List of usage examples for edu.stanford.nlp.trees Tree label
@Override
public Label label()
From source file:CollapseUnaryTransformer.java
License:Apache License
public Tree transformTree(Tree tree) { if (tree.isPreTerminal() || tree.isLeaf()) { return tree.deepCopy(); }/*from www.jav a 2s .c o m*/ Label label = tree.label().labelFactory().newLabel(tree.label()); Tree[] children = tree.children(); while (children.length == 1 && !children[0].isLeaf()) { children = children[0].children(); } List<Tree> processedChildren = Generics.newArrayList(); for (Tree child : children) { processedChildren.add(transformTree(child)); } return tree.treeFactory().newTreeNode(label, processedChildren); }
From source file:Treeparse.java
public static List<Tree> GetNounPhrases(Tree parse) { List<Tree> phraseList = new ArrayList<Tree>(); for (Tree subtree : parse) { if (subtree.label().value().equals("NP")) { String str = subtree.toString(); if (str.contains("JJ") || str.contains("JJR") || str.contains("JJS") || str.contains("RB") || str.contains("RBR") || str.contains("RBS")) { phraseList.add(subtree); }/*from w w w . ja v a 2s .c om*/ } } return phraseList; }
From source file:BuildBinarizedDataset.java
/** * Sets all of the labels on a tree to the given default value. *///from www . ja v a2 s. c om public static void setUnknownLabels(Tree tree, Integer defaultLabel) { if (tree.isLeaf()) { return; } for (Tree child : tree.children()) { setUnknownLabels(child, defaultLabel); } tree.label().setValue(defaultLabel.toString()); }
From source file:BuildBinarizedDataset.java
public static void setPredictedLabels(Tree tree) { if (tree.isLeaf()) { return;//from ww w.j a v a2 s . c o m } for (Tree child : tree.children()) { setPredictedLabels(child); } tree.label().setValue(Integer.toString(RNNCoreAnnotations.getPredictedClass(tree))); }
From source file:BuildBinarizedDataset.java
public static boolean setSpanLabel(Tree tree, Pair<Integer, Integer> span, String value) { if (!(tree.label() instanceof CoreLabel)) { throw new AssertionError("Expected CoreLabels"); }/* w w w . j av a 2 s .c o m*/ CoreLabel label = (CoreLabel) tree.label(); if (label.get(CoreAnnotations.BeginIndexAnnotation.class).equals(span.first) && label.get(CoreAnnotations.EndIndexAnnotation.class).equals(span.second)) { label.setValue(value); return true; } if (label.get(CoreAnnotations.BeginIndexAnnotation.class) > span.first && label.get(CoreAnnotations.EndIndexAnnotation.class) < span.second) { return false; } for (Tree child : tree.children()) { if (setSpanLabel(child, span, value)) { return true; } } return false; }
From source file:Anaphora_Resolution.ParseAllXMLDocuments.java
public static Tree HobbsResolve(Tree pronoun, ArrayList<Tree> forest) { Tree wholetree = forest.get(forest.size() - 1); // The last one is the one I am going to start from ArrayList<Tree> candidates = new ArrayList<Tree>(); List<Tree> path = wholetree.pathNodeToNode(wholetree, pronoun); System.out.println(path);//from ww w . j a v a2 s. c om // Step 1 Tree ancestor = pronoun.parent(wholetree); // This one locates the NP the pronoun is in, therefore we need one more "parenting" ! // Step 2 ancestor = ancestor.parent(wholetree); //System.out.println("LABEL: "+pronoun.label().value() + "\n\tVALUE: "+pronoun.firstChild()); while (!ancestor.label().value().equals("NP") && !ancestor.label().value().equals("S")) ancestor = ancestor.parent(wholetree); Tree X = ancestor; path = X.pathNodeToNode(wholetree, pronoun); System.out.println(path); // Step 3 for (Tree relative : X.children()) { for (Tree candidate : relative) { if (candidate.contains(pronoun)) break; // I am looking to all the nodes to the LEFT (i.e. coming before) the path leading to X. contain <-> in the path //System.out.println("LABEL: "+relative.label().value() + "\n\tVALUE: "+relative.firstChild()); if ((candidate.parent(wholetree) != X) && (candidate.parent(wholetree).label().value().equals("NP") || candidate.parent(wholetree).label().value().equals("S"))) if (candidate.label().value().equals("NP")) // "Propose as the antecedent any NP node that is encountered which has an NP or S node between it and X" candidates.add(candidate); } } // Step 9 is a GOTO step 4, hence I will envelope steps 4 to 8 inside a while statement. while (true) { // It is NOT an infinite loop. // Step 4 if (X.parent(wholetree) == wholetree) { for (int q = 1; q < MAXPREVSENTENCES; ++q) {// I am looking for the prev sentence (hence we start with 1) if (forest.size() - 1 < q) break; // If I don't have it, break Tree prevTree = forest.get(forest.size() - 1 - q); // go to previous tree // Now we look for each S subtree, in order of recency (hence right-to-left, hence opposite order of that of .children() ). ArrayList<Tree> backlist = new ArrayList<Tree>(); for (Tree child : prevTree.children()) { for (Tree subtree : child) { if (subtree.label().value().equals("S")) { backlist.add(child); break; } } } for (int i = backlist.size() - 1; i >= 0; --i) { Tree Treetovisit = backlist.get(i); for (Tree relative : Treetovisit.children()) { for (Tree candidate : relative) { if (candidate.contains(pronoun)) continue; // I am looking to all the nodes to the LEFT (i.e. coming before) the path leading to X. contain <-> in the path //System.out.println("LABEL: "+relative.label().value() + "\n\tVALUE: "+relative.firstChild()); if (candidate.label().value().equals("NP")) { // "Propose as the antecedent any NP node that you find" if (!candidates.contains(candidate)) candidates.add(candidate); } } } } } break; // It will always come here eventually } // Step 5 ancestor = X.parent(wholetree); //System.out.println("LABEL: "+pronoun.label().value() + "\n\tVALUE: "+pronoun.firstChild()); while (!ancestor.label().value().equals("NP") && !ancestor.label().value().equals("S")) ancestor = ancestor.parent(wholetree); X = ancestor; // Step 6 if (X.label().value().equals("NP")) { // If X is an NP for (Tree child : X.children()) { // Find the nominal nodes that X directly dominates if (child.label().value().equals("NN") || child.label().value().equals("NNS") || child.label().value().equals("NNP") || child.label().value().equals("NNPS")) if (!child.contains(pronoun)) candidates.add(X); // If one of them is not in the path between X and the pronoun, add X to the antecedents } } // Step SETTE for (Tree relative : X.children()) { for (Tree candidate : relative) { if (candidate.contains(pronoun)) continue; // I am looking to all the nodes to the LEFT (i.e. coming before) the path leading to X. contain <-> in the path //System.out.println("LABEL: "+relative.label().value() + "\n\tVALUE: "+relative.firstChild()); if (candidate.label().value().equals("NP")) { // "Propose as the antecedent any NP node that you find" boolean contains = false; for (Tree oldercandidate : candidates) { if (oldercandidate.contains(candidate)) { contains = true; break; } } if (!contains) candidates.add(candidate); } } } // Step 8 if (X.label().value().equals("S")) { boolean right = false; // Now we want all branches to the RIGHT of the path pronoun -> X. for (Tree relative : X.children()) { if (relative.contains(pronoun)) { right = true; continue; } if (!right) continue; for (Tree child : relative) { // Go in but do not go below any NP or S node. Go below the rest if (child.label().value().equals("NP")) { candidates.add(child); break; // not sure if this means avoid going below NP but continuing with the rest of non-NP children. Should be since its DFS. } if (child.label().value().equals("S")) break; // Same } } } } // Step 9 is a GOTO, so we use a while. System.out.println(pronoun + ": CHAIN IS " + candidates.toString()); ArrayList<Integer> scores = new ArrayList<Integer>(); for (int j = 0; j < candidates.size(); ++j) { Tree candidate = candidates.get(j); Tree parent = null; int parent_index = 0; for (Tree tree : forest) { if (tree.contains(candidate)) { parent = tree; break; } ++parent_index; } scores.add(0); if (parent_index == 0) scores.set(j, scores.get(j) + 