public override RawCorefSample read()
{
RawCorefSample sample = samples.read();
if (sample != null)
{
IList <Parse> enhancedParses = new List <Parse>();
IList <string[]> sentences = sample.Texts;
foreach (var sentence in sentences)
{
Parse p = FullParseCorefEnhancerStream.createIncompleteParse(sentence);
p.Type = AbstractBottomUpParser.TOP_NODE;
Parse[] parseTokens = p.Children;
// construct incomplete parse here ..
string[] tags = posTagger.tag(sentence);
for (int i = 0; i < parseTokens.Length; i++)
{
p.insert(new Parse(p.Text, parseTokens[i].Span, tags[i], 1d, parseTokens[i].HeadIndex));
}
// insert tags into incomplete parse
Span[] chunks = chunker.chunkAsSpans(sentence, tags);
foreach (Span chunk in chunks)
{
if ("NP".Equals(chunk.Type))
{
p.insert(new Parse(p.Text, new Span(0, 0), chunk.Type, 1d, p.HeadIndex));
}
}
enhancedParses.Add(p);
}
sample.Parses = enhancedParses;
return(sample);
}
else
{
return(null);
}
}
public override ObjectStream <CorefSample> create(string[] args)
{
Parameters @params = ArgumentParser.parse(args, typeof(Parameters));
ParserModel parserModel = (new ParserModelLoader()).load(@params.ParserModel);
Parser parser = ParserFactory.create(parserModel);
TokenizerModel tokenizerModel = (new TokenizerModelLoader()).load(@params.TokenizerModel);
Tokenizer tokenizer = new TokenizerME(tokenizerModel);
ObjectStream <string> mucDocStream = new FileToStringSampleStream(new DirectorySampleStream(@params.Data, new FileFilterAnonymousInnerClassHelper(this), false), Charset.forName("UTF-8"));
ObjectStream <RawCorefSample> rawSamples = new MucCorefSampleStream(tokenizer, mucDocStream);
ObjectStream <RawCorefSample> parsedSamples = new FullParseCorefEnhancerStream(parser, rawSamples);
// How to load all these nameFinder models ?!
// Lets make a param per model, not that nice, but ok!
IDictionary <string, Jfile> modelFileTagMap = new Dictionary <string, Jfile>();
modelFileTagMap["person"] = @params.PersonModel;
modelFileTagMap["organization"] = @params.OrganizationModel;
IList <TokenNameFinder> nameFinders = new List <TokenNameFinder>();
IList <string> tags = new List <string>();
foreach (KeyValuePair <string, Jfile> entry in modelFileTagMap)
{
nameFinders.Add(new NameFinderME((new TokenNameFinderModelLoader()).load(entry.Value)));
tags.Add(entry.Key);
}
return(new MucMentionInserterStream(new NameFinderCorefEnhancerStream(nameFinders.ToArray(), tags.ToArray(), parsedSamples)));
}