!gold && (document.corefClusters.get(m.corefClusterID)==null || document.corefClusters.get(m.corefClusterID).getCorefMentions().size()<=1)) {
        //        continue;
        //      }
        startCounts.incrementCount(m.startIndex);
        endCounts.incrementCount(m.endIndex);
        if (!endMentions.containsKey(m.endIndex))
            endMentions.put(m.endIndex, new ArrayDeque<>());
        endMentions.get(m.endIndex).push(m);
    }
    for (int j = 0; j < tokens.length; j++) {
        if (endMentions.containsKey(j)) {
            for (Mention m : endMentions.get(j)) {
                int id = (gold) ? m.goldCorefClusterID : m.corefClusterID;
                id = (printClusterID) ? id : m.mentionID;
                sentStr.append("]_").append(id);
            }
        }
        for (int k = 0; k < startCounts.getCount(j); k++) {
            if (sentStr.length() > 0 && sentStr.charAt(sentStr.length() - 1) != '[')
                sentStr.append(" ");
            sentStr.append("[");
        }
        if (sentStr.length() > 0 && sentStr.charAt(sentStr.length() - 1) != '[')
            sentStr.append(" ");
        sentStr.append(tokens[j]);
    }
    if (endMentions.containsKey(tokens.length)) {
        for (Mention m : endMentions.get(tokens.length)) {
            int id = (gold) ? m.goldCorefClusterID : m.corefClusterID;
            id = (printClusterID) ? id : m.mentionID;
            //append("_").append(m.mentionID);
            sentStr.append("]_").append(id);
        }
    }
    return sentStr.toString();
}

Example 17 with Mention

use of edu.stanford.nlp.coref.data.Mention in project CoreNLP by stanfordnlp.

the class HybridCorefMentionFinder method extractNPorPRP.

private static void extractNPorPRP(CoreMap s, List<Mention> mentions, Set<IntPair> mentionSpanSet, Set<IntPair> namedEntitySpanSet) {
    List<CoreLabel> sent = s.get(CoreAnnotations.TokensAnnotation.class);
    Tree tree = s.get(TreeCoreAnnotations.TreeAnnotation.class);
    tree.indexLeaves();
    SemanticGraph basicDependency = s.get(SemanticGraphCoreAnnotations.BasicDependenciesAnnotation.class);
    SemanticGraph enhancedDependency = s.get(SemanticGraphCoreAnnotations.EnhancedDependenciesAnnotation.class);
    if (enhancedDependency == null) {
        enhancedDependency = s.get(SemanticGraphCoreAnnotations.BasicDependenciesAnnotation.class);
    }
    TregexPattern tgrepPattern = npOrPrpMentionPattern;
    TregexMatcher matcher = tgrepPattern.matcher(tree);
    while (matcher.find()) {
        Tree t = matcher.getMatch();
        List<Tree> mLeaves = t.getLeaves();
        int beginIdx = ((CoreLabel) mLeaves.get(0).label()).get(CoreAnnotations.IndexAnnotation.class) - 1;
        int endIdx = ((CoreLabel) mLeaves.get(mLeaves.size() - 1).label()).get(CoreAnnotations.IndexAnnotation.class);
        // try not to have span that ends with ,
        if (",".equals(sent.get(endIdx - 1).word())) {
            endIdx--;
        }
        IntPair mSpan = new IntPair(beginIdx, endIdx);
        //      if(!mentionSpanSet.contains(mSpan) && (!insideNE(mSpan, namedEntitySpanSet)) ) {
        if (!mentionSpanSet.contains(mSpan) && (!insideNE(mSpan, namedEntitySpanSet) || t.value().startsWith("PRP"))) {
            int dummyMentionId = -1;
            Mention m = new Mention(dummyMentionId, beginIdx, endIdx, sent, basicDependency, enhancedDependency, new ArrayList<>(sent.subList(beginIdx, endIdx)), t);
            mentions.add(m);
            mentionSpanSet.add(mSpan);
            if (m.originalSpan.size() > 1) {
                boolean isNE = true;
                for (CoreLabel cl : m.originalSpan) {
                    if (!cl.tag().startsWith("NNP"))
                        isNE = false;
                }
                if (isNE) {
                    namedEntitySpanSet.add(mSpan);
                }
            }
        }
    }
}

Example 18 with Mention

use of edu.stanford.nlp.coref.data.Mention in project CoreNLP by stanfordnlp.

the class HybridCorefMentionFinder method extractNamedEntityMentions.

