List of usage examples for edu.stanford.nlp.trees TreebankLanguagePack grammaticalStructureFactory
GrammaticalStructureFactory grammaticalStructureFactory(Predicate<String> puncFilter,
HeadFinder typedDependencyHF);
From source file:de.tudarmstadt.ukp.dkpro.core.stanfordnlp.StanfordCoreferenceResolver.java
License:Open Source License
@Override public void process(JCas aJCas) throws AnalysisEngineProcessException { modelProvider.configure(aJCas.getCas()); List<Tree> trees = new ArrayList<Tree>(); List<CoreMap> sentences = new ArrayList<CoreMap>(); List<List<CoreLabel>> sentenceTokens = new ArrayList<List<CoreLabel>>(); for (ROOT root : select(aJCas, ROOT.class)) { // Copy all relevant information from the tokens List<CoreLabel> tokens = new ArrayList<CoreLabel>(); for (Token token : selectCovered(Token.class, root)) { tokens.add(tokenToWord(token)); }//from w w w. j a v a 2 s . c o m sentenceTokens.add(tokens); // SemanticHeadFinder (nonTerminalInfo) does not know about PRN0, so we have to replace // it with PRN to avoid NPEs. TreeFactory tFact = new LabeledScoredTreeFactory(CoreLabel.factory()) { @Override public Tree newTreeNode(String aParent, List<Tree> aChildren) { String parent = aParent; if ("PRN0".equals(parent)) { parent = "PRN"; } Tree node = super.newTreeNode(parent, aChildren); return node; } }; // deep copy of the tree. These are modified inside coref! Tree treeCopy = TreeUtils.createStanfordTree(root, tFact).treeSkeletonCopy(); treeCopy.indexSpans(); trees.add(treeCopy); // Build the sentence CoreMap sentence = new CoreLabel(); sentence.set(TreeAnnotation.class, treeCopy); sentence.set(TokensAnnotation.class, tokens); sentence.set(RootKey.class, root); sentences.add(sentence); // https://code.google.com/p/dkpro-core-asl/issues/detail?id=590 // We currently do not copy over dependencies from the CAS. This is supposed to fill // in the dependencies so we do not get NPEs. TreebankLanguagePack tlp = new PennTreebankLanguagePack(); GrammaticalStructureFactory gsf = tlp.grammaticalStructureFactory(tlp.punctuationWordRejectFilter(), tlp.typedDependencyHeadFinder()); ParserAnnotatorUtils.fillInParseAnnotations(false, true, gsf, sentence, treeCopy, GrammaticalStructure.Extras.NONE); // https://code.google.com/p/dkpro-core-asl/issues/detail?id=582 SemanticGraph deps = sentence.get(SemanticGraphCoreAnnotations.CollapsedDependenciesAnnotation.class); for (IndexedWord vertex : deps.vertexSet()) { vertex.setWord(vertex.value()); } // merge the new CoreLabels with the tree leaves MentionExtractor.mergeLabels(treeCopy, tokens); MentionExtractor.initializeUtterance(tokens); } Annotation document = new Annotation(aJCas.getDocumentText()); document.set(SentencesAnnotation.class, sentences); Coreferencer coref = modelProvider.getResource(); // extract all possible mentions // Reparsing only works when the full CoreNLP pipeline system is set up! Passing false here // disables reparsing. RuleBasedCorefMentionFinder finder = new RuleBasedCorefMentionFinder(false); List<List<Mention>> allUnprocessedMentions = finder.extractPredictedMentions(document, 0, coref.corefSystem.dictionaries()); // add the relevant info to mentions and order them for coref Map<Integer, CorefChain> result; try { Document doc = coref.mentionExtractor.arrange(document, sentenceTokens, trees, allUnprocessedMentions); result = coref.corefSystem.coref(doc); } catch (Exception e) { throw new AnalysisEngineProcessException(e); } for (CorefChain chain : result.values()) { CoreferenceLink last = null; for (CorefMention mention : chain.getMentionsInTextualOrder()) { CoreLabel beginLabel = sentences.get(mention.sentNum - 1).get(TokensAnnotation.class) .get(mention.startIndex - 1); CoreLabel endLabel = sentences.get(mention.sentNum - 1).get(TokensAnnotation.class) .get(mention.endIndex - 2); CoreferenceLink link = new CoreferenceLink(aJCas, beginLabel.get(TokenKey.class).getBegin(), endLabel.get(TokenKey.class).getEnd()); if (mention.mentionType != null) { link.setReferenceType(mention.mentionType.toString()); } if (last == null) { // This is the first mention. Here we'll initialize the chain CoreferenceChain corefChain = new CoreferenceChain(aJCas); corefChain.setFirst(link); corefChain.addToIndexes(); } else { // For the other mentions, we'll add them to the chain. last.setNext(link); } last = link; link.addToIndexes(); } } }
