I am migrating from OpenNLP 1.6 to 1.8 and wondering how best to train the 
NameFinder.  
I used the following API in 1.6
  public static TokenNameFinderModel train(String languageCode, String type, 
ObjectStream<NameSample> samples,
          TrainingParameters trainParams, AdaptiveFeatureGenerator generator, 
final Map<String, Object> resources)
          throws IOException 

In 1.8 the API is 
  public static TokenNameFinderModel train(String languageCode, String type,
          ObjectStream<NameSample> samples, TrainingParameters trainParams,
          TokenNameFinderFactory factory) throws IOException

where the TokenNameFinderFactory constructor is
  public TokenNameFinderFactory(byte[] featureGeneratorBytes, final Map<String, 
Object> resources,
      SequenceCodec<String> seqCodec) 

In opennlp.tools.util.featuregen.eneratorFactory, the 
TokenFeatureGeneratorFactory does not support the lowercase attribute, which 
prevents me from using the XML based definition.  
Any suggestions ?
Thanks
- viraf

Reply via email to