edu.umass.cs.mallet.base.fst.confidence

Class NBestViterbiConfidenceEstimator

java.lang.Object

edu.umass.cs.mallet.base.fst.confidence.TransducerSequenceConfidenceEstimator

edu.umass.cs.mallet.base.fst.confidence.NBestViterbiConfidenceEstimator

public class NBestViterbiConfidenceEstimator
extends TransducerSequenceConfidenceEstimator

Estimates the confidence of an entire sequence by the probability that one of the the Viterbi paths rank 2->N is correct. Note that this is a strange definition of confidence, and is mainly used for MultipleChoiceCRFActiveLearner, where we want to find Instances that are mislabeled, but are likely to have a correct labeling in the top N Viterbi paths.

Constructor Summary

Constructors

Constructor and Description
NBestViterbiConfidenceEstimator(Transducer model, int N)

Method Summary

Modifier and Type	Method and Description
double	estimateConfidenceFor(Instance instance, Object[] startTags, Object[] inTags) Calculates the confidence in the tagging of a Instance.

Methods inherited from class edu.umass.cs.mallet.base.fst.confidence.TransducerSequenceConfidenceEstimator

rankInstancesByConfidence

Methods inherited from class java.lang.Object

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

Constructor Detail

NBestViterbiConfidenceEstimator

public NBestViterbiConfidenceEstimator(Transducer model,
                                       int N)

Method Detail

estimateConfidenceFor

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

Calculates the confidence in the tagging of a Instance.

Specified by:

estimateConfidenceFor in class TransducerSequenceConfidenceEstimator