@Override
public List<Constituent> transform(Constituent input) {
int tokenId = input.getEndSpan();
TextAnnotation ta = input.getTextAnnotation();
Sentence sentence = ta.getSentence(input.getSentenceId());
if (tokenId < sentence.size()) {
Constituent c = new Constituent("", "", ta, tokenId, tokenId + 1);
return Collections.singletonList(addPointerToSource(input, c));
} else return new ArrayList<>();
}
public List<Constituent> getConnectedNPs(Problem prob) {
List<Constituent> npList = new ArrayList<>();
List<Constituent> npListQuantRemoved = new ArrayList<>();
boolean onlyQuantityInSentence = true;
int sentId = prob.ta.getSentenceFromToken(quantPhrase.getStartSpan()).getSentenceId();
for (QuantSpan qs : prob.quantities) {
int tokenId = prob.ta.getTokenIdFromCharacterOffset(qs.start);
if (prob.ta.getSentenceFromToken(tokenId).getSentenceId() == sentId
&& !(quantPhrase.getStartSpan() <= tokenId && quantPhrase.getEndSpan() > tokenId)) {
onlyQuantityInSentence = false;
break;
}
}
// Find NPs from children of verb
if (verbPhrase != null) {
List<Relation> relations = verbPhrase.getOutgoingRelations();
for (Relation relation : relations) {
if (!relation.getRelationName().equals("nsubj")) continue;
Constituent dst = relation.getTarget();
for (Constituent cons : prob.chunks) {
if (cons.getStartSpan() <= dst.getStartSpan()
&& cons.getEndSpan() > dst.getStartSpan()
&& cons.getLabel().equals("NP")
&& !npList.contains(cons)) {
npList.add(cons);
subject = cons;
break;
}
}
}
}
// Find NPs from PP NP connection
int quantPhraseId = getChunkIndex(prob, quantPhrase.getStartSpan());
if (quantPhraseId + 2 < prob.chunks.size()
&& !prob.chunks.get(quantPhraseId + 1).getSurfaceForm().trim().equals("of")
&& prob.chunks.get(quantPhraseId + 1).getLabel().equals("PP")
&& prob.chunks.get(quantPhraseId + 2).getLabel().equals("NP")
&& !npList.contains(prob.chunks.get(quantPhraseId + 2))) {
npList.add(prob.chunks.get(quantPhraseId + 2));
}
if (quantPhraseId - 2 >= 0
&& prob.chunks.get(quantPhraseId - 1).getLabel().equals("PP")
&& prob.chunks.get(quantPhraseId - 2).getLabel().equals("NP")
&& !npList.contains(prob.chunks.get(quantPhraseId - 2))) {
npList.add(prob.chunks.get(quantPhraseId - 2));
}
// Get preceding NP
if (quantPhraseId - 1 >= 0
&& prob.chunks.get(quantPhraseId - 1).getLabel().equals("NP")
&& !prob.posTags
.get(prob.chunks.get(quantPhraseId - 1).getEndSpan())
.getLabel()
.equals("CC")
&& !npList.contains(prob.chunks.get(quantPhraseId - 1))) {
npList.add(prob.chunks.get(quantPhraseId - 1));
}
// Get succeeding NP
if (quantPhraseId + 1 < prob.chunks.size()
&& prob.chunks.get(quantPhraseId + 1).getLabel().equals("NP")
&& !prob.posTags.get(prob.chunks.get(quantPhraseId).getEndSpan()).getLabel().equals("CC")
&& !npList.contains(prob.chunks.get(quantPhraseId + 1))) {
npList.add(prob.chunks.get(quantPhraseId + 1));
}
// If only quantity in sentence, all NPs are connected
if (onlyQuantityInSentence) {
for (int i = 0; i < prob.chunks.size(); ++i) {
Constituent cons = prob.chunks.get(i);
if (cons.getSentenceId() == sentId
&& (i > quantPhraseId + 2 || i < quantPhraseId - 2)
&& !npList.contains(cons)
&& cons.getLabel().equals("NP")) {
npList.add(cons);
}
}
}
// Remove quantity phrases from npList
for (Constituent cons : npList) {
boolean allow = true;
for (QuantSpan qs : prob.quantities) {
int index = prob.ta.getTokenIdFromCharacterOffset(qs.start);
if (index >= cons.getStartSpan() && index < cons.getEndSpan()) {
allow = false;
break;
}
}
if (allow) {
npListQuantRemoved.add(cons);
}
}
return npListQuantRemoved;
}