/// <summary>Determine the case of the pronoun "you" or "it".</summary>
private static string PronounCase(SemanticGraph sg, IndexedWord word)
{
word = sg.GetNodeByIndex(word.Index());
IndexedWord parent = sg.GetParent(word);
if (parent != null)
{
SemanticGraphEdge edge = sg.GetEdge(parent, word);
if (edge != null)
{
if (UniversalEnglishGrammaticalRelations.Object.IsAncestor(edge.GetRelation()))
{
/* "you" is an object. */
return("Acc");
}
else
{
if (UniversalEnglishGrammaticalRelations.NominalModifier.IsAncestor(edge.GetRelation()) || edge.GetRelation() == GrammaticalRelation.Root)
{
if (sg.HasChildWithReln(word, UniversalEnglishGrammaticalRelations.CaseMarker))
{
/* "you" is the head of a prepositional phrase. */
return("Acc");
}
}
}
}
}
return("Nom");
}
/// <summary>Extracts features from relative and interrogative pronouns.</summary>
private static Dictionary <string, string> GetRelAndIntPronFeatures(SemanticGraph sg, IndexedWord word)
{
Dictionary <string, string> features = new Dictionary <string, string>();
if (word.Tag().StartsWith("W"))
{
bool isRel = false;
IndexedWord parent = sg.GetParent(word);
if (parent != null)
{
IndexedWord parentParent = sg.GetParent(parent);
if (parentParent != null)
{
SemanticGraphEdge edge = sg.GetEdge(parentParent, parent);
isRel = edge.GetRelation().Equals(UniversalEnglishGrammaticalRelations.RelativeClauseModifier);
}
}
if (isRel)
{
features["PronType"] = "Rel";
}
else
{
if (Sharpen.Runtime.EqualsIgnoreCase(word.Value(), "that"))
{
features["PronType"] = "Dem";
}
else
{
features["PronType"] = "Int";
}
}
}
return(features);
}
private static Dictionary <string, string> GetGraphFeatures(SemanticGraph sg, IndexedWord word)
{
Dictionary <string, string> features = new Dictionary <string, string>();
/* Determine the case of "you". */
if (word.Tag().Equals("PRP") && (Sharpen.Runtime.EqualsIgnoreCase(word.Value(), "you") || Sharpen.Runtime.EqualsIgnoreCase(word.Value(), "it")))
{
features["Case"] = PronounCase(sg, word);
}
/* Determine the person of "was". */
if (word.Tag().Equals("VBD") && Sharpen.Runtime.EqualsIgnoreCase(word.Value(), "was"))
{
string person = WasPerson(sg, word);
if (person != null)
{
features["Person"] = person;
}
}
/* Determine features of relative and interrogative pronouns. */
features.PutAll(GetRelAndIntPronFeatures(sg, word));
/* Determine features of gerunds and present participles. */
if (word.Tag().Equals("VBG"))
{
if (HasBeAux(sg, word))
{
features["VerbForm"] = "Part";
features["Tense"] = "Pres";
}
else
{
features["VerbForm"] = "Ger";
}
}
/* Determine whether reflexive pronoun is reflexive or intensive. */
if (word.Value().Matches(SelfRegex) && word.Tag().Equals("PRP"))
{
IndexedWord parent = sg.GetParent(word);
if (parent != null)
{
SemanticGraphEdge edge = sg.GetEdge(parent, word);
if (edge.GetRelation() != UniversalEnglishGrammaticalRelations.NpAdverbialModifier)
{
features["Case"] = "Acc";
features["Reflex"] = "Yes";
}
}
}
/* Voice feature. */
if (word.Tag().Equals("VBN"))
{
if (sg.HasChildWithReln(word, UniversalEnglishGrammaticalRelations.AuxPassiveModifier))
{
features["Voice"] = "Pass";
}
}
return(features);
}
public override void Evaluate(SemanticGraph sg, SemgrexMatcher sm)
{
string relation = sm.GetRelnString(edgeName);
IndexedWord govNode = GetNamedNode(govName, sm);
IndexedWord depNode = GetNamedNode(depName, sm);
SemanticGraphEdge edge = sg.GetEdge(govNode, depNode, GrammaticalRelation.ValueOf(relation));
if (edge != null)
{
sg.RemoveEdge(edge);
}
}
