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

Class ConfidenceCorrectorEvaluator

java.lang.Object

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

public class ConfidenceCorrectorEvaluator
extends Object

Calculates the effectiveness of "constrained viterbi" in propagating corrections in one segment of a sequence to other segments.

Constructor Summary

Constructors

Constructor and Description
ConfidenceCorrectorEvaluator(Object[] startTags, Object[] inTags)

Method Summary

Modifier and Type	Method and Description
void	evaluate(Transducer model, ArrayList predictions, InstanceList ilist, ArrayList correctedSegments, String description, PrintStream outputStream, boolean errorsInUncorrected) Only evaluates over sequences which contain errors.

Methods inherited from class java.lang.Object

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

Constructor Detail

ConfidenceCorrectorEvaluator

public ConfidenceCorrectorEvaluator(Object[] startTags,
                                    Object[] inTags)

Method Detail

evaluate

public void evaluate(Transducer model,
                     ArrayList predictions,
                     InstanceList ilist,
                     ArrayList correctedSegments,
                     String description,
                     PrintStream outputStream,
                     boolean errorsInUncorrected)

Only evaluates over sequences which contain errors. Examine region not directly corrected by correctedSegments to measure effects of error propagation.

Parameters:

model - used to segment input sequence

predictions - list of the corrected segmentation

ilist - list of testing data

correctedSegments - list of Segments in each sequence that were corrected...currently only allows one segment per instance.

uncorrected - true if we only evaluate sequences where errors remain after correction