private IntPair GetNPSpanOld(IndexedWord headword, SemanticGraph dep, IList <CoreLabel> sent)
{
IndexedWord cop = dep.GetChildWithReln(headword, UniversalEnglishGrammaticalRelations.Copula);
Pair <IndexedWord, IndexedWord> leftRight = SemanticGraphUtils.LeftRightMostChildVertices(headword, dep);
// headword can be first or last word
int beginIdx = Math.Min(headword.Index() - 1, leftRight.first.Index() - 1);
int endIdx = Math.Max(headword.Index() - 1, leftRight.second.Index() - 1);
// no copula relation
if (cop == null)
{
return(new IntPair(beginIdx, endIdx));
}
// if we have copula relation
IList <IndexedWord> children = dep.GetChildList(headword);
int copIdx = children.IndexOf(cop);
if (copIdx + 1 < children.Count)
{
beginIdx = Math.Min(headword.Index() - 1, SemanticGraphUtils.LeftMostChildVertice(children[copIdx + 1], dep).Index() - 1);
}
else
{
beginIdx = headword.Index() - 1;
}
return(new IntPair(beginIdx, endIdx));
}
/// <summary>Determine the person of "was".</summary>
private static string WasPerson(SemanticGraph sg, IndexedWord word)
{
IndexedWord subj = sg.GetChildWithReln(word, UniversalEnglishGrammaticalRelations.NominalSubject);
if (subj == null)
{
subj = sg.GetChildWithReln(word, UniversalEnglishGrammaticalRelations.NominalPassiveSubject);
}
if (subj != null)
{
if (Sharpen.Runtime.EqualsIgnoreCase(subj.Word(), "i"))
{
/* "I" is the subject of "was". */
return("1");
}
}
IndexedWord parent = sg.GetParent(word);
if (parent == null)
{
return(subj != null ? "3" : null);
}
SemanticGraphEdge edge = sg.GetEdge(parent, word);
if (edge == null)
{
return(subj != null ? "3" : null);
}
if (UniversalEnglishGrammaticalRelations.AuxModifier.Equals(edge.GetRelation()) || UniversalEnglishGrammaticalRelations.AuxPassiveModifier.Equals(edge.GetRelation()))
{
return(WasPerson(sg, parent));
}
if (UniversalEnglishGrammaticalRelations.Conjunct.IsAncestor(edge.GetRelation()))
{
/* Check if the subject of the head of a conjunction is "I". */
return(WasPerson(sg, parent));
}
return("3");
}
/// <summary>
/// return the left and right most node except copula relation (nsubj & cop) and some others (maybe discourse?)
/// e.g., you are the person -> return "the person"
/// </summary>
private IntPair GetNPSpan(IndexedWord headword, SemanticGraph dep, IList <CoreLabel> sent)
{
int headwordIdx = headword.Index() - 1;
IList <IndexedWord> children = dep.GetChildList(headword);
// if(children.size()==0) return new IntPair(headwordIdx, headwordIdx); // the headword is the only word
// check if we have copula relation
IndexedWord cop = dep.GetChildWithReln(headword, UniversalEnglishGrammaticalRelations.Copula);
int startIdx = (cop == null) ? 0 : children.IndexOf(cop) + 1;
// children which will be inside of NP
IList <IndexedWord> insideNP = Generics.NewArrayList();
for (int i = startIdx; i < children.Count; i++)
{
IndexedWord child = children[i];
SemanticGraphEdge edge = dep.GetEdge(headword, child);
if (edge.GetRelation().GetShortName().Matches("dep|discourse|punct"))
{
continue;
}
else
{
// skip
insideNP.Add(child);
}
}
if (insideNP.Count == 0)
{
return(new IntPair(headwordIdx, headwordIdx));
}
// the headword is the only word
Pair <IndexedWord, IndexedWord> firstChildLeftRight = SemanticGraphUtils.LeftRightMostChildVertices(insideNP[0], dep);
Pair <IndexedWord, IndexedWord> lastChildLeftRight = SemanticGraphUtils.LeftRightMostChildVertices(insideNP[insideNP.Count - 1], dep);
// headword can be first or last word
int beginIdx = Math.Min(headwordIdx, firstChildLeftRight.first.Index() - 1);
int endIdx = Math.Max(headwordIdx, lastChildLeftRight.second.Index() - 1);
return(new IntPair(beginIdx, endIdx));
}
private void ExtractMentionForHeadword(IndexedWord headword, SemanticGraph dep, ICoreMap s, IList <Mention> mentions, ICollection <IntPair> mentionSpanSet, ICollection <IntPair> namedEntitySpanSet)
{
IList <CoreLabel> sent = s.Get(typeof(CoreAnnotations.TokensAnnotation));
SemanticGraph basic = s.Get(typeof(SemanticGraphCoreAnnotations.BasicDependenciesAnnotation));
SemanticGraph enhanced = s.Get(typeof(SemanticGraphCoreAnnotations.EnhancedDependenciesAnnotation));
if (enhanced == null)
{
enhanced = s.Get(typeof(SemanticGraphCoreAnnotations.BasicDependenciesAnnotation));
}
// pronoun
if (headword.Tag().StartsWith("PRP"))
{
ExtractPronounForHeadword(headword, dep, s, mentions, mentionSpanSet, namedEntitySpanSet);
return;
}
// add NP mention
IntPair npSpan = GetNPSpan(headword, dep, sent);
int beginIdx = npSpan.Get(0);
int endIdx = npSpan.Get(1) + 1;
if (",".Equals(sent[endIdx - 1].Word()))
{
endIdx--;
}
// try not to have span that ends with ,
if ("IN".Equals(sent[beginIdx].Tag()))
{
beginIdx++;
}
// try to remove first IN.
AddMention(beginIdx, endIdx, headword, mentions, mentionSpanSet, namedEntitySpanSet, sent, basic, enhanced);
//
// extract the first element in conjunction (A and B -> extract A here "A and B", "B" will be extracted above)
//
// to make sure we find the first conjunction
ICollection <IndexedWord> conjChildren = dep.GetChildrenWithReln(headword, UniversalEnglishGrammaticalRelations.Conjunct);
if (conjChildren.Count > 0)
{
IndexedWord conjChild = dep.GetChildWithReln(headword, UniversalEnglishGrammaticalRelations.Conjunct);
foreach (IndexedWord c in conjChildren)
{
if (c.Index() < conjChild.Index())
{
conjChild = c;
}
}
IndexedWord left = SemanticGraphUtils.LeftMostChildVertice(conjChild, dep);
for (int endIdxFirstElement = left.Index() - 1; endIdxFirstElement > beginIdx; endIdxFirstElement--)
{
if (!sent[endIdxFirstElement - 1].Tag().Matches("CC|,"))
{
if (headword.Index() - 1 < endIdxFirstElement)
{
AddMention(beginIdx, endIdxFirstElement, headword, mentions, mentionSpanSet, namedEntitySpanSet, sent, basic, enhanced);
}
break;
}
}
}
}