private static SemanticGraph ConvertTreeToBasic(Tree tree)
{
AddLemmata(tree);
AddNERTags(tree);
SemanticGraph sg = SemanticGraphFactory.MakeFromTree(tree, SemanticGraphFactory.Mode.Basic, GrammaticalStructure.Extras.None, null, false, true);
AddLemmata(sg);
if (UseName)
{
AddNERTags(sg);
}
return(sg);
}
protected internal override void DoOneSentence(Annotation annotation, ICoreMap sentence)
{
GrammaticalStructure gs = parser.Predict(sentence);
SemanticGraph deps = SemanticGraphFactory.MakeFromTree(gs, SemanticGraphFactory.Mode.Collapsed, extraDependencies, null);
SemanticGraph uncollapsedDeps = SemanticGraphFactory.MakeFromTree(gs, SemanticGraphFactory.Mode.Basic, extraDependencies, null);
SemanticGraph ccDeps = SemanticGraphFactory.MakeFromTree(gs, SemanticGraphFactory.Mode.Ccprocessed, extraDependencies, null);
SemanticGraph enhancedDeps = SemanticGraphFactory.MakeFromTree(gs, SemanticGraphFactory.Mode.Enhanced, extraDependencies, null);
SemanticGraph enhancedPlusPlusDeps = SemanticGraphFactory.MakeFromTree(gs, SemanticGraphFactory.Mode.EnhancedPlusPlus, extraDependencies, null);
sentence.Set(typeof(SemanticGraphCoreAnnotations.CollapsedDependenciesAnnotation), deps);
sentence.Set(typeof(SemanticGraphCoreAnnotations.BasicDependenciesAnnotation), uncollapsedDeps);
sentence.Set(typeof(SemanticGraphCoreAnnotations.CollapsedCCProcessedDependenciesAnnotation), ccDeps);
sentence.Set(typeof(SemanticGraphCoreAnnotations.EnhancedDependenciesAnnotation), enhancedDeps);
sentence.Set(typeof(SemanticGraphCoreAnnotations.EnhancedPlusPlusDependenciesAnnotation), enhancedPlusPlusDeps);
}
/// <exception cref="System.Exception"/>
public override Document NextDoc()
{
IList <IList <CoreLabel> > allWords = new List <IList <CoreLabel> >();
IList <Tree> allTrees = new List <Tree>();
CoNLL2011DocumentReader.Document conllDoc = reader.GetNextDocument();
if (conllDoc == null)
{
return(null);
}
Annotation anno = conllDoc.GetAnnotation();
IList <ICoreMap> sentences = anno.Get(typeof(CoreAnnotations.SentencesAnnotation));
foreach (ICoreMap sentence in sentences)
{
if (!Constants.UseGoldParses && !replicateCoNLL)
{
// Remove tree from annotation and replace with parse using stanford parser
sentence.Remove(typeof(TreeCoreAnnotations.TreeAnnotation));
}
else
{
Tree tree = sentence.Get(typeof(TreeCoreAnnotations.TreeAnnotation));
treeLemmatizer.TransformTree(tree);
// generate the dependency graph
try
{
SemanticGraph deps = SemanticGraphFactory.MakeFromTree(tree, SemanticGraphFactory.Mode.Enhanced, GrammaticalStructure.Extras.None);
SemanticGraph basicDeps = SemanticGraphFactory.MakeFromTree(tree, SemanticGraphFactory.Mode.Basic, GrammaticalStructure.Extras.None);
sentence.Set(typeof(SemanticGraphCoreAnnotations.BasicDependenciesAnnotation), basicDeps);
sentence.Set(typeof(SemanticGraphCoreAnnotations.EnhancedDependenciesAnnotation), deps);
}
catch (Exception e)
{
logger.Log(Level.Warning, "Exception caught during extraction of Stanford dependencies. Will ignore and continue...", e);
}
}
}
string preSpeaker = null;
int utterance = -1;
foreach (CoreLabel token in anno.Get(typeof(CoreAnnotations.TokensAnnotation)))
{
if (!token.ContainsKey(typeof(CoreAnnotations.SpeakerAnnotation)))
{
token.Set(typeof(CoreAnnotations.SpeakerAnnotation), string.Empty);
}
string curSpeaker = token.Get(typeof(CoreAnnotations.SpeakerAnnotation));
if (!curSpeaker.Equals(preSpeaker))
{
utterance++;
preSpeaker = curSpeaker;
}
token.Set(typeof(CoreAnnotations.UtteranceAnnotation), utterance);
}
// Run pipeline
stanfordProcessor.Annotate(anno);
foreach (ICoreMap sentence_1 in anno.Get(typeof(CoreAnnotations.SentencesAnnotation)))
{
allWords.Add(sentence_1.Get(typeof(CoreAnnotations.TokensAnnotation)));
allTrees.Add(sentence_1.Get(typeof(TreeCoreAnnotations.TreeAnnotation)));