100); // If in the last sentence, +100 points scores.set(j, scores.get(j) + syntacticScore(candidate, parent)); if (existentialEmphasis(candidate)) // Example: "There was a dog standing outside" scores.set(j, scores.get(j) + 70); if (!adverbialEmphasis(candidate, parent)) scores.set(j, scores.get(j) + 50); if (headNounEmphasis(candidate, parent)) scores.set(j, scores.get(j) + 80); int sz = forest.size() - 1; // System.out.println("pronoun in sentence " + sz + "(sz). Candidate in sentence "+parent_index+" (parent_index)"); int dividend = 1; for (int u = 0; u < sz - parent_index; ++u) dividend *= 2; //System.out.println("\t"+dividend); scores.set(j, scores.get(j) / dividend); System.out.println(candidate + " -> " + scores.get(j)); } int max = -1; int max_index = -1; for (int i = 0; i < scores.size(); ++i) { if (scores.get(i) > max) { max_index = i; max = scores.get(i); } } Tree final_candidate = candidates.get(max_index); System.out.println("My decision for " + pronoun + " is: " + final_candidate); // Decide what candidate, with both gender resolution and Lappin and Leass ranking. Tree pronounparent = pronoun.parent(wholetree).parent(wholetree); // 1 parent gives me the NP of the pronoun int pos = 0; for (Tree sibling : pronounparent.children()) { System.out.println("Sibling " + pos + ": " + sibling); if (sibling.contains(pronoun)) break; ++pos; } System.out.println("Before setchild: " + pronounparent); @SuppressWarnings("unused") Tree returnval = pronounparent.setChild(pos, final_candidate); System.out.println("After setchild: " + pronounparent); return wholetree; // wholetree is already modified, since it contains pronounparent }
From source file:Anaphora_Resolution.ParseAllXMLDocuments.java
private static int syntacticScore(Tree candidate, Tree root) { // We will check whether the NP is inside an S (hence it would be a subject) // a VP (direct object) // a PP inside a VP (an indirect obj) Tree parent = candidate; while (!parent.label().value().equals("S")) { if (parent.label().value().equals("VP")) return 50; // direct obj if (parent.label().value().equals("PP")) { Tree grandparent = parent.parent(root); while (!grandparent.label().value().equals("S")) { if (parent.label().value().equals("VP")) // indirect obj is a PP inside a VP return 40; parent = grandparent;/*from w w w .j a v a2 s . c om*/ grandparent = grandparent.parent(root); } } parent = parent.parent(root); } return 80; // If nothing remains, it must be the subject }
From source file:Anaphora_Resolution.ParseAllXMLDocuments.java
private static boolean existentialEmphasis(Tree candidate) { // We want to check whether our NP's Dets are "a" or "an". for (Tree child : candidate) { if (child.label().value().equals("DT")) { for (Tree leaf : child) { if (leaf.value().equals("a") || leaf.value().equals("an") || leaf.value().equals("A") || leaf.value().equals("An")) { //System.out.println("Existential emphasis!"); return true; }/*from ww w.j a v a2 s. c o m*/ } } } return false; }
From source file:Anaphora_Resolution.ParseAllXMLDocuments.java
private static boolean headNounEmphasis(Tree candidate, Tree root) { Tree parent = candidate.parent(root); while (!parent.label().value().equals("S")) { // If it is the head NP, it is not contained in another NP (that's exactly how the original algorithm does it) if (parent.label().value().equals("NP")) return false; parent = parent.parent(root);//from w w w . j a v a2 s . com } return true; }
From source file:Anaphora_Resolution.ParseAllXMLDocuments.java
private static boolean adverbialEmphasis(Tree candidate, Tree root) { // Like in "Inside the castle, King Arthur was invincible". "Castle" has the adv emph. Tree parent = candidate; while (!parent.label().value().equals("S")) { if (parent.label().value().equals("PP")) { for (Tree sibling : parent.siblings(root)) { if ((sibling.label().value().equals(","))) { //System.out.println("adv Emph!"); return true; }// w ww . j ava 2 s . c o m } } parent = parent.parent(root); } return false; }