Examples with CoreLabel edu.stanford.nlp.ling.CoreLabel used on opensource projects

Example 46 with CoreLabel

use of edu.stanford.nlp.ling.CoreLabel in project CoreNLP by stanfordnlp.

the class Mention method lowestNPIncludesHead.

public String lowestNPIncludesHead() {
    String ret = "";
    Tree head = this.contextParseTree.getLeaves().get(this.headIndex);
    Tree lowestNP = head;
    String s;
    while (true) {
        if (lowestNP == null)
            return ret;
        s = ((CoreLabel) lowestNP.label()).get(CoreAnnotations.ValueAnnotation.class);
        if (s.equals("NP") || s.equals("ROOT"))
            break;
        lowestNP = lowestNP.ancestor(1, this.contextParseTree);
    }
    if (s.equals("ROOT"))
        lowestNP = head;
    for (Tree t : lowestNP.getLeaves()) {
        if (!ret.equals(""))
            ret = ret + " ";
        ret = ret + ((CoreLabel) t.label()).get(CoreAnnotations.TextAnnotation.class);
    }
    if (!this.spanToString().contains(ret))
        return this.sentenceWords.get(this.headIndex).get(CoreAnnotations.TextAnnotation.class);
    return ret;
}

Example 47 with CoreLabel

use of edu.stanford.nlp.ling.CoreLabel in project CoreNLP by stanfordnlp.

the class Mention method nerTokens.

// Retrieves part of the span that corresponds to the NER (going out from head)
public List<CoreLabel> nerTokens() {
    if (nerString == null || "O".equals(nerString))
        return null;
    int start = headIndex - startIndex;
    int end = headIndex - startIndex + 1;
    while (start > 0) {
        CoreLabel prev = originalSpan.get(start - 1);
        if (nerString.equals(prev.ner())) {
            start--;
        } else {
            break;
        }
    }
    while (end < originalSpan.size()) {
        CoreLabel next = originalSpan.get(end);
        if (nerString.equals(next.ner())) {
            end++;
        } else {
            break;
        }
    }
    return originalSpan.subList(start, end);
}

Example 48 with CoreLabel

use of edu.stanford.nlp.ling.CoreLabel in project CoreNLP by stanfordnlp.

the class CoNLLDocumentReader method writeTabSep.

public static void writeTabSep(PrintWriter pw, CoreMap sentence, CollectionValuedMap<String, CoreMap> chainmap) {
    HeadFinder headFinder = new ModCollinsHeadFinder();
    List<CoreLabel> sentenceAnno = sentence.get(CoreAnnotations.TokensAnnotation.class);
    Tree sentenceTree = sentence.get(TreeCoreAnnotations.TreeAnnotation.class);
    Map<Pair<Integer, Integer>, String> sentenceInfo = Generics.newHashMap();
    Set<Tree> sentenceSubTrees = sentenceTree.subTrees();
    sentenceTree.setSpans();
    Map<Pair<Integer, Integer>, Tree> treeSpanMap = Generics.newHashMap();
    Map<Pair<Integer, Integer>, List<Tree>> wordSpanMap = Generics.newHashMap();
    for (Tree ctree : sentenceSubTrees) {
        IntPair span = ctree.getSpan();
        if (span != null) {
            treeSpanMap.put(Pair.makePair(span.getSource(), span.getTarget()), ctree);
            wordSpanMap.put(Pair.makePair(span.getSource(), span.getTarget()), ctree.getLeaves());
        }
    }
    String[][] finalSentence;
    finalSentence = new String[sentenceAnno.size()][];
    Map<Pair<Integer, Integer>, String> allHeads = Generics.newHashMap();
    int index = -1;
    for (CoreLabel newAnno : sentenceAnno) {
        index += 1;
        String word = newAnno.word();
        String tag = newAnno.tag();
        String cat = newAnno.ner();
        String coref = newAnno.get(CorefCoreAnnotations.CorefAnnotation.class);
        finalSentence[index] = new String[4];
        finalSentence[index][0] = word;
        finalSentence[index][1] = tag;
        finalSentence[index][2] = cat;
        finalSentence[index][3] = coref;
        if (coref == null) {
            sentenceInfo.put(Pair.makePair(index, index), coref);
            finalSentence[index][3] = "O";
        } else {
            String[] allC = coref.split("\\|");
            for (String corefG : allC) {
                Pair<Integer, Integer> mention = getMention(index, corefG, sentenceAnno);
                if (!include(sentenceInfo, mention, corefG)) {
                    // find largest NP in mention
                    sentenceInfo.put(mention, corefG);
                    Tree mentionTree = treeSpanMap.get(mention);
                    String head = null;
                    if (mentionTree != null) {
                        head = mentionTree.headTerminal(headFinder).nodeString();
                    } else if (mention.first.equals(mention.second)) {
                        head = word;
                    }
                    allHeads.put(mention, head);
                }
            }
            if (allHeads.values().contains(word)) {
                finalSentence[index][3] = "MENTION";
            } else {
                finalSentence[index][3] = "O";
            }
        }
    }
    for (int i = 0; i < finalSentence.length; i++) {
        String[] wordInfo = finalSentence[i];
        if (i < finalSentence.length - 1) {
            String[] nextWordInfo = finalSentence[i + 1];
            if (nextWordInfo[3].equals("MENTION") && nextWordInfo[0].equals("'s")) {
                wordInfo[3] = "MENTION";
                finalSentence[i + 1][3] = "O";
            }
        }
        pw.println(wordInfo[0] + "\t" + wordInfo[1] + "\t" + wordInfo[2] + "\t" + wordInfo[3]);
    }
    pw.println("");
}