private void ExtractPronounForHeadword(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));
}
int beginIdx = headword.Index() - 1;
int endIdx = headword.Index();
// handle "you all", "they both" etc
if (sent.Count > headword.Index() && sent[headword.Index()].Word().Matches("all|both"))
{
IndexedWord c = dep.GetNodeByIndex(headword.Index() + 1);
SemanticGraphEdge edge = dep.GetEdge(headword, c);
if (edge != null)
{
endIdx++;
}
}
IntPair mSpan = new IntPair(beginIdx, endIdx);
if (!mentionSpanSet.Contains(mSpan) && (!InsideNE(mSpan, namedEntitySpanSet)))
{
int dummyMentionId = -1;
Mention m = new Mention(dummyMentionId, beginIdx, endIdx, sent, basic, enhanced, new List <CoreLabel>(sent.SubList(beginIdx, endIdx)));
m.headIndex = headword.Index() - 1;
m.headWord = sent[m.headIndex];
m.headString = m.headWord.Word().ToLower(Locale.English);
mentions.Add(m);
mentionSpanSet.Add(mSpan);
}
// when pronoun is a part of conjunction (e.g., you and I)
ICollection <IndexedWord> conjChildren = dep.GetChildrenWithReln(headword, UniversalEnglishGrammaticalRelations.Conjunct);
if (conjChildren.Count > 0)
{
IntPair npSpan = GetNPSpan(headword, dep, sent);
beginIdx = npSpan.Get(0);
endIdx = npSpan.Get(1) + 1;
if (",".Equals(sent[endIdx - 1].Word()))
{
endIdx--;
}
// try not to have span that ends with ,
AddMention(beginIdx, endIdx, headword, mentions, mentionSpanSet, namedEntitySpanSet, sent, basic, enhanced);
}
}
/// <summary>
/// Returns true if
/// <paramref name="word"/>
/// has an auxiliary verb attached to it.
/// </summary>
private static bool HasAux(SemanticGraph sg, IndexedWord word)
{
if (sg.HasChildWithReln(word, UniversalEnglishGrammaticalRelations.AuxModifier))
{
return(true);
}
IndexedWord gov = sg.GetParent(word);
if (gov != null)
{
SemanticGraphEdge edge = sg.GetEdge(gov, word);
if (UniversalEnglishGrammaticalRelations.Conjunct.IsAncestor(edge.GetRelation()) || UniversalEnglishGrammaticalRelations.Copula.Equals(edge.GetRelation()))
{
return(HasAux(sg, gov));
}
}
return(false);
}
public override void Evaluate(SemanticGraph sg, SemgrexMatcher sm)
{
bool govWild = govName.Equals(WildcardNode);
bool depWild = depName.Equals(WildcardNode);
IndexedWord govNode = GetNamedNode(govName, sm);
IndexedWord depNode = GetNamedNode(depName, sm);
if (govNode != null && depNode != null)
{
SemanticGraphEdge edge = sg.GetEdge(govNode, depNode, relation);
if (edge != null)
{
bool successFlag = sg.RemoveEdge(edge);
}
}
else
{
if (depNode != null && govWild)
{
// dep known, wildcard gov
foreach (SemanticGraphEdge edge in sg.IncomingEdgeIterable(depNode))
{
if (edge.GetRelation().Equals(relation) && sg.ContainsEdge(edge))
{
sg.RemoveEdge(edge);
}
}
}
else
{
if (govNode != null && depWild)
{
// gov known, wildcard dep
foreach (SemanticGraphEdge edge in sg.OutgoingEdgeIterable(govNode))
{
if (edge.GetRelation().Equals(relation) && sg.ContainsEdge(edge))
{
sg.RemoveEdge(edge);
}
}
}
}
}
}
private void ExtractNPorPRPFromDependency(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));
IList <IndexedWord> nounsOrPrp = basic.GetAllNodesByPartOfSpeechPattern("N.*|PRP.*|DT");
// DT is for "this, these, etc"
Tree tree = s.Get(typeof(TreeCoreAnnotations.TreeAnnotation));
foreach (IndexedWord w in nounsOrPrp)
{
SemanticGraphEdge edge = basic.GetEdge(basic.GetParent(w), w);
GrammaticalRelation rel = null;
string shortname = "root";
// if edge is null, it's root
if (edge != null)
{
rel = edge.GetRelation();
shortname = rel.GetShortName();
}
// TODO: what to remove? remove more?