}
// Initialize gold mentions
IList <IList <Mention> > allGoldMentions = ExtractGoldMentions(conllDoc);
IList <IList <Mention> > allPredictedMentions;
//allPredictedMentions = allGoldMentions;
// Make copy of gold mentions since mentions may be later merged, mentionID's changed and stuff
allPredictedMentions = mentionFinder.ExtractPredictedMentions(anno, maxID, dictionaries);
try
{
RecallErrors(allGoldMentions, allPredictedMentions, anno);
}
catch (IOException e)
{
throw new Exception(e);
}
Document doc = Arrange(anno, allWords, allTrees, allPredictedMentions, allGoldMentions, true);
doc.conllDoc = conllDoc;
return(doc);
}
public virtual void Annotate(Annotation annotation)
{
// temporarily set the primary named entity tag to the coarse tag
SetNamedEntityTagGranularity(annotation, "coarse");
if (performMentionDetection)
{
mentionAnnotator.Annotate(annotation);
}
try
{
IList <Tree> trees = new List <Tree>();
IList <IList <CoreLabel> > sentences = new List <IList <CoreLabel> >();
// extract trees and sentence words
// we are only supporting the new annotation standard for this Annotator!
bool hasSpeakerAnnotations = false;
if (annotation.ContainsKey(typeof(CoreAnnotations.SentencesAnnotation)))
{
// int sentNum = 0;
foreach (ICoreMap sentence in annotation.Get(typeof(CoreAnnotations.SentencesAnnotation)))
{
IList <CoreLabel> tokens = sentence.Get(typeof(CoreAnnotations.TokensAnnotation));
sentences.Add(tokens);
Tree tree = sentence.Get(typeof(TreeCoreAnnotations.TreeAnnotation));
trees.Add(tree);
SemanticGraph dependencies = SemanticGraphFactory.MakeFromTree(tree, SemanticGraphFactory.Mode.Collapsed, GrammaticalStructure.Extras.None, null, true);
// locking here is crucial for correct threading!
sentence.Set(typeof(SemanticGraphCoreAnnotations.AlternativeDependenciesAnnotation), dependencies);
if (!hasSpeakerAnnotations)
{
// check for speaker annotations
foreach (CoreLabel t in tokens)
{
if (t.Get(typeof(CoreAnnotations.SpeakerAnnotation)) != null)
{
hasSpeakerAnnotations = true;
break;
}
}
}
MentionExtractor.MergeLabels(tree, tokens);
MentionExtractor.InitializeUtterance(tokens);
}
}
else
{
log.Error("this coreference resolution system requires SentencesAnnotation!");
return;
}
if (hasSpeakerAnnotations)
{
annotation.Set(typeof(CoreAnnotations.UseMarkedDiscourseAnnotation), true);
}
// extract all possible mentions
// this is created for each new annotation because it is not threadsafe
RuleBasedCorefMentionFinder finder = new RuleBasedCorefMentionFinder(allowReparsing);
IList <IList <Mention> > allUnprocessedMentions = finder.ExtractPredictedMentions(annotation, 0, corefSystem.Dictionaries());
// add the relevant info to mentions and order them for coref
Document document = mentionExtractor.Arrange(annotation, sentences, trees, allUnprocessedMentions);
IList <IList <Mention> > orderedMentions = document.GetOrderedMentions();
IDictionary <int, CorefChain> result = corefSystem.CorefReturnHybridOutput(document);
annotation.Set(typeof(CorefCoreAnnotations.CorefChainAnnotation), result);
if (OldFormat)
{
IDictionary <int, CorefChain> oldResult = corefSystem.Coref(document);
AddObsoleteCoreferenceAnnotations(annotation, orderedMentions, oldResult);
}
}
catch (Exception e)
{
throw;
}
catch (Exception e)
{
throw new Exception(e);
}
finally
{
// restore to the fine-grained
SetNamedEntityTagGranularity(annotation, "fine");
}
}
/// <summary>Prints out all matches of a semgrex pattern on a file of dependencies.</summary>
/// <remarks>
/// Prints out all matches of a semgrex pattern on a file of dependencies.
/// <p>
/// Usage:<br />
/// java edu.stanford.nlp.semgraph.semgrex.SemgrexPattern [args]
/// <br />
/// See the help() function for a list of possible arguments to provide.