Example 49 with CoreLabel

use of edu.stanford.nlp.ling.CoreLabel in project CoreNLP by stanfordnlp.

the class DocumentMaker method makeDocument.

public Document makeDocument(InputDoc input) throws Exception {
    List<List<Mention>> mentions = new ArrayList<>();
    if (CorefProperties.useGoldMentions(props)) {
        List<CoreMap> sentences = input.annotation.get(CoreAnnotations.SentencesAnnotation.class);
        for (int i = 0; i < sentences.size(); i++) {
            CoreMap sentence = sentences.get(i);
            List<CoreLabel> sentenceWords = sentence.get(CoreAnnotations.TokensAnnotation.class);
            List<Mention> sentenceMentions = new ArrayList<>();
            mentions.add(sentenceMentions);
            for (Mention g : input.goldMentions.get(i)) {
                sentenceMentions.add(new Mention(-1, g.startIndex, g.endIndex, sentenceWords, null, null, new ArrayList<>(sentenceWords.subList(g.startIndex, g.endIndex))));
            }
            md.findHead(sentence, sentenceMentions);
        }
    } else {
        for (CoreMap sentence : input.annotation.get(CoreAnnotations.SentencesAnnotation.class)) {
            mentions.add(sentence.get(CorefCoreAnnotations.CorefMentionsAnnotation.class));
        }
    }
    Document doc = new Document(input, mentions);
    if (input.goldMentions != null) {
        findGoldMentionHeads(doc);
    }
    DocumentPreprocessor.preprocess(doc, dict, null, headFinder);
    return doc;
}

Example 50 with CoreLabel

use of edu.stanford.nlp.ling.CoreLabel in project CoreNLP by stanfordnlp.

the class DocumentPreprocessor method setParagraphAnnotation.

/** Set paragraph index */
private static void setParagraphAnnotation(Document doc) {
    int paragraphIndex = 0;
    int previousOffset = -10;
    for (CoreMap sent : doc.annotation.get(CoreAnnotations.SentencesAnnotation.class)) {
        for (CoreLabel w : sent.get(CoreAnnotations.TokensAnnotation.class)) {
            if (w.containsKey(CoreAnnotations.CharacterOffsetBeginAnnotation.class)) {
                if (w.get(CoreAnnotations.CharacterOffsetBeginAnnotation.class) > previousOffset + 2)
                    paragraphIndex++;
                w.set(CoreAnnotations.ParagraphAnnotation.class, paragraphIndex);
                previousOffset = w.get(CoreAnnotations.CharacterOffsetEndAnnotation.class);
            } else {
                w.set(CoreAnnotations.ParagraphAnnotation.class, -1);
            }
        }
    }
    for (List<Mention> l : doc.predictedMentions) {
        for (Mention m : l) {
            m.paragraph = m.headWord.get(CoreAnnotations.ParagraphAnnotation.class);
        }
    }
    doc.numParagraph = paragraphIndex;
}