private PropertyList addConstituentFeatures(
PropertyList pl,
Document doc,
Pair<Integer, Integer> candidate,
int arg2Line,
int arg2HeadPos,
int connStart,
int connEnd) {
Sentence arg2Sentence = doc.getSentence(arg2Line);
String conn = arg2Sentence.toString(connStart, connEnd);
int connHeadPos = connAnalyzer.getHeadWord(arg2Sentence.getParseTree(), connStart, connEnd);
int arg1Line = candidate.first();
Tree arg1Tree = doc.getTree(arg1Line);
int arg1HeadPos = candidate.second();
List<String> path = new ArrayList<String>();
List<String> pathWithoutPOS = new ArrayList<String>();
if (arg1Line == arg2Line) {
Tree root = arg1Tree;
List<Tree> leaves = root.getLeaves();
List<Tree> treePath = root.pathNodeToNode(leaves.get(connHeadPos), leaves.get(arg1HeadPos));
if (treePath != null) {
for (Tree t : treePath) {
if (!t.isLeaf()) {
path.add(t.value());
if (!t.isPreTerminal()) {
pathWithoutPOS.add(t.value());
}
}
}
}
} else {
Tree arg2Root = arg2Sentence.getParseTree();
Tree mainHead = headAnalyzer.getCollinsHead(arg2Root.getChild(0));
List<Tree> leaves = arg2Root.getLeaves();
int mainHeadPos = treeAnalyzer.getLeafPosition(arg2Root, mainHead);
if (mainHeadPos != -1) {
List<Tree> treePath =
arg2Root.pathNodeToNode(leaves.get(connHeadPos), leaves.get(mainHeadPos));
if (treePath != null) {
for (Tree t : treePath) {
if (!t.isLeaf()) {
path.add(t.value());
if (!t.isPreTerminal()) {
pathWithoutPOS.add(t.value());
}
}
}
}
}
for (int i = 0; i < Math.abs(arg1Line - arg2Line); i++) {
path.add("SENT");
pathWithoutPOS.add("SENT");
}
Tree arg1Root = arg1Tree;
mainHead = headAnalyzer.getCollinsHead(arg1Root.getChild(0));
leaves = arg1Root.getLeaves();
mainHeadPos = treeAnalyzer.getLeafPosition(arg1Root, mainHead);
if (mainHeadPos != -1) {
List<Tree> treePath =
arg1Root.pathNodeToNode(leaves.get(mainHeadPos), leaves.get(arg1HeadPos));
if (treePath != null) {
for (Tree t : treePath) {
if (!t.isLeaf()) {
path.add(t.value());
if (!t.isPreTerminal()) {
pathWithoutPOS.add(t.value());
}
}
}
}
}
}
// H-full path
// L-C&H
StringBuilder fullPath = new StringBuilder();
for (String node : path) {
fullPath.append(node).append(":");
}
pl = PropertyList.add("H=" + fullPath.toString(), 1.0, pl);
pl = PropertyList.add("L=CONN-" + conn + "&" + "H-" + fullPath.toString(), 1.0, pl);
// I-length of path
pl = PropertyList.add("I=" + path.size(), 1.0, pl);
// J-collapsed path without part of speech
// K-collapsed path without repititions
fullPath = new StringBuilder();
StringBuilder collapsedPath = new StringBuilder();
String prev = "";
for (String node : pathWithoutPOS) {
fullPath.append(node).append(":");
if (!node.equals(prev)) {
collapsedPath.append(node).append(":");
}
prev = node;
}
pl = PropertyList.add("J=" + fullPath.toString(), 1.0, pl);
pl = PropertyList.add("K=" + collapsedPath.toString(), 1.0, pl);
return pl;
}
