edu.stanford.nlp.coref.neural.EmbeddingExtractor Maven / Gradle / Ivy

Show more of this group Show more artifacts with this name
Show all versions of stanford-corenlp Show documentation

Stanford CoreNLP provides a set of natural language analysis tools which can take raw English language text input and give the base forms of words, their parts of speech, whether they are names of companies, people, etc., normalize dates, times, and numeric quantities, mark up the structure of sentences in terms of phrases and word dependencies, and indicate which noun phrases refer to the same entities. It provides the foundational building blocks for higher level text understanding applications.

There is a newer version: 4.5.7

Show newest version

package edu.stanford.nlp.coref.neural;

import java.util.ArrayList;
import java.util.HashSet;
import java.util.Iterator;
import java.util.List;
import java.util.Set;

import org.ejml.simple.SimpleMatrix;

import edu.stanford.nlp.coref.data.Document;
import edu.stanford.nlp.coref.data.Mention;
import edu.stanford.nlp.ling.CoreLabel;
import edu.stanford.nlp.neural.Embedding;
import edu.stanford.nlp.neural.NeuralUtils;
import edu.stanford.nlp.semgraph.SemanticGraphEdge;

/**
 * Extracts word-embedding features from mentions.
 * @author Kevin Clark
 */
public class EmbeddingExtractor {
  private final boolean conll;
  private final Embedding staticWordEmbeddings;
  private final Embedding tunedWordEmbeddings;

  public EmbeddingExtractor(boolean conll, Embedding staticWordEmbeddings,
      Embedding tunedWordEmbeddings) {
    this.conll = conll;
    this.staticWordEmbeddings = staticWordEmbeddings;
    this.tunedWordEmbeddings = tunedWordEmbeddings;
  }

  public SimpleMatrix getDocumentEmbedding(Document document) {
    if (!conll) {
      return new SimpleMatrix(staticWordEmbeddings.getEmbeddingSize(), 1);
    }
    List words = new ArrayList<>();
    Set seenSentences = new HashSet<>();
    for (Mention m : document.predictedMentionsByID.values()) {
      if (!seenSentences.contains(m.sentNum)) {
        seenSentences.add(m.sentNum);
        words.addAll(m.sentenceWords);
      }
    }
    return getAverageEmbedding(words);
  }

  public SimpleMatrix getMentionEmbeddings(Mention m, SimpleMatrix docEmbedding) {
    Iterator depIterator =
        m.enhancedDependency.incomingEdgeIterator(m.headIndexedWord);
    SemanticGraphEdge depRelation = depIterator.hasNext() ? depIterator.next() : null;

    return NeuralUtils.concatenate(
        getAverageEmbedding(m.sentenceWords, m.startIndex, m.endIndex),
        getAverageEmbedding(m.sentenceWords, m.startIndex - 5, m.startIndex),
        getAverageEmbedding(m.sentenceWords, m.endIndex, m.endIndex + 5),
        getAverageEmbedding(m.sentenceWords.subList(0, m.sentenceWords.size() - 1)),
        docEmbedding,
        getWordEmbedding(m.sentenceWords, m.headIndex),
        getWordEmbedding(m.sentenceWords, m.startIndex),
        getWordEmbedding(m.sentenceWords, m.endIndex - 1),
        getWordEmbedding(m.sentenceWords, m.startIndex - 1),
        getWordEmbedding(m.sentenceWords, m.endIndex),
        getWordEmbedding(m.sentenceWords, m.startIndex - 2),
        getWordEmbedding(m.sentenceWords, m.endIndex + 1),
        getWordEmbedding(depRelation == null ? null : depRelation.getSource().word())
    );
  }

  private SimpleMatrix getAverageEmbedding(List words) {
    SimpleMatrix emb = new SimpleMatrix(staticWordEmbeddings.getEmbeddingSize(), 1);
    for (CoreLabel word : words) {
      emb = emb.plus(getStaticWordEmbedding(word.word()));
    }
    return emb.divide(Math.max(1, words.size()));
  }

  private SimpleMatrix getAverageEmbedding(List sentence, int start, int end) {
    return getAverageEmbedding(sentence.subList(Math.max(Math.min(start, sentence.size() - 1), 0),
        Math.max(Math.min(end, sentence.size() - 1), 0)));
  }

  private SimpleMatrix getWordEmbedding(List sentence, int i) {
    return getWordEmbedding(i < 0 || i >= sentence.size() ? null : sentence.get(i).word());
  }

  public SimpleMatrix getWordEmbedding(String word) {
    word = normalizeWord(word);
    return tunedWordEmbeddings.containsWord(word) ? tunedWordEmbeddings.get(word) :
      staticWordEmbeddings.get(word);
  }

  public SimpleMatrix getStaticWordEmbedding(String word) {
    return staticWordEmbeddings.get(normalizeWord(word));
  }

  private static String normalizeWord(String w) {
    if (w == null) {
      return "";
    } else if (w.equals("/.")) {
      return ".";
    } else if (w.equals("/?")) {
      return "?";
    } else if (w.equals("-LRB-")) {
      return "(";
    } else if (w.equals("-RRB-")) {
      return ")";
    } else if (w.equals("-LCB-")) {
      return "{";
    } else if (w.equals("-RCB-")) {
      return "}";
    } else if (w.equals("-LSB-")) {
      return "[";
    } else if (w.equals("-RSB-")) {
      return "]";
    }
    return w.replaceAll("\\d", "0").toLowerCase();
  }
}