edu.stanford.nlp.ie.hmm
Class MergeTrainer

java.lang.Object
  |
  +--edu.stanford.nlp.ie.hmm.MergeTrainer

public class MergeTrainer

extends Object

Main class for building a single HMM by combining multiple target HMMs and a context HMM. Can either simply load and merge serialized HMMs or actually do the training given target fields.

See Also:

main(java.lang.String[])

Method Summary

static void

main(String[] args) Command-line utility for creating a single merged HMM from a set of single-target HMMs and a context HMM.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Method Detail

main

public static void main(String[] args)

Command-line utility for creating a single merged HMM from a set of single-target HMMs and a context HMM. There are two basic ways to run MergeTrainer, one in which the target HMMs are trained and then merged, and one in which previously trained and serialized target HMMs are simply loaded and merged:

Usage: java MergeTrainer newHMMFile trainingFile targetField [targetField2 ...]
       java MergeTrainer -f newHMMFile contextHMMFile targetHMMFile [targetHMMFIle2 ...]

In both cases, newHMMFile specifies the output for the serialized merged HMM. If -f is the first arg, the given context HMM and all given target HMMs are read in and merged together with no additional training. Otherwise the given training corpus is read in, a single target HMM is trained for each target field given, and then a single context HMM is trained and everything is glued together. Regardless of how the merged HMM is constructed, afterwards the transitions and states are printed out, and the merged HMM is serialized.