/// </remarks>
/// <exception cref="System.IO.IOException"/>
public static void Main(string[] args)
{
IDictionary <string, int> flagMap = Generics.NewHashMap();
flagMap[Pattern] = 1;
flagMap[TreeFile] = 1;
flagMap[Mode] = 1;
flagMap[Extras] = 1;
flagMap[ConlluFile] = 1;
flagMap[OutputFormatOption] = 1;
IDictionary <string, string[]> argsMap = StringUtils.ArgsToMap(args, flagMap);
// args = argsMap.get(null);
// TODO: allow patterns to be extracted from a file
if (!(argsMap.Contains(Pattern)) || argsMap[Pattern].Length == 0)
{
Help();
System.Environment.Exit(2);
}
Edu.Stanford.Nlp.Semgraph.Semgrex.SemgrexPattern semgrex = Edu.Stanford.Nlp.Semgraph.Semgrex.SemgrexPattern.Compile(argsMap[Pattern][0]);
string modeString = DefaultMode;
if (argsMap.Contains(Mode) && argsMap[Mode].Length > 0)
{
modeString = argsMap[Mode][0].ToUpper();
}
SemanticGraphFactory.Mode mode = SemanticGraphFactory.Mode.ValueOf(modeString);
string outputFormatString = DefaultOutputFormat;
if (argsMap.Contains(OutputFormatOption) && argsMap[OutputFormatOption].Length > 0)
{
outputFormatString = argsMap[OutputFormatOption][0].ToUpper();
}
SemgrexPattern.OutputFormat outputFormat = SemgrexPattern.OutputFormat.ValueOf(outputFormatString);
bool useExtras = true;
if (argsMap.Contains(Extras) && argsMap[Extras].Length > 0)
{
useExtras = bool.ValueOf(argsMap[Extras][0]);
}
IList <SemanticGraph> graphs = Generics.NewArrayList();
// TODO: allow other sources of graphs, such as dependency files
if (argsMap.Contains(TreeFile) && argsMap[TreeFile].Length > 0)
{
foreach (string treeFile in argsMap[TreeFile])
{
log.Info("Loading file " + treeFile);
MemoryTreebank treebank = new MemoryTreebank(new TreeNormalizer());
treebank.LoadPath(treeFile);
foreach (Tree tree in treebank)
{
// TODO: allow other languages... this defaults to English
SemanticGraph graph = SemanticGraphFactory.MakeFromTree(tree, mode, useExtras ? GrammaticalStructure.Extras.Maximal : GrammaticalStructure.Extras.None);
graphs.Add(graph);
}
}
}
if (argsMap.Contains(ConlluFile) && argsMap[ConlluFile].Length > 0)
{
CoNLLUDocumentReader reader = new CoNLLUDocumentReader();
foreach (string conlluFile in argsMap[ConlluFile])
{
log.Info("Loading file " + conlluFile);
IEnumerator <SemanticGraph> it = reader.GetIterator(IOUtils.ReaderFromString(conlluFile));
while (it.MoveNext())
{
SemanticGraph graph = it.Current;
graphs.Add(graph);
}
}
}
foreach (SemanticGraph graph_1 in graphs)
{
SemgrexMatcher matcher = semgrex.Matcher(graph_1);
if (!matcher.Find())
{
continue;
}
if (outputFormat == SemgrexPattern.OutputFormat.List)
{
log.Info("Matched graph:" + Runtime.LineSeparator() + graph_1.ToString(SemanticGraph.OutputFormat.List));
int i = 1;
bool found = true;
while (found)
{
log.Info("Match " + i + " at: " + matcher.GetMatch().ToString(CoreLabel.OutputFormat.ValueIndex));
IList <string> nodeNames = Generics.NewArrayList();
Sharpen.Collections.AddAll(nodeNames, matcher.GetNodeNames());
nodeNames.Sort();
foreach (string name in nodeNames)
{
log.Info(" " + name + ": " + matcher.GetNode(name).ToString(CoreLabel.OutputFormat.ValueIndex));
}
log.Info(" ");
found = matcher.Find();
}
}
else
{
if (outputFormat == SemgrexPattern.OutputFormat.Offset)
{
if (graph_1.VertexListSorted().IsEmpty())
{
continue;
}
System.Console.Out.Printf("+%d %s%n", graph_1.VertexListSorted()[0].Get(typeof(CoreAnnotations.LineNumberAnnotation)), argsMap[ConlluFile][0]);
}
}
}
}
public virtual void TestFind()
{
SemanticGraph h = SemanticGraph.ValueOf("[married/VBN nsubjpass>Hughes/NNP auxpass>was/VBD prep_to>Gracia/NNP]");
