List<Util.Pair<Integer, Integer>> getCandidates(
Document document, int arg2Line, int connStart, int connEnd, int arg1Line) {
List<Util.Pair<Integer, Integer>> candidates = new ArrayList<Util.Pair<Integer, Integer>>();
int distance = 10;
Sentence arg2Sentence = document.getSentence(arg2Line);
// String conn = arg2Sentence.toString(connStart, connEnd).toLowerCase();
// String category = connAnalyzer.getCategory(conn);
int connHeadPos = connAnalyzer.getHeadWord(arg2Sentence.getParseTree(), connStart, connEnd);
SimpleDepGraph arg2DepGraph = document.getDepGraph(arg2Line);
List<Integer> reachable = arg2DepGraph.getReachableIndices(connHeadPos, false, distance);
for (Integer i : reachable) {
if (arg2Sentence.get(i).getTag("POS").matches("VB.*|NNS?|JJ.*|MD")) {
candidates.add(new Util.Pair<Integer, Integer>(arg2Line, i));
}
}
Tree mainHead = headAnalyzer.getCollinsHead(arg2Sentence.getParseTree().getChild(0));
if (mainHead != null) {
int mainHeadPos = treeAnalyzer.getLeafPosition(arg2Sentence.getParseTree(), mainHead);
List<String> pathAsList = arg2DepGraph.getPathAsList(connHeadPos, mainHeadPos, false);
if (pathAsList != null) {
distance = distance - (1 + pathAsList.size());
} else {
// System.out.println("No path from connHead to mainHead!");
distance--;
}
}
// if (arg1Line == arg2Line) return candidates;
for (int i = arg2Line - 1; i >= 0 && distance >= 0; i--) {
Sentence sentence = document.getSentence(i);
SimpleDepGraph depGraph = document.getDepGraph(i);
mainHead = headAnalyzer.getCollinsHead(sentence.getParseTree().getChild(0));
if (mainHead == null) {
distance--;
continue;
}
int mainHeadPos = treeAnalyzer.getLeafPosition(sentence.getParseTree(), mainHead);
reachable = depGraph.getReachableIndices(mainHeadPos, false, distance);
if (reachable == null) {
distance--;
continue;
}
for (Integer j : reachable) {
if (sentence.get(j).getTag("POS").matches("VB.*|NNS?|JJ.*|MD")) {
candidates.add(new Util.Pair<Integer, Integer>(i, j));
}
}
distance -= 2;
}
return candidates;
}
private PropertyList addLexicoSyntacticFeatures(
PropertyList pl,
Document doc,
Pair<Integer, Integer> candidate,
int arg2Line,
int arg2HeadPos,
int connStart,
int connEnd) {
int arg1Line = candidate.first();
Tree root = doc.getTree(arg1Line);
int arg1HeadPos = candidate.second();
boolean attributive = false;
String head = root.getLeaves().get(arg1HeadPos).value();
for (String verb : attributiveVerb) {
if (head.matches(verb)) {
attributive = true;
break;
}
}
pl = PropertyList.add("U=" + attributive, 1.0, pl);
SimpleDepGraph depGraph = doc.getDepGraph(arg1Line);
boolean hasClausalComp = false;
List<SimpleDependency> govDependencies = depGraph.getGovDependencies(arg1HeadPos);
for (SimpleDependency dep : govDependencies) {
if (dep.reln().equals("ccomp")) {
hasClausalComp = true;
break;
}
}
pl = PropertyList.add("V=" + hasClausalComp, 1.0, pl);
pl = PropertyList.add("W=" + attributive + "&" + hasClausalComp, 1.0, pl);
boolean isClausalComp = false;
List<SimpleDependency> depDependencies = depGraph.getDepDependencies(arg1HeadPos);
SimpleDependency clausalComp = null;
for (SimpleDependency dep : depDependencies) {
if (dep.reln().equals("ccomp")) {
isClausalComp = true;
clausalComp = dep;
break;
}
}
pl = PropertyList.add("X=" + isClausalComp, 1.0, pl);
if (isClausalComp) {
int gov = clausalComp.gov();
String govWord = root.getLeaves().get(gov).value();
boolean isGovAttributive = false;
for (String verb : attributiveVerb) {
if (govWord.matches(verb)) {
isGovAttributive = true;
break;
}
}
pl = PropertyList.add("Y=" + isClausalComp + "&" + isGovAttributive, 1.0, pl);
}
return pl;
}
private PropertyList addDependencyFeatures(
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);
String category = connAnalyzer.getCategory(conn.toLowerCase());
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>();
if (arg1Line == arg2Line) {
SimpleDepGraph depGraph = doc.getDepGraph(arg1Line);
List<String> tmpPath = depGraph.getPathAsList(connHeadPos, arg1HeadPos, false);
if (tmpPath != null) {
path.addAll(tmpPath);
} else {
path.add("null");
}
} else {
Tree arg2Root = arg2Sentence.getParseTree();
Tree mainHead = headAnalyzer.getCollinsHead(arg2Root.getChild(0));
int mainHeadPos = treeAnalyzer.getLeafPosition(arg2Root, mainHead);
if (mainHeadPos != -1) {
SimpleDepGraph depGraph = doc.getDepGraph(arg2Line);
List<String> tmpPath = depGraph.getPathAsList(connHeadPos, mainHeadPos, false);
if (tmpPath != null) {
path.addAll(tmpPath);
} else {
path.add("null");
}
}
for (int i = 0; i < Math.abs(arg1Line - arg2Line); i++) {
path.add("SENT");
}
Tree arg1Root = arg1Tree;
mainHead = headAnalyzer.getCollinsHead(arg1Root.getChild(0));
mainHeadPos = treeAnalyzer.getLeafPosition(arg1Root, mainHead);
if (mainHeadPos != -1) {
SimpleDepGraph depGraph = doc.getDepGraph(arg1Line);
List<String> tmpPath = depGraph.getPathAsList(mainHeadPos, arg1HeadPos, false);
if (tmpPath != null) {
path.addAll(tmpPath);
} else {
path.add("null");
}
}
}
StringBuilder sbPath = new StringBuilder();
StringBuilder sbPathWithoutCC = new StringBuilder();
StringBuilder sbPathWithoutRep = new StringBuilder();
String prev = "";
for (String node : path) {
sbPath.append(node).append(":");
if (!node.matches("cc|-cc")) {
sbPathWithoutCC.append(node).append(":");
}
if (!node.equals(prev)) {
sbPathWithoutRep.append(node).append(":");
}
prev = node;
}
// M-dependency path
pl = PropertyList.add("M=" + sbPath.toString(), 1.0, pl);
// Q-M&C
pl = PropertyList.add("Q=" + "CONN-" + conn + '&' + "M-" + sbPath.toString(), 1.0, pl);
// T-M&R
pl = PropertyList.add("T=" + "CAT-" + category + '&' + "M-" + sbPath.toString(), 1.0, pl);
// O-collapsed path without cc
pl = PropertyList.add("O=" + sbPathWithoutCC.toString(), 1.0, pl);
// P-collapsed path without repetition
pl = PropertyList.add("P=" + sbPathWithoutRep.toString(), 1.0, pl);
return pl;
}
// added 11-03-2012
int getMainHead(Sentence s, SimpleDepGraph depGraph) {
for (int i = 0; i < s.size(); i++) {
if (depGraph.getParent(i) == -1) return i;
}
return -1;
}