private void AddAcronyms(Annotation ann)
{
// Find all the organizations in a document
IList <ICoreMap> allMentionsSoFar = new List <ICoreMap>();
foreach (ICoreMap sentence in ann.Get(typeof(CoreAnnotations.SentencesAnnotation)))
{
Sharpen.Collections.AddAll(allMentionsSoFar, sentence.Get(typeof(CoreAnnotations.MentionsAnnotation)));
}
IList <IList <CoreLabel> > organizations = new List <IList <CoreLabel> >();
foreach (ICoreMap mention in allMentionsSoFar)
{
if ("ORGANIZATION".Equals(mention.Get(nerCoreAnnotationClass)))
{
organizations.Add(mention.Get(typeof(CoreAnnotations.TokensAnnotation)));
}
}
// Skip very long documents
if (organizations.Count > 100)
{
return;
}
// Iterate over tokens...
foreach (ICoreMap sentence_1 in ann.Get(typeof(CoreAnnotations.SentencesAnnotation)))
{
IList <ICoreMap> sentenceMentions = new List <ICoreMap>();
IList <CoreLabel> tokens = sentence_1.Get(typeof(CoreAnnotations.TokensAnnotation));
int totalTokensOffset = sentence_1.Get(typeof(CoreAnnotations.TokenBeginAnnotation));
for (int i = 0; i < tokens.Count; ++i)
{
// ... that look like they might be an acronym and are not already a mention
CoreLabel token = tokens[i];
if ("O".Equals(token.Ner()) && token.Word().ToUpper().Equals(token.Word()) && token.Word().Length >= 3)
{
foreach (IList <CoreLabel> org in organizations)
{
// ... and actually are an acronym
if (AcronymMatcher.IsAcronym(token.Word(), org))
{
// ... and add them.
// System.out.println("found ACRONYM ORG");
token.SetNER("ORGANIZATION");
ICoreMap chunk = ChunkAnnotationUtils.GetAnnotatedChunk(tokens, i, i + 1, totalTokensOffset, null, null, null);
chunk.Set(typeof(CoreAnnotations.NamedEntityTagAnnotation), "ORGANIZATION");
sentenceMentions.Add(chunk);
}
}
}
}
}
}
public static ICounter <string> ExtractFeatures(Mention p, ICollection <Mention> shares, ICollection <string> neStrings, Dictionaries dict, Properties props)
{
ICounter <string> features = new ClassicCounter <string>();
string span = p.LowercaseNormalizedSpanString();
string ner = p.headWord.Ner();
int sIdx = p.startIndex;
int eIdx = p.endIndex;
IList <CoreLabel> sent = p.sentenceWords;
CoreLabel preWord = (sIdx == 0) ? null : sent[sIdx - 1];
CoreLabel nextWord = (eIdx == sent.Count) ? null : sent[eIdx];
CoreLabel firstWord = p.originalSpan[0];
CoreLabel lastWord = p.originalSpan[p.originalSpan.Count - 1];
features.IncrementCount("B-NETYPE-" + ner);
if (neStrings.Contains(span))
{
features.IncrementCount("B-NE-STRING-EXIST");
if ((preWord == null || !preWord.Ner().Equals(ner)) && (nextWord == null || !nextWord.Ner().Equals(ner)))
{
features.IncrementCount("B-NE-FULLSPAN");
}
}
if (preWord != null)
{
features.IncrementCount("B-PRECEDINGWORD-" + preWord.Word());
}
if (nextWord != null)
{
features.IncrementCount("B-FOLLOWINGWORD-" + nextWord.Word());
}
if (preWord != null)
{
features.IncrementCount("B-PRECEDINGPOS-" + preWord.Tag());
}
if (nextWord != null)
{
features.IncrementCount("B-FOLLOWINGPOS-" + nextWord.Tag());
}
features.IncrementCount("B-FIRSTWORD-" + firstWord.Word());
features.IncrementCount("B-FIRSTPOS-" + firstWord.Tag());
features.IncrementCount("B-LASTWORD-" + lastWord.Word());
features.IncrementCount("B-LASTWORD-" + lastWord.Tag());
foreach (Mention s in shares)
{
if (s == p)
{
continue;
}
if (s.InsideIn(p))
{
features.IncrementCount("B-BIGGER-THAN-ANOTHER");
break;
}
}
foreach (Mention s_1 in shares)
{
if (s_1 == p)
{
continue;
}
if (p.InsideIn(s_1))
{
features.IncrementCount("B-SMALLER-THAN-ANOTHER");
break;
}
}
return(features);
}