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

Class TransducerSequenceConfidenceEstimator

java.lang.Object

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

Direct Known Subclasses:

MaxEntSequenceConfidenceEstimator, MinSegmentConfidenceEstimator, NBestViterbiConfidenceEstimator, RandomSequenceConfidenceEstimator, SegmentProductConfidenceEstimator, ViterbiConfidenceEstimator, ViterbiRatioConfidenceEstimator

public abstract class TransducerSequenceConfidenceEstimator
extends Object

Abstract class that estimates the confidence of a Sequence extracted by a Transducer.Note that this is different from TransducerConfidenceEstimator, which estimates the confidence for a single Segment.

Constructor Summary

Constructors

Constructor and Description
TransducerSequenceConfidenceEstimator()

Method Summary

Modifier and Type	Method and Description
abstract double	estimateConfidenceFor(Instance instance, Object[] startTags, Object[] inTags) Calculates the confidence in the tagging of a Sequence.
InstanceWithConfidence[]	rankInstancesByConfidence(InstanceList ilist, Object[] startTags, Object[] continueTags) Ranks all Sequencess in this InstanceList by confidence estimate.

Methods inherited from class java.lang.Object

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

Constructor Detail

TransducerSequenceConfidenceEstimator

public TransducerSequenceConfidenceEstimator()

Method Detail

estimateConfidenceFor

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

Calculates the confidence in the tagging of a Sequence.

rankInstancesByConfidence

public InstanceWithConfidence[] rankInstancesByConfidence(InstanceList ilist,
                                                          Object[] startTags,
                                                          Object[] continueTags)

Ranks all Sequencess in this InstanceList by confidence estimate.

Parameters:

ilist - list of segmentation instances

startTags - represent the labels for the start states (B-) of all segments

continueTags - represent the labels for the continue state (I-) of all segments

Returns:

array of InstanceWithConfidences ordered by non-decreasing confidence scores, as calculated by estimateConfidenceFor