Examples with SemgrexMatcher edu.stanford.nlp.semgraph.semgrex.SemgrexMatcher used on opensource projects

Example 16 with SemgrexMatcher

use of edu.stanford.nlp.semgraph.semgrex.SemgrexMatcher in project CoreNLP by stanfordnlp.

the class ExtractPhraseFromPattern method getSemGrexPatternNodes.

/*
   * Given a SemanticGraph g and a SemgrexPattern pattern
   * And a bunch of other parameters,
   * run the pattern matcher (get SemgrexMatcher m)
   * Iterate through to get matching words/phrases
   * 
   * Next, gets matchedGraphsForPattern.get(pattern),
   * a list of matched (String, semgraph) pairs
   * and adds the new graph and tokens if matched.
   * 
   * I need to clarify what's going on with tokens.
   */
public Set<IndexedWord> getSemGrexPatternNodes(SemanticGraph g, List<String> tokens, Collection<String> outputNodes, Collection<IntPair> outputIndices, SemgrexPattern pattern, boolean findSubTrees, Collection<ExtractedPhrase> extractedPhrases, boolean lowercase, Function<CoreLabel, Boolean> acceptWord) {
    Set<IndexedWord> foundWordsParents = new HashSet<>();
    SemgrexMatcher m = pattern.matcher(g, lowercase);
    while (m.find()) {
        IndexedWord w = m.getNode("node");
        //System.out.println("found a match for " + pattern.pattern());
        IndexedWord parent = m.getNode("parent");
        boolean ifSatisfiedMaxDepth = checkIfSatisfiedMaxDepth(g, parent, w, new IntPair(maxDepth, 0));
        if (ifSatisfiedMaxDepth == false)
            continue;
        if (DEBUG > 3) {
            List<Pair<String, SemanticGraph>> matchedGraphs = matchedGraphsForPattern.get(pattern);
            if (matchedGraphs == null)
                matchedGraphs = new ArrayList<>();
            matchedGraphs.add(new Pair<>(StringUtils.join(tokens, " "), g));
            //if (DEBUG >= 3)
            //  System.out.println("matched pattern is " + pattern);
            matchedGraphsForPattern.put(pattern, matchedGraphs);
        }
        foundWordsParents.add(parent);
        // String relationName = m.getRelnString("reln");
        // System.out.println("word is " + w.lemma() + " and " + w.tag());
        ArrayList<IndexedWord> seenNodes = new ArrayList<>();
        List<String> cutoffrelations = new ArrayList<>();
        //      if (elementStr.equalsIgnoreCase("technique"))
        //        cutoffrelations = cutoffRelationsForTech;
        //      if (elementStr.equalsIgnoreCase("app"))
        //        cutoffrelations = this.cuttoffRelationsForApp;
        //System.out.println("g is ");
        //g.prettyPrint();
        printSubGraph(g, w, cutoffrelations, tokens, outputNodes, outputIndices, seenNodes, new ArrayList<>(), findSubTrees, extractedPhrases, pattern, acceptWord);
    }
    return foundWordsParents;
}

Example 17 with SemgrexMatcher

use of edu.stanford.nlp.semgraph.semgrex.SemgrexMatcher in project CoreNLP by stanfordnlp.

the class SemgrexDemo method main.

public static void main(String[] args) {
    String treeString = "(ROOT  (S (NP (PRP$ My) (NN dog)) (ADVP (RB also)) (VP (VBZ likes) (S (VP (VBG eating) (NP (NN sausage))))) (. .)))";
    // Typically the tree is constructed by parsing or reading a
    // treebank.  This is just for example purposes
    Tree tree = Tree.valueOf(treeString);
    // This creates English uncollapsed dependencies as a
    // SemanticGraph.  If you are creating many SemanticGraphs, you
    // should use a GrammaticalStructureFactory and use it to generate
    // the intermediate GrammaticalStructure instead
    SemanticGraph graph = SemanticGraphFactory.generateUncollapsedDependencies(tree);
    // Alternatively, this could have been the Chinese params or any
    // other language supported.  As of 2014, only English and Chinese
    TreebankLangParserParams params = new EnglishTreebankParserParams();
    GrammaticalStructureFactory gsf = params.treebankLanguagePack().grammaticalStructureFactory(params.treebankLanguagePack().punctuationWordRejectFilter(), params.typedDependencyHeadFinder());
    GrammaticalStructure gs = gsf.newGrammaticalStructure(tree);
    log.info(graph);
    SemgrexPattern semgrex = SemgrexPattern.compile("{}=A <<nsubj {}=B");
    SemgrexMatcher matcher = semgrex.matcher(graph);
    // ancestor of both "dog" and "my" via the nsubj relation
    while (matcher.find()) {
        log.info(matcher.getNode("A") + " <<nsubj " + matcher.getNode("B"));
    }
}

Example 18 with SemgrexMatcher

use of edu.stanford.nlp.semgraph.semgrex.SemgrexMatcher in project CoreNLP by stanfordnlp.

the class Sentence method semgrex.