SemanticGraph t = SemanticGraph.ValueOf("[loved/VBD\nnsubj>Hughes/NNP\ndobj>[wife/NN poss>his/PRP$ appos>Gracia/NNP]\nconj_and>[obsessed/JJ\ncop>was/VBD\nadvmod>absolutely/RB\nprep_with>[Elicia/NN poss>his/PRP$ amod>little/JJ compound>daughter/NN]]]"
);
string s = "(ROOT\n(S\n(NP (DT The) (NN chimney) (NNS sweeps))\n(VP (VBP do) (RB not)\n(VP (VB like)\n(S\n(VP (VBG working)\n(PP (IN on)\n(NP (DT an) (JJ empty) (NN stomach)))))))\n(. .)))";
Tree tree = Tree.ValueOf(s);
SemanticGraph sg = SemanticGraphFactory.MakeFromTree(tree, SemanticGraphFactory.Mode.Collapsed, GrammaticalStructure.Extras.Maximal, null);
SemgrexPattern pat = SemgrexPattern.Compile("{}=gov ![>det {}] & > {word:/^(?!not).*$/}=dep");
sg.PrettyPrint();
// SemgrexPattern pat =
// SemgrexPattern.compile("{} [[<prep_to ({word:married} >nsubjpass {})] | [<nsubjpass ({word:married} >prep_to {})]]");
pat.PrettyPrint();
SemgrexMatcher mat = pat.Matcher(sg);
while (mat.Find())
{
// String match = mat.getMatch().word();
string gov = mat.GetNode("gov").Word();
// String reln = mat.getRelnString("reln");
string dep = mat.GetNode("dep").Word();
// System.out.println(match);
System.Console.Out.WriteLine(dep + ' ' + gov);
}
SemgrexPattern pat2 = SemgrexPattern.Compile("{} [[>/nn|appos/ ({lemma:/wife|husband|partner/} >/poss/ {}=txtPartner)] | [<poss ({}=txtPartner >/nn|appos/ {lemma:/wife|husband|partner/})]" + "| [<nsubj ({$} >> ({word:/wife|husband|partner/} >poss {word:/his|her/} >/nn|appos/ {}))]]"
);
SemgrexMatcher mat2 = pat2.Matcher(t);
while (mat2.Find())
{
string match = mat2.GetMatch().Word();
// String gov = mat.getNode("gov").word();
// String reln = mat.getRelnString("reln");
// String dep = mat.getNode("dep").word();
System.Console.Out.WriteLine(match);
}
// System.out.println(dep + " " + gov);
Dictionary <IndexedWord, IndexedWord> map = new Dictionary <IndexedWord, IndexedWord>();
map[h.GetNodeByWordPattern("Hughes")] = t.GetNodeByWordPattern("Hughes");
map[h.GetNodeByWordPattern("Gracia")] = t.GetNodeByWordPattern("Gracia");
Alignment alignment = new Alignment(map, 0, string.Empty);
SemgrexPattern fullPat = SemgrexPattern.Compile("({}=partnerOne [[<prep_to ({word:married} >nsubjpass {}=partnerTwo)] | [<nsubjpass ({word:married} >prep_to {}=partnerTwo)]]) @ ({} [[>/nn|appos/ ({lemma:/wife|husband|partner/} >/poss/ {}=txtPartner)] | [<poss ({}=txtPartner >/nn|appos/ {lemma:/wife|husband|partner/})]"
+ "| [<nsubj ({$} >> ({word:/wife|husband|partner/} >poss {word:/his|her/} >/nn|appos/ {}=txtPartner))]])");
fullPat.PrettyPrint();
SemgrexMatcher fullMat = fullPat.Matcher(h, alignment, t);
if (fullMat.Find())
{
System.Console.Out.WriteLine("woo: " + fullMat.GetMatch().Word());
System.Console.Out.WriteLine(fullMat.GetNode("txtPartner"));
System.Console.Out.WriteLine(fullMat.GetNode("partnerOne"));
System.Console.Out.WriteLine(fullMat.GetNode("partnerTwo"));
}
else
{
System.Console.Out.WriteLine("boo");
}
SemgrexPattern pat3 = SemgrexPattern.Compile("({word:LIKE}=parent >>/aux.*/ {word:/do/}=node)");
System.Console.Out.WriteLine("pattern is ");
pat3.PrettyPrint();
System.Console.Out.WriteLine("tree is ");
sg.PrettyPrint();
//checking if ignoring case or not
SemgrexMatcher mat3 = pat3.Matcher(sg, true);
if (mat3.Find())
{
string parent = mat3.GetNode("parent").Word();
string node = mat3.GetNode("node").Word();
System.Console.Out.WriteLine("Result: parent is " + parent + " and node is " + node);
NUnit.Framework.Assert.AreEqual(parent, "like");
NUnit.Framework.Assert.AreEqual(node, "do");
}
else
{
NUnit.Framework.Assert.Fail();
}
}