protected static void extractNamedEntityMentions(CoreMap s, List<Mention> mentions, Set<IntPair> mentionSpanSet, Set<IntPair> namedEntitySpanSet) {
    List<CoreLabel> sent = s.get(CoreAnnotations.TokensAnnotation.class);
    SemanticGraph basicDependency = s.get(SemanticGraphCoreAnnotations.BasicDependenciesAnnotation.class);
    SemanticGraph enhancedDependency = s.get(SemanticGraphCoreAnnotations.EnhancedDependenciesAnnotation.class);
    if (enhancedDependency == null) {
        enhancedDependency = s.get(SemanticGraphCoreAnnotations.BasicDependenciesAnnotation.class);
    }
    String preNE = "O";
    int beginIndex = -1;
    for (CoreLabel w : sent) {
        String nerString = w.ner();
        if (!nerString.equals(preNE)) {
            int endIndex = w.get(CoreAnnotations.IndexAnnotation.class) - 1;
            if (!preNE.matches("O")) {
                if (w.get(CoreAnnotations.TextAnnotation.class).equals("'s") && w.tag().equals("POS")) {
                    endIndex++;
                }
                IntPair mSpan = new IntPair(beginIndex, endIndex);
                // attached to the previous NER by the earlier heuristic
                if (beginIndex < endIndex && !mentionSpanSet.contains(mSpan)) {
                    int dummyMentionId = -1;
                    Mention m = new Mention(dummyMentionId, beginIndex, endIndex, sent, basicDependency, enhancedDependency, new ArrayList<>(sent.subList(beginIndex, endIndex)));
                    mentions.add(m);
                    mentionSpanSet.add(mSpan);
                    namedEntitySpanSet.add(mSpan);
                }
            }
            beginIndex = endIndex;
            preNE = nerString;
        }
    }
    // NE at the end of sentence
    if (!preNE.matches("O")) {
        IntPair mSpan = new IntPair(beginIndex, sent.size());
        if (!mentionSpanSet.contains(mSpan)) {
            int dummyMentionId = -1;
            Mention m = new Mention(dummyMentionId, beginIndex, sent.size(), sent, basicDependency, enhancedDependency, new ArrayList<>(sent.subList(beginIndex, sent.size())));
            mentions.add(m);
            mentionSpanSet.add(mSpan);
            namedEntitySpanSet.add(mSpan);
        }
    }
}

Example 19 with Mention

use of edu.stanford.nlp.coref.data.Mention in project CoreNLP by stanfordnlp.

the class HybridCorefMentionFinder method findMentions.

@Override
public List<List<Mention>> findMentions(Annotation doc, Dictionaries dict, Properties props) {
    List<List<Mention>> predictedMentions = new ArrayList<>();
    Set<String> neStrings = Generics.newHashSet();
    List<Set<IntPair>> mentionSpanSetList = Generics.newArrayList();
    List<CoreMap> sentences = doc.get(CoreAnnotations.SentencesAnnotation.class);
    // extract premarked mentions, NP/PRP, named entity, enumerations
    for (CoreMap s : sentences) {
        List<Mention> mentions = new ArrayList<>();
        predictedMentions.add(mentions);
        Set<IntPair> mentionSpanSet = Generics.newHashSet();
        Set<IntPair> namedEntitySpanSet = Generics.newHashSet();
        extractPremarkedEntityMentions(s, mentions, mentionSpanSet, namedEntitySpanSet);
        extractNamedEntityMentions(s, mentions, mentionSpanSet, namedEntitySpanSet);
        extractNPorPRP(s, mentions, mentionSpanSet, namedEntitySpanSet);
        extractEnumerations(s, mentions, mentionSpanSet, namedEntitySpanSet);
        addNamedEntityStrings(s, neStrings, namedEntitySpanSet);
        mentionSpanSetList.add(mentionSpanSet);
    }
    extractNamedEntityModifiers(sentences, mentionSpanSetList, predictedMentions, neStrings);
    // find head
    for (int i = 0; i < sentences.size(); i++) {
        findHead(sentences.get(i), predictedMentions.get(i));
    }
    // mention selection based on document-wise info
    removeSpuriousMentions(doc, predictedMentions, dict, CorefProperties.removeNestedMentions(props), lang);
    // if this is for MD training, skip classification
    if (!CorefProperties.isMentionDetectionTraining(props)) {
        mdClassifier.classifyMentions(predictedMentions, dict, props);
    }
    return predictedMentions;
}

Example 20 with Mention

use of edu.stanford.nlp.coref.data.Mention in project CoreNLP by stanfordnlp.

the class HybridCorefSystem method postProcessing.

/** Remove singletons, appositive, predicate nominatives, relative pronouns. */
private static void postProcessing(Document document) {
    Set<Mention> removeSet = Generics.newHashSet();
    Set<Integer> removeClusterSet = Generics.newHashSet();
    for (CorefCluster c : document.corefClusters.values()) {
        Set<Mention> removeMentions = Generics.newHashSet();
        for (Mention m : c.getCorefMentions()) {
            if (HybridCorefProperties.REMOVE_APPOSITION_PREDICATENOMINATIVES && ((m.appositions != null && m.appositions.size() > 0) || (m.predicateNominatives != null && m.predicateNominatives.size() > 0) || (m.relativePronouns != null && m.relativePronouns.size() > 0))) {
                removeMentions.add(m);
                removeSet.add(m);
                m.corefClusterID = m.mentionID;
            }
        }
        c.corefMentions.removeAll(removeMentions);
        if (HybridCorefProperties.REMOVE_SINGLETONS && c.getCorefMentions().size() == 1) {
            removeClusterSet.add(c.clusterID);
        }
    }
    for (int removeId : removeClusterSet) {
        document.corefClusters.remove(removeId);
    }
    for (Mention m : removeSet) {
        document.positions.remove(m);
    }
}