/// <summary>Generate the training features from the CoNLL input file.</summary>
/// <returns>Dataset of feature vectors</returns>
/// <exception cref="System.Exception"/>
private static GeneralDataset <string, string> GenerateFeatureVectors(Properties props)
{
GeneralDataset <string, string> dataset = new Dataset <string, string>();
Dictionaries dict = new Dictionaries(props);
DocumentMaker docMaker = new DocumentMaker(props, dict);
Document document;
while ((document = docMaker.NextDoc()) != null)
{
SetTokenIndices(document);
IDictionary <int, CorefCluster> entities = document.goldCorefClusters;
// Generate features for coreferent mentions with class label 1
foreach (CorefCluster entity in entities.Values)
{
foreach (Mention mention in entity.GetCorefMentions())
{
// Ignore verbal mentions
if (mention.headWord.Tag().StartsWith("V"))
{
continue;
}
IndexedWord head = mention.enhancedDependency.GetNodeByIndexSafe(mention.headWord.Index());
if (head == null)
{
continue;
}
List <string> feats = mention.GetSingletonFeatures(dict);
dataset.Add(new BasicDatum <string, string>(feats, "1"));
}
}
// Generate features for singletons with class label 0
List <CoreLabel> gold_heads = new List <CoreLabel>();
foreach (Mention gold_men in document.goldMentionsByID.Values)
{
gold_heads.Add(gold_men.headWord);
}
foreach (Mention predicted_men in document.predictedMentionsByID.Values)
{
SemanticGraph dep = predicted_men.enhancedDependency;
IndexedWord head = dep.GetNodeByIndexSafe(predicted_men.headWord.Index());
if (head == null || !dep.VertexSet().Contains(head))
{
continue;
}
// Ignore verbal mentions
if (predicted_men.headWord.Tag().StartsWith("V"))
{
continue;
}
// If the mention is in the gold set, it is not a singleton and thus ignore
if (gold_heads.Contains(predicted_men.headWord))
{
continue;
}
dataset.Add(new BasicDatum <string, string>(predicted_men.GetSingletonFeatures(dict), "0"));
}
}
dataset.SummaryStatistics();
return(dataset);
}
public CorefSystem(DocumentMaker docMaker, ICorefAlgorithm corefAlgorithm, bool removeSingletonClusters, bool verbose)
{
this.docMaker = docMaker;
this.corefAlgorithm = corefAlgorithm;
this.removeSingletonClusters = removeSingletonClusters;
this.verbose = verbose;
}
public CorefSystem(Properties props)
{
try
{
Dictionaries dictionaries = new Dictionaries(props);
docMaker = new DocumentMaker(props, dictionaries);
corefAlgorithm = ICorefAlgorithm.FromProps(props, dictionaries);
removeSingletonClusters = CorefProperties.RemoveSingletonClusters(props);
verbose = CorefProperties.Verbose(props);
}
catch (Exception e)
{
throw new Exception("Error initializing coref system", e);
}
}
/// <exception cref="System.Exception"/>
public HybridCorefSystem(Properties props)
{
this.props = props;
sieves = Edu.Stanford.Nlp.Coref.Hybrid.Sieve.Sieve.LoadSieves(props);
// set semantics loading
foreach (Edu.Stanford.Nlp.Coref.Hybrid.Sieve.Sieve sieve in sieves)
{
if (sieve.classifierType == Sieve.ClassifierType.Rule)
{
continue;
}
if (HybridCorefProperties.UseWordEmbedding(props, sieve.sievename))
{
props.SetProperty(HybridCorefProperties.LoadWordEmbeddingProp, "true");
}
}
dictionaries = new Edu.Stanford.Nlp.Coref.Data.Dictionaries(props);
docMaker = new DocumentMaker(props, dictionaries);
}