/**
   * Apply a semgrex pattern to the sentence
   * @param pattern The Semgrex pattern to match against.
   * @param fn The action to do on each match.
   * @return the list of matches, after run through the function.
   */
public <T> List<T> semgrex(SemgrexPattern pattern, Function<SemgrexMatcher, T> fn) {
    SemgrexMatcher matcher = pattern.matcher(dependencyGraph());
    List<T> lst = new ArrayList<>();
    while (matcher.findNextMatchingNode()) {
        lst.add(fn.apply(matcher));
    }
    return lst;
}

Example 19 with SemgrexMatcher

use of edu.stanford.nlp.semgraph.semgrex.SemgrexMatcher in project CoreNLP by stanfordnlp.

the class EnglishGrammaticalStructure method correctWHAttachment.

/**
   * Tries to correct complicated cases of WH-movement in
   * sentences such as "What does Mary seem to have?" in
   * which "What" should attach to "have" instead of the
   * control verb.
   *
   * @param sg The Semantic graph to operate on.
   */
private static void correctWHAttachment(SemanticGraph sg) {
    /* Semgrexes require a graph with a root. */
    if (sg.getRoots().isEmpty())
        return;
    SemanticGraph sgCopy = sg.makeSoftCopy();
    SemgrexMatcher matcher = XCOMP_PATTERN.matcher(sgCopy);
    while (matcher.findNextMatchingNode()) {
        IndexedWord root = matcher.getNode("root");
        IndexedWord embeddedVerb = matcher.getNode("embedded");
        IndexedWord wh = matcher.getNode("wh");
        IndexedWord dobj = matcher.getNode("obj");
        /* Check if the object is a WH-word. */
        if (wh.tag().startsWith("W")) {
            boolean reattach = false;
            /* If the control verb already has an object, then
           we have to reattach th WH-word to the verb in the embedded clause. */
            if (dobj != null) {
                reattach = true;
            } else {
                /* If the control verb can't have an object, we also have to reattach. */
                String lemma = Morphology.lemmaStatic(root.value(), root.tag());
                if (lemma.matches(EnglishPatterns.NP_V_S_INF_VERBS_REGEX)) {
                    reattach = true;
                }
            }
            if (reattach) {
                SemanticGraphEdge edge = sg.getEdge(root, wh);
                if (edge != null) {
                    sg.removeEdge(edge);
                    sg.addEdge(embeddedVerb, wh, DIRECT_OBJECT, Double.NEGATIVE_INFINITY, false);
                }
            }
        }
    }
}

Example 20 with SemgrexMatcher

use of edu.stanford.nlp.semgraph.semgrex.SemgrexMatcher in project CoreNLP by stanfordnlp.

the class UniversalEnglishGrammaticalStructure method processSimple2WP.

/**
   * Processes all the two-word prepositions in TWO_WORD_PREPS_REGULAR.
   */
private static void processSimple2WP(SemanticGraph sg, HashMap<String, HashSet<Integer>> bigrams) {
    for (String bigram : TWO_WORD_PREPS_REGULAR) {
        if (bigrams.get(bigram) == null) {
            continue;
        }
        for (Integer i : bigrams.get(bigram)) {
            IndexedWord w1 = sg.getNodeByIndexSafe(i);
            IndexedWord w2 = sg.getNodeByIndexSafe(i + 1);
            if (w1 == null || w2 == null) {
                continue;
            }
            SemgrexMatcher matcher = TWO_WORD_PREPS_REGULAR_PATTERN.matcher(sg);
            IndexedWord gov = null;
            while (matcher.find()) {
                if (w1.equals(matcher.getNode("w1")) && w2.equals(matcher.getNode("w2"))) {
                    gov = matcher.getNode("gov");
                    break;
                }
            }
            if (gov == null) {
                continue;
            }
            createMultiWordExpression(sg, gov, CASE_MARKER, w1, w2);
        }
    }
}