public DeterministicCorefAnnotator(Properties props)
{
// for backward compatibility
try
{
corefSystem = new SieveCoreferenceSystem(props);
mentionExtractor = new MentionExtractor(corefSystem.Dictionaries(), corefSystem.Semantics());
OldFormat = bool.Parse(props.GetProperty("oldCorefFormat", "false"));
allowReparsing = PropertiesUtils.GetBool(props, Constants.AllowReparsingProp, Constants.AllowReparsing);
// unless custom mention detection is set, just use the default coref mention detector
performMentionDetection = !PropertiesUtils.GetBool(props, "dcoref.useCustomMentionDetection", false);
if (performMentionDetection)
{
mentionAnnotator = new CorefMentionAnnotator(props);
}
}
catch (Exception e)
{
log.Error("cannot create DeterministicCorefAnnotator!");
log.Error(e);
throw new Exception(e);
}
}
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");
}
}