Custom Feature Generation in OpenNLP Namefinder API

Question

I am trying to use the Custom Feature generation of OpenNLP for Named Finder API.

http://opennlp.apache.org/documentation/1.5.3/manual/opennlp.html

I went through the documentation but I was not able to understand how to specify the different features.

It just says:

AdaptiveFeatureGenerator featureGenerator = new CachedFeatureGenerator(
     new AdaptiveFeatureGenerator[]{
       new WindowFeatureGenerator(new TokenFeatureGenerator(), 2, 2),
       new WindowFeatureGenerator(new TokenClassFeatureGenerator(true), 2, 2),
       new OutcomePriorFeatureGenerator(),
       new PreviousMapFeatureGenerator(),
       new BigramNameFeatureGenerator(),
       new SentenceFeatureGenerator(true, false)
       });

But how do you actually use each of these different feature generators to create my own custom features. Can somebody post a sample code defining these feature generators and how do we use it for custom feature generation. Thanks in anticipation.

Orbicular

I am also stuck with the same. Were you able to solve this issue ? – Waylin

Answer 1

I did not find any documentation either, but for most feature generators the code is quite self explanatory. Here are some links to the source repository:

• TokenFeatureGenerator
• TokenClassFeatureGenerator
• OutcomePriorFeatureGenerator
• PreviousMapFeatureGenerator
• SentenceFeatureGenerator
• BigramNameFeatureGenerator
• WindowFeatureGenerator
• CachedFeatureGenerator

Furthermore, the documentation you refer to states that the quoted feature generator is similar to the default feature generator. However, the source code shows that it is actually identical to the default feature generator.

Portsalut