Implement your own applications

Steps for training

1. Create a dataset object from your training data.
2. Create a label map object from your training data.
3. Create an array list of feature types.
4. Create a parameter object from your configuration file and your label map.
5. Create a feature generator object from your features types and the parameters.
6. Initialize your feature generator using your training data.
7. Create a high-order semi-CRF model from your feature generator.
8. Call the train(...) method of the high-order semi-CRF model.

Steps for testing

1. Read the label map, feature generator, and the high-order semi-CRF model from files.
2. Create a dataset object from your testing data.
3. Call the runViterbi(...) method of the high-order semi-CRF model.
4. Score the results by using the Scorer class.

Examples

• PunctuationPredictor.java
• ReferenceTagger.java