if (shortname.Matches("det|compound"))
{
// // for debug ---------------
// Tree t = tree.getLeaves().get(w.index()-1);
// for(Tree p : tree.pathNodeToNode(t, tree)) {
// if(p.label().value().equals("NP")) {
// HeadFinder headFinder = new SemanticHeadFinder();
// Tree head = headFinder.determineHead(p);
// if(head == t.parent(tree)) {
// log.info();
// }
// break;
// }
// } // for debug -------------
continue;
}
else
{
ExtractMentionForHeadword(w, basic, s, mentions, mentionSpanSet, namedEntitySpanSet);
}
}
}
public override void Evaluate(SemanticGraph sg, SemgrexMatcher sm)
{
IndexedWord govNode = GetNamedNode(govName, sm);
IndexedWord depNode = GetNamedNode(depName, sm);
SemanticGraphEdge existingEdge = sg.GetEdge(govNode, depNode, relation);
if (existingEdge == null)
{
// When adding the edge, check to see if the gov/dep nodes are presently in the graph.
//
if (!sg.ContainsVertex(govNode))
{
sg.AddVertex(govNode);
}
if (!sg.ContainsVertex(depNode))
{
sg.AddVertex(depNode);
}
sg.AddEdge(govNode, depNode, relation, weight, false);
}
}
/// <summary>
/// Returns true if
/// <paramref name="word"/>
/// has an inflection of "be" as an auxiliary.
/// </summary>
private static bool HasBeAux(SemanticGraph sg, IndexedWord word)
{
foreach (IndexedWord aux in sg.GetChildrenWithReln(word, UniversalEnglishGrammaticalRelations.AuxModifier))
{
if (aux.Value().Matches(BeRegex))
{
return(true);
}
}
/* Check if head of conjunction has an auxiliary in case the word is part of a conjunction */
IndexedWord gov = sg.GetParent(word);
if (gov != null)
{
SemanticGraphEdge edge = sg.GetEdge(gov, word);
if (UniversalEnglishGrammaticalRelations.Conjunct.IsAncestor(edge.GetRelation()))
{
return(HasBeAux(sg, gov));
}
}
return(false);
}
/// <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));
}
//using quote-removed depparses
public virtual void DependencyParses(Annotation doc)
{
IList <ICoreMap> quotes = doc.Get(typeof(CoreAnnotations.QuotationsAnnotation));
IList <CoreLabel> tokens = doc.Get(typeof(CoreAnnotations.TokensAnnotation));
IList <ICoreMap> sentences = doc.Get(typeof(CoreAnnotations.SentencesAnnotation));
foreach (ICoreMap quote in quotes)
{
if (quote.Get(typeof(QuoteAttributionAnnotator.MentionAnnotation)) != null)
{
continue;
}
Pair <int, int> range = QuoteAttributionUtils.GetRemainderInSentence(doc, quote);
if (range == null)
{
continue;
}
//search for mentions in the first run
Pair <List <string>, List <Pair <int, int> > > namesAndNameIndices = ScanForNames(range);
List <string> names = namesAndNameIndices.first;
List <Pair <int, int> > nameIndices = namesAndNameIndices.second;
SemanticGraph graph = quote.Get(typeof(SemanticGraphCoreAnnotations.EnhancedPlusPlusDependenciesAnnotation));
SemgrexMatcher matcher = subjVerbPattern.Matcher(graph);
IList <Pair <IndexedWord, IndexedWord> > subjVerbPairs = new List <Pair <IndexedWord, IndexedWord> >();
//TODO: check and see if this is necessary
while (matcher.Find())
{
IndexedWord subj = matcher.GetNode("SUBJ");
IndexedWord verb = matcher.GetNode("VERB");
subjVerbPairs.Add(new Pair <IndexedWord, IndexedWord>(subj, verb));
}
IList <IndexedWord> vbs = graph.GetAllNodesByPartOfSpeechPattern("VB.*");
foreach (IndexedWord iw in vbs)
{
// does it have an nsubj child?
