de.aitools.dm.clusterlabeling.validation.external
Interface ExternalMeasure

All Known Implementing Classes:

MatchAtN, MeanAveragePrecision, MeanReciprocalRank, PrecisionAtN, PreferencedMeasure

public interface ExternalMeasure

The abstract class PreferencedMeasure is an advanced implementation and should be used if possible.

Version:

$Id: ExternalMeasure.java,v 1.1 2011/02/18 16:08:24 hoppe Exp $

Author:

johannes.kiesel(/\t)uni-weimar.de

Method Summary

double	eval() This will evaluate depending on the concrete implementation.
double	evalAt(int n) This will evaluate depending on the concrete implementation.
double	evalClusterLabels(java.lang.String[] clusterLabels) This will set the clusterLabels to be used on validation, most likely via setClusterLabels(clusterLabels) and the evaluate the pair referenceLabel - clusterLabels depending on the concrete implementation.
double	evalReferenceLabel(java.lang.String referenceLabel) This will set the referenceLabel to be used on validation, most likely via setReferenceLabel(referenceLabel) and the evaluate the pair referenceLabel - clusterLabels depending on the concrete implementation.
void	setClusterLabels(java.lang.String[] clusterLabels) This will set the clusterLabels to be used for Validation.
void	setReferenceLabel(java.lang.String referenceLabel) This will set the referenceLabel to be used on validation.

Method Detail

setReferenceLabel

void setReferenceLabel(java.lang.String referenceLabel)

This will set the referenceLabel to be used on validation. A referenceLabel is a phrase created for extern evaluation. Depending on the measure, the closer a clusterLabel fits to the referenceLabel, the higher the score taken.
Processing (stemming/converting to lower case) will take place according to the concrete implementation of the interface.

Parameters:

referenceLabel - the best descriptive phrase for the cluster

setClusterLabels

void setClusterLabels(java.lang.String[] clusterLabels)

This will set the clusterLabels to be used for Validation. A clusterLabel is a automatically generated label, which to evaluate is the subject of the classes of this package.
More clusterLabels can be given.
The order will certainly be important.
Processing (creating synonyms/stemming/converting to lower case) will take place according to the concrete implementation of the interface.

Parameters:

clusterLabels - generated labels for the cluster

evalReferenceLabel

double evalReferenceLabel(java.lang.String referenceLabel)

This will set the referenceLabel to be used on validation, most likely via setReferenceLabel(referenceLabel) and the evaluate the pair referenceLabel - clusterLabels depending on the concrete implementation.
Attention: The return-value may vary significantly in its meaning depending on the implementation. Higher values should suggest closer pairs.

Parameters:

referenceLabel - the best descriptive phrase for the cluster

evalClusterLabels

double evalClusterLabels(java.lang.String[] clusterLabels)

This will set the clusterLabels to be used on validation, most likely via setClusterLabels(clusterLabels) and the evaluate the pair referenceLabel - clusterLabels depending on the concrete implementation.
Attention: The return-value may vary significantly in its meaning depending on the implementation. Higher values should suggest closer pairs.

Parameters:

clusterLabels - generated labels for the cluster

eval

double eval()

This will evaluate depending on the concrete implementation.
Attention: The return-value may vary significantly in its meaning depending on the implementation. Higher values should suggest closer pairs.

evalAt

double evalAt(int n)

This will evaluate depending on the concrete implementation.
Attention: The return-value may vary significantly in its meaning depending on the implementation. Higher values should suggest closer pairs.