mark.nlp.features
Class ClassificationOrderer

java.lang.Object
  |
  +--mark.nlp.features.ClassificationOrderer

All Implemented Interfaces:

Orderer

Direct Known Subclasses:

SimpleOrderer, TwoLevelOrderer

public abstract class ClassificationOrderer

extends java.lang.Object

implements Orderer

An Orderer that orders according to usefullness during classification. After ordering, potential features with lower indices are more useful.

Field Summary

protected BagCorpusCounter[]	fCatCounters
protected BagCorpusCounter	fCounter
protected ObjectMap	fPotentialFeatures

Constructor Summary

ClassificationOrderer()

Method Summary

protected abstract java.util.ArrayList	computeScores() Subclasses implement the following function which returns a Score for each potential feature in the set of potential features.
ObjectMap	order(BagCorpusCounter counter, BagCorpusCounter[] catCounters, ObjectMap potentialFeatures, java.lang.String parameter) Given corpus counters for a data set and a set of potential features, orders the potential features according to how useful they will be during classification.
protected void	setParameter(java.lang.String parameter) Subclasses override the following function if they use the parameter.

Methods inherited from class java.lang.Object

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

Field Detail

fCounter

protected BagCorpusCounter fCounter

fCatCounters

protected BagCorpusCounter[] fCatCounters

fPotentialFeatures

protected ObjectMap fPotentialFeatures

Constructor Detail

ClassificationOrderer

public ClassificationOrderer()

Method Detail

setParameter

protected void setParameter(java.lang.String parameter)

Subclasses override the following function if they use the parameter.

Parameters:

parameter - the parameter.

computeScores

protected abstract java.util.ArrayList computeScores()
                                              throws java.lang.Exception

Subclasses implement the following function which returns a Score for each potential feature in the set of potential features.

Returns:

an array list of Scores. The array list is parallel to fVocabulary.

java.lang.Exception

order

public ObjectMap order(BagCorpusCounter counter,
                       BagCorpusCounter[] catCounters,
                       ObjectMap potentialFeatures,
                       java.lang.String parameter)
                throws java.lang.Exception

Given corpus counters for a data set and a set of potential features, orders the potential features according to how useful they will be during classification.

Specified by:

order in interface Orderer

Parameters:

counter - a corpus counter in which each bag corresponds to a category.

catCounters - a set of counters. For the counter with index i, bags correspond to documents within category i.

potentialFeatures - the set of potential features.

Returns:

the same set of potential features ordered according to the estimated usefullness during classification. Lower indices correspond to more useful features.

java.lang.Exception