ICollection <IndexedWord> children = graph.GetChildren(iw);
IList <IndexedWord> deps = Generics.NewArrayList();
IndexedWord nsubj = null;
foreach (IndexedWord child in children)
{
SemanticGraphEdge sge = graph.GetEdge(iw, child);
if (sge.GetRelation().GetShortName().Equals("dep") && child.Tag().StartsWith("VB"))
{
deps.Add(child);
}
else
{
if (sge.GetRelation().GetShortName().Equals("nsubj"))
{
nsubj = child;
}
}
}
if (nsubj != null)
{
foreach (IndexedWord dep in deps)
{
subjVerbPairs.Add(new Pair(nsubj, dep));
}
}
}
//look for a speech verb
foreach (Pair <IndexedWord, IndexedWord> SVPair in subjVerbPairs)
{
IndexedWord verb = SVPair.second;
IndexedWord subj = SVPair.first;
//check if subj and verb outside of quote
int verbTokPos = TokenToLocation(verb.BackingLabel());
int subjTokPos = TokenToLocation(verb.BackingLabel());
if (InRange(range, verbTokPos) && InRange(range, subjTokPos) && commonSpeechWords.Contains(verb.Lemma()))
{
if (subj.Tag().Equals("NNP"))
{
int startChar = subj.BeginPosition();
for (int i = 0; i < names.Count; i++)
{
Pair <int, int> nameIndex = nameIndices[i];
//avoid names that don't actually exist in
if (RangeContainsCharIndex(nameIndex, startChar))
{
FillInMention(quote, TokenRangeToString(nameIndex), nameIndex.first, nameIndex.second, sieveName, Name);
break;
}
}
}
else
{
if (subj.Tag().Equals("PRP"))
{
int loc = TokenToLocation(subj.BackingLabel());
FillInMention(quote, subj.Word(), loc, loc, sieveName, Pronoun);
break;
}
else
{
if (subj.Tag().Equals("NN") && animacySet.Contains(subj.Word()))
{
int loc = TokenToLocation(subj.BackingLabel());
FillInMention(quote, subj.Word(), loc, loc, sieveName, AnimateNoun);
break;
}
}
}
}
}
}
}
public virtual string PrintSemanticGraph(SemanticGraph sg, bool unescapeParenthesis)
{
bool isTree = SemanticGraphUtils.IsTree(sg);
StringBuilder sb = new StringBuilder();
/* Print comments. */
foreach (string comment in sg.GetComments())
{
sb.Append(comment).Append("\n");
}
foreach (IndexedWord token in sg.VertexListSorted())
{
/* Check for multiword tokens. */
if (token.ContainsKey(typeof(CoreAnnotations.CoNLLUTokenSpanAnnotation)))
{
IntPair tokenSpan = token.Get(typeof(CoreAnnotations.CoNLLUTokenSpanAnnotation));
if (tokenSpan.GetSource() == token.Index())
{
string range = string.Format("%d-%d", tokenSpan.GetSource(), tokenSpan.GetTarget());
sb.Append(string.Format("%s\t%s\t_\t_\t_\t_\t_\t_\t_\t_%n", range, token.OriginalText()));
}
}
/* Try to find main governor and additional dependencies. */
string govIdx = null;
GrammaticalRelation reln = null;
Dictionary <string, string> enhancedDependencies = new Dictionary <string, string>();
foreach (IndexedWord parent in sg.GetParents(token))
{
SemanticGraphEdge edge = sg.GetEdge(parent, token);
if (govIdx == null && !edge.IsExtra())
{
govIdx = parent.ToCopyIndex();
reln = edge.GetRelation();
}
enhancedDependencies[parent.ToCopyIndex()] = edge.GetRelation().ToString();
}
string additionalDepsString = isTree ? "_" : CoNLLUUtils.ToExtraDepsString(enhancedDependencies);
string word = token.Word();
string featuresString = CoNLLUUtils.ToFeatureString(token.Get(typeof(CoreAnnotations.CoNLLUFeats)));
string pos = token.GetString <CoreAnnotations.PartOfSpeechAnnotation>("_");
string upos = token.GetString <CoreAnnotations.CoarseTagAnnotation>("_");
string misc = token.GetString <CoreAnnotations.CoNLLUMisc>("_");
string lemma = token.GetString <CoreAnnotations.LemmaAnnotation>("_");
string relnName = reln == null ? "_" : reln.ToString();
/* Root. */
if (govIdx == null && sg.GetRoots().Contains(token))
{
govIdx = "0";
relnName = GrammaticalRelation.Root.ToString();
additionalDepsString = isTree ? "_" : "0:" + relnName;
}
else
{
if (govIdx == null)
{
govIdx = "_";
relnName = "_";
}
}
if (unescapeParenthesis)
{
word = word.ReplaceAll(LrbPattern, "(");
word = word.ReplaceAll(RrbPattern, ")");
lemma = lemma.ReplaceAll(LrbPattern, "(");
lemma = lemma.ReplaceAll(RrbPattern, ")");
}
sb.Append(string.Format("%s\t%s\t%s\t%s\t%s\t%s\t%s\t%s\t%s\t%s%n", token.ToCopyIndex(), word, lemma, upos, pos, featuresString, govIdx, relnName, additionalDepsString, misc));
}
sb.Append("\n");
return(sb.ToString());
}