From source file:de.tudarmstadt.ukp.dkpro.core.stanfordnlp.StanfordDependencyConverter.java
License:Open Source License
protected void doCreateDependencyTags(JCas aJCas, TreebankLanguagePack aLP, Tree parseTree, List<Token> tokens) { GrammaticalStructure gs;/*from w w w . j a va 2 s .c om*/ try { gs = aLP.grammaticalStructureFactory(aLP.punctuationWordRejectFilter(), aLP.typedDependencyHeadFinder()) .newGrammaticalStructure(parseTree); } catch (UnsupportedOperationException e) { // We already warned in the model provider if dependencies are not supported, so here // we just do nothing and skip the dependencies. return; } Collection<TypedDependency> dependencies = null; switch (mode) { case BASIC: dependencies = gs.typedDependencies(); // gs.typedDependencies(false); break; case NON_COLLAPSED: dependencies = gs.allTypedDependencies(); // gs.typedDependencies(true); break; case COLLAPSED_WITH_EXTRA: dependencies = gs.typedDependenciesCollapsed(true); break; case COLLAPSED: dependencies = gs.typedDependenciesCollapsed(false); break; case CC_PROPAGATED: dependencies = gs.typedDependenciesCCprocessed(true); break; case CC_PROPAGATED_NO_EXTRA: dependencies = gs.typedDependenciesCCprocessed(false); break; case TREE: dependencies = gs.typedDependenciesCollapsedTree(); break; } for (TypedDependency currTypedDep : dependencies) { int govIndex = currTypedDep.gov().index(); int depIndex = currTypedDep.dep().index(); if (govIndex != 0) { // Stanford CoreNLP produces a dependency relation between a verb and ROOT-0 which // is not token at all! Token govToken = tokens.get(govIndex - 1); Token depToken = tokens.get(depIndex - 1); StanfordAnnotator.createDependencyAnnotation(aJCas, currTypedDep.reln(), govToken, depToken); } } }
From source file:de.tudarmstadt.ukp.dkpro.core.stanfordnlp.StanfordParser.java
License:Open Source License
protected void doCreateDependencyTags(ParserGrammar aParser, StanfordAnnotator sfAnnotator, Tree parseTree, List<Token> tokens) { GrammaticalStructure gs;//from w w w. j a va 2 s . c o m try { TreebankLanguagePack tlp = aParser.getTLPParams().treebankLanguagePack(); gs = tlp.grammaticalStructureFactory(tlp.punctuationWordRejectFilter(), tlp.typedDependencyHeadFinder()) .newGrammaticalStructure(parseTree); } catch (UnsupportedOperationException e) { // We already warned in the model provider if dependencies are not supported, so here // we just do nothing and skip the dependencies. return; } Collection<TypedDependency> dependencies = null; switch (mode) { case BASIC: dependencies = gs.typedDependencies(); // gs.typedDependencies(false); break; case NON_COLLAPSED: dependencies = gs.allTypedDependencies(); // gs.typedDependencies(true); break; case COLLAPSED_WITH_EXTRA: dependencies = gs.typedDependenciesCollapsed(true); break; case COLLAPSED: dependencies = gs.typedDependenciesCollapsed(false); break; case CC_PROPAGATED: dependencies = gs.typedDependenciesCCprocessed(true); break; case CC_PROPAGATED_NO_EXTRA: dependencies = gs.typedDependenciesCCprocessed(false); break; case TREE: dependencies = gs.typedDependenciesCollapsedTree(); break; } for (TypedDependency currTypedDep : dependencies) { int govIndex = currTypedDep.gov().index(); int depIndex = currTypedDep.dep().index(); if (govIndex != 0) { // Stanford CoreNLP produces a dependency relation between a verb and ROOT-0 which // is not token at all! Token govToken = tokens.get(govIndex - 1); Token depToken = tokens.get(depIndex - 1); sfAnnotator.createDependencyAnnotation(currTypedDep.reln(), govToken, depToken); } } }