Package cc.mallet.fst.confidence

Class MaxEntSequenceConfidenceEstimator

java.lang.Object

cc.mallet.fst.confidence.TransducerSequenceConfidenceEstimator

cc.mallet.fst.confidence.MaxEntSequenceConfidenceEstimator

public class MaxEntSequenceConfidenceEstimator
extends TransducerSequenceConfidenceEstimator

Estimates the confidence of a Sequence extracted by a Transducer using a MaxEnt classifier to classify Sequences as "correct" or "incorrect." xxx needs some interface work.

Field Summary

Fields inherited from class cc.mallet.fst.confidence.TransducerSequenceConfidenceEstimator

model

Constructor Summary

Constructors

Constructor	Description
MaxEntSequenceConfidenceEstimator(Transducer model)
MaxEntSequenceConfidenceEstimator(Transducer model, double gaussianVariance)

Method Summary

Modifier and Type	Method	Description
double	estimateConfidenceFor(Instance instance, java.lang.Object[] startTags, java.lang.Object[] inTags)	Calculates the confidence in the tagging of an Instance.
MaxEnt	getClassifier()
PipedInstanceWithConfidence[]	rankPipedInstancesByConfidence(InstanceList ilist, java.lang.Object[] startTags, java.lang.Object[] continueTags)
MaxEnt	trainClassifier(InstanceList ilist, java.lang.String correct, java.lang.String incorrect)	Train underlying classifier on ilist.

Methods inherited from class cc.mallet.fst.confidence.TransducerSequenceConfidenceEstimator

rankInstancesByConfidence

Methods inherited from class java.lang.Object

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

Constructor Detail

MaxEntSequenceConfidenceEstimator

public MaxEntSequenceConfidenceEstimator(Transducer model,
                                         double gaussianVariance)

MaxEntSequenceConfidenceEstimator

public MaxEntSequenceConfidenceEstimator(Transducer model)

Method Detail

getClassifier

public MaxEnt getClassifier()

trainClassifier

public MaxEnt trainClassifier(InstanceList ilist,
                              java.lang.String correct,
                              java.lang.String incorrect)

Train underlying classifier on ilist. Assumes ilist has targst correct or incorrect.

Parameters:

ilist - training list to build correct/incorrect classifier

correct - "correct" label

incorrect - "incorrect" label

estimateConfidenceFor

public double estimateConfidenceFor(Instance instance,
                                    java.lang.Object[] startTags,
                                    java.lang.Object[] inTags)

Calculates the confidence in the tagging of an Instance.

Specified by:

estimateConfidenceFor in class TransducerSequenceConfidenceEstimator

rankPipedInstancesByConfidence

public PipedInstanceWithConfidence[] rankPipedInstancesByConfidence(InstanceList ilist,
                                                                    java.lang.Object[] startTags,
                                                                    java.lang.Object[] continueTags)