Examples of edu.stanford.nlp.parser.common.ParserConstraint

edu.stanford.nlp.parser.common.ParserConstraint
A Constraint represents a restriction on possible parse trees to consider. It says that a parse cannot be postulated that would contradict having a constituent from position start to end, and that any constituent postulated with this span must match the regular expression given by state. Note, however, that it does not strictly guarantee that such a constituent exists in a returned parse.
The words in constraint bounds are counted starting from 0. Furthermore, the end number is not included in the constraint. For example, a constraint covering the first two words of a sentence has a start of 0 and an end of 2.
state must successfully match states used internal to the parser, not just the final states, so the pattern "NP" will not match anything. Better is "NP.*". Note that this does run the risk of matching states with the same prefix. For example, there is SBAR and SBARQ, so "SBAR.*" could have unexpected results. The states used internal to the parser extend the state name with non-alphanumeric characters, so a fancy expression such as "SBAR|SBAR[^a-zA-Z].*" would match SBAR but not SBARQ constituents.

      // constrain the parse to the part we're interested in.
      // Starting from ADDED_WORDS comes from skipping "It was".
      // -1 to exclude the period.
      // We now let it be any kind of nominal constituent, since there
      // are VP and S ones
      ParserConstraint constraint = new ParserConstraint(ADDED_WORDS, extentTokens.size() - 1, Pattern.compile(".*"));
      List<ParserConstraint> constraints = Collections.singletonList(constraint);
      Tree tree = parse(extentTokens, constraints);
      convertToCoreLabels(tree);  // now unnecessary, as parser uses CoreLabels?
      tree.indexSpans(m.startIndex - ADDED_WORDS);  // remember it has ADDED_WORDS extra words at the beginning
      Tree subtree = findPartialSpan(tree, m.startIndex);

View Full Code Here

    CoreMap sentence = annotation.get(CoreAnnotations.SentencesAnnotation.class).get(0);


    parserOnlyPipeline.annotate(annotation);
    assertEquals(expectedResult, sentence.get(TreeCoreAnnotations.TreeAnnotation.class).toString());


    ParserConstraint constraint = new ParserConstraint(0, 2, "SBAR|SBAR[^a-zA-Z].*");
    List<ParserConstraint> constraints = new ArrayList<ParserConstraint>();
    constraints.add(constraint);
    sentence.set(ConstraintAnnotation.class, constraints);


    parserOnlyPipeline.annotate(annotation);

View Full Code Here

    // constrain the parse to the part we're interested in.
    // Starting from ADDED_WORDS comes from skipping "It was".
    // -1 to exclude the period.
    // We now let it be any kind of nominal constituent, since there
    // are VP and S ones
    ParserConstraint constraint = new ParserConstraint(ADDED_WORDS, extentTokens.size() - 1, ".*");
    List<ParserConstraint> constraints = Collections.singletonList(constraint);
    Tree tree = parse(extentTokens, constraints);
    logger.fine("No exact match found. Local parse:\n" + tree.pennString());
    convertToCoreLabels(tree);
    tree.indexSpans(ent.getExtentTokenStart() - ADDED_WORDS);  // remember it has ADDED_WORDS extra words at the beginning

View Full Code Here

    englishTagger.tagCoreLabels(sentence);
    Tree result = englishParser.apply(sentence);
    // pretty much need to make the test rely on the parser being consistent
    assertEquals("(ROOT (S (NP (PRP It)) (VP (VBD was) (NP (NNP Carolina) (NNP Reapers))) (. .)))", result.toString());


    ParserConstraint constraint = new ParserConstraint(2, 4, ".*");
    List<ParserConstraint> constraints = Collections.singletonList(constraint);
    ParserQuery pq = englishParser.parserQuery();
    pq.setConstraints(constraints);
    assertTrue(pq.parse(sentence));
    result = pq.getBestParse();
    assertEquals("(ROOT (S (NP (PRP It)) (VP (VBD was) (NP (NNP Carolina) (NNP Reapers))) (. .)))", result.toString());


    constraint = new ParserConstraint(2, 4, "NP");
    constraints = Collections.singletonList(constraint);
    pq = englishParser.parserQuery();
    pq.setConstraints(constraints);
    assertTrue(pq.parse(sentence));
    result = pq.getBestParse();
    assertEquals("(ROOT (S (NP (PRP It)) (VP (VBD was) (NP (NNP Carolina) (NNP Reapers))) (. .)))", result.toString());


    constraint = new ParserConstraint(2, 4, "ADJP");
    constraints = Collections.singletonList(constraint);
    pq = englishParser.parserQuery();
    pq.setConstraints(constraints);
    assertTrue(pq.parse(sentence));
    result = pq.getBestParse();
    assertEquals("(ROOT (S (NP (PRP It)) (VP (VBD was) (ADJP (NP (NNP Carolina) (NNP Reapers)))) (. .)))", result.toString());


    constraint = new ParserConstraint(1, 3, "VP");
    constraints = Collections.singletonList(constraint);
    pq = englishParser.parserQuery();
    pq.setConstraints(constraints);
    assertTrue(pq.parse(sentence));
    result = pq.getBestParse();

View Full Code Here

  public void testConstraints() {
    List<CoreLabel> sentence = sampleSausage();


    ParserQuery pq = englishParser.parserQuery();


    ParserConstraint constraint =
      new ParserConstraint(0, 2, "SBAR|SBAR[^a-zA-Z].*");
    List<ParserConstraint> constraints = new ArrayList<ParserConstraint>();
    constraints.add(constraint);
    pq.setConstraints(constraints);


    pq.parse(sentence);

View Full Code Here

        for(int nodeName : nodes) {
          if(lastBoundaryNode == -1) {
            assert nodeName % NODE_OFFSET == 0;
            lastBoundaryNode = realNodeIdx;
          } else if(nodeName % NODE_OFFSET == 0) {
            ParserConstraint c = new ParserConstraint(lastBoundaryNode, realNodeIdx, ".*");
            lattice.addConstraint(c);
          }


          nodeMap.put(nodeName, realNodeIdx);
          realNodeIdx++;

View Full Code Here

TOP

Related Classes of edu.stanford.nlp.parser.common.ParserConstraint

edu.stanford.nlp.dcoref.RuleBasedCorefMentionFinder

edu.stanford.nlp.ie.machinereading.GenericDataSetReader

edu.stanford.nlp.parser.lexparser.LatticeXMLReader

edu.stanford.nlp.parser.lexparser.LexicalizedParserITest

edu.stanford.nlp.parser.shiftreduce.ShiftReduceParserITest

edu.stanford.nlp.pipeline.ParserAnnotatorITest

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.