Example usage for edu.stanford.nlp.trees TypedDependency setExtra

Introduction

In this page you can find the example usage for edu.stanford.nlp.trees TypedDependency setExtra.

Prototype

public void setExtra() 

Source Link

Usage

From source file:de.tudarmstadt.ukp.dkpro.core.corenlp.internal.DKPro2CoreNlp.java

License:Open Source License

public Annotation convert(JCas aSource, Annotation aTarget) {
    // Document annotation
    aTarget.set(CoreAnnotations.TextAnnotation.class, aSource.getDocumentText());

    // Sentences/*from  ww  w .j  a  v  a 2  s  . co m*/
    List<CoreMap> sentences = new ArrayList<>();
    for (Sentence s : select(aSource, Sentence.class)) {
        if (StringUtils.isBlank(s.getCoveredText())) {
            continue;
        }

        String sentenceText = s.getCoveredText();
        if (encoding != null && !"UTF-8".equals(encoding.name())) {
            sentenceText = new String(sentenceText.getBytes(StandardCharsets.UTF_8), encoding);
        }

        Annotation sentence = new Annotation(sentenceText);
        sentence.set(CharacterOffsetBeginAnnotation.class, s.getBegin());
        sentence.set(CharacterOffsetEndAnnotation.class, s.getEnd());
        sentence.set(SentenceIndexAnnotation.class, sentences.size());

        // Tokens
        Map<Token, IndexedWord> idxTokens = new HashMap<>();
        List<CoreLabel> tokens = new ArrayList<>();
        for (Token t : selectCovered(Token.class, s)) {
            String tokenText = t.getCoveredText();
            if (encoding != null && !"UTF-8".equals(encoding.name())) {
                tokenText = new String(tokenText.getBytes(StandardCharsets.UTF_8), encoding);
            }

            CoreLabel token = tokenFactory.makeToken(tokenText, t.getBegin(), t.getEnd() - t.getBegin());
            // First add token so that tokens.size() returns a 1-based counting as required
            // by IndexAnnotation
            tokens.add(token);
            token.set(SentenceIndexAnnotation.class, sentences.size());
            token.set(IndexAnnotation.class, tokens.size());
            token.set(TokenKey.class, t);
            idxTokens.put(t, new IndexedWord(token));

            // POS tags
            if (readPos && t.getPos() != null) {
                token.set(PartOfSpeechAnnotation.class, t.getPos().getPosValue());
            }

            // Lemma
            if (t.getLemma() != null) {
                token.set(LemmaAnnotation.class, t.getLemma().getValue());
            }

            // Stem
            if (t.getStem() != null) {
                token.set(StemAnnotation.class, t.getStem().getValue());
            }

            // NamedEntity
            // TODO: only token-based NEs are supported, but not multi-token NEs
            // Supporting multi-token NEs via selectCovering would be very slow. To support
            // them, another approach would need to be implemented, e.g. via indexCovering.
            List<NamedEntity> nes = selectCovered(NamedEntity.class, t);
            if (nes.size() > 0) {
                token.set(NamedEntityTagAnnotation.class, nes.get(0).getValue());
            } else {
                token.set(NamedEntityTagAnnotation.class, "O");
            }
        }

        // Constituents
        for (ROOT r : selectCovered(ROOT.class, s)) {
            Tree tree = createStanfordTree(r, idxTokens);
            tree.indexSpans();
            sentence.set(TreeAnnotation.class, tree);
        }

        // Dependencies
        List<TypedDependency> dependencies = new ArrayList<>();
        for (Dependency d : selectCovered(Dependency.class, s)) {
            TypedDependency dep = new TypedDependency(GrammaticalRelation.valueOf(d.getDependencyType()),
                    idxTokens.get(d.getGovernor()), idxTokens.get(d.getDependent()));
            if (DependencyFlavor.ENHANCED.equals(d.getFlavor())) {
                dep.setExtra();
            }
            dependencies.add(dep);
        }
        sentence.set(EnhancedDependenciesAnnotation.class, new SemanticGraph(dependencies));

        if (ptb3Escaping) {
            tokens = applyPtbEscaping(tokens, quoteBegin, quoteEnd);
        }

        sentence.set(TokensAnnotation.class, tokens);
        sentences.add(sentence);
    }
    aTarget.set(SentencesAnnotation.class, sentences);

    return aTarget;
}

---