Package edu.stanford.nlp.trees

Examples of edu.stanford.nlp.trees.LabeledScoredTreeNode$TreeFactoryHolder


      // Internally this parser uses the index, so we have to
      // overwrite incorrect indices if the label is already indexed
      wordLabel.setIndex(index + 1);
      tagLabel.setIndex(index + 1);

      LabeledScoredTreeNode wordNode = new LabeledScoredTreeNode(wordLabel);
      LabeledScoredTreeNode tagNode = new LabeledScoredTreeNode(tagLabel);
      tagNode.addChild(wordNode);

      wordLabel.set(TreeCoreAnnotations.HeadWordAnnotation.class, wordNode);
      wordLabel.set(TreeCoreAnnotations.HeadTagAnnotation.class, tagNode);
      tagLabel.set(TreeCoreAnnotations.HeadWordAnnotation.class, wordNode);
      tagLabel.set(TreeCoreAnnotations.HeadTagAnnotation.class, tagNode);
View Full Code Here


   *
   * @param proto The serialized tree.
   * @return A Tree object corresponding to the saved tree. This will always be a {@link LabeledScoredTreeNode}.
   */
  public Tree fromProto(CoreNLPProtos.ParseTree proto) {
    LabeledScoredTreeNode node = new LabeledScoredTreeNode();
    // Set label
    if (proto.hasValue()) {
      CoreLabel value = new CoreLabel();
      value.setCategory(proto.getValue());
      value.setValue(proto.getValue());
      node.setLabel(value);
      // Set span
      if (proto.hasYieldBeginIndex() && proto.hasYieldEndIndex()) {
        IntPair span = new IntPair(proto.getYieldBeginIndex(), proto.getYieldEndIndex());
        value.set(SpanAnnotation.class, span);
      }
    }
    // Set score
    if (proto.hasScore()) { node.setScore(proto.getScore()); }
    // Set children
    Tree[] children = new LabeledScoredTreeNode[proto.getChildCount()];
    for (int i = 0; i < children.length; ++i) {
      children[i] = fromProto(proto.getChild(i));
    }
    node.setChildren(children);
    // Return
    return node;
  }
View Full Code Here

    Tree[] subtrees = new Tree[maxNode + 1];
    for (int i = 0; i < sentence.size(); ++i) {
      CoreLabel word = new CoreLabel();
      word.setValue(sentence.get(i));
      Tree leaf = new LabeledScoredTreeNode(word);
      subtrees[i] = new LabeledScoredTreeNode(new CoreLabel());
      subtrees[i].addChild(leaf);
    }

    for (int i = sentence.size(); i <= maxNode; ++i) {
      subtrees[i] = new LabeledScoredTreeNode(new CoreLabel());
    }

    boolean[] connected = new boolean[maxNode + 1];
    Tree root = null;
    for (int index = 0; index < parentPointers.size(); ++index) {
      if (parentPointers.get(index) == -1) {
        if (root != null) {
          throw new RuntimeException("Found two roots for sentence " + sentence);
        }
        root = subtrees[index];
      } else {
        // Walk up the tree structure to make sure that leftmost
        // phrases are added first.  Otherwise, if the numbers are
        // inverted, we might get the right phrase added to a parent
        // first, resulting in "case zero in this", for example,
        // instead of "in this case zero"
        // Note that because we keep track of which ones are already
        // connected, we process this at most once per parent, so the
        // overall construction time is still efficient.
        connect(parentPointers, subtrees, connected, index);
      }
    }

    for (int i = 0; i <= maxNode; ++i) {
      List<Tree> leaves = subtrees[i].getLeaves();
      List<String> words = CollectionUtils.transformAsList(leaves, TRANSFORM_TREE_TO_WORD);
      // First we look for a copy of the phrase with -LRB- -RRB-
      // instead of ().  The sentiment trees sometimes have both, and
      // the escaped versions seem to have more reasonable scores.
      // If a particular phrase doesn't have -LRB- -RRB- we fall back
      // to the unescaped versions.
      Integer phraseId = phraseIds.get(CollectionUtils.transformAsList(words, TRANSFORM_PARENS));
      if (phraseId == null) {
        phraseId = phraseIds.get(words);
      }
      if (phraseId == null) {
        throw new RuntimeException("Could not find phrase id for phrase " + sentence);
      }
      // TODO: should we make this an option?  Perhaps we want cases
      // where the trees have the phrase id and not their class
      Double score = sentimentScores.get(phraseId);
      if (score == null) {
        throw new RuntimeException("Could not find sentiment score for phrase id " + phraseId);
      }
      // TODO: make this a numClasses option
      int classLabel = Math.round((float) Math.floor(score * 5.0));
      if (classLabel > 4) {
        classLabel = 4;
      }
      subtrees[i].label().setValue(Integer.toString(classLabel));
    }

    for (int i = 0; i < sentence.size(); ++i) {
      Tree leaf = subtrees[i].children()[0];
      leaf.label().setValue(escaper.escapeString(leaf.label().value()));
    }

    for (int i = 0; i < tregexPatterns.length; ++i) {
      root = Tsurgeon.processPattern(tregexPatterns[i], tsurgeonPatterns[i], root);
    }
View Full Code Here

    CoreLabel headLabel = (CoreLabel) top.label();
    CoreLabel production = new CoreLabel();
    production.setValue(label);
    production.set(TreeCoreAnnotations.HeadWordAnnotation.class, headLabel.get(TreeCoreAnnotations.HeadWordAnnotation.class));
    production.set(TreeCoreAnnotations.HeadTagAnnotation.class, headLabel.get(TreeCoreAnnotations.HeadTagAnnotation.class));
    Tree newTop = new LabeledScoredTreeNode(production);
    for (Tree child : children) {
      newTop.addChild(child);
    }
    return newTop;
  }
View Full Code Here

    CoreLabel production = new CoreLabel();
    production.setValue(label);
    production.set(TreeCoreAnnotations.HeadWordAnnotation.class, headLabel.get(TreeCoreAnnotations.HeadWordAnnotation.class));
    production.set(TreeCoreAnnotations.HeadTagAnnotation.class, headLabel.get(TreeCoreAnnotations.HeadTagAnnotation.class));
    Tree newTop = new LabeledScoredTreeNode(production);
    newTop.addChild(left);
    newTop.addChild(right);

    stack = stack.push(newTop);

    return new State(stack, state.transitions.push(this), state.separators, state.sentence, state.tokenPosition, state.score + scoreDelta, false);   
  }
View Full Code Here

TOP

Related Classes of edu.stanford.nlp.trees.LabeledScoredTreeNode$TreeFactoryHolder

Copyright © 2018 www.massapicom. All rights reserved.
All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.