csli.util.classify.stanford
Class ExternalClassifierFactory

java.lang.Object
  csli.util.classify.stanford.ExternalClassifierFactory

All Implemented Interfaces:

edu.stanford.nlp.classify.ClassifierFactory

Direct Known Subclasses:

C4_5ClassifierFactory, SvmLightClassifierFactory

public abstract class ExternalClassifierFactory
extends Object
implements edu.stanford.nlp.classify.ClassifierFactory

An extension for the standard Stanford ClassifierFactory class for use with external stand-alone classifiers. Includes result cacheing, feature normalization, and mapping from feature objects to numerical indices (for external classifier packages which like that sort of thing)

Author:

mpurver

Constructor Summary

ExternalClassifierFactory(String keyStem, String fileStem)

Method Summary

String	getClassifyCommand()
String	getFileStem()
String	getTrainCommand()
boolean	isBag()
boolean	isNormalize()
void	setBag(boolean bag)
void	setNormalize(boolean normalize)

Methods inherited from class java.lang.Object

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

Methods inherited from interface edu.stanford.nlp.classify.ClassifierFactory

trainClassifier

Constructor Detail

ExternalClassifierFactory

public ExternalClassifierFactory(String keyStem,
                                 String fileStem)

Method Detail

getTrainCommand

public String getTrainCommand()

Returns:

the trainCommand

getClassifyCommand

public String getClassifyCommand()

Returns:

the classifyCommand

getFileStem

public String getFileStem()

Returns:

the fileStem

isBag

public boolean isBag()

Returns:

if true, features are to be specified as a bag of Objects or ScoredObjects; one feature per unique Object will be created, with its value set as the number of instances of that Object in the bag, or the total score of the ScoredObject instances. If false, features are to be specified as a list of numerical values in consistent order. (default false)

setBag

public void setBag(boolean bag)

Parameters:

bag - if true, features are to be specified as a bag of Objects or ScoredObjects; one feature per unique Object will be created, with its value set as the number of instances of that Object in the bag, or the total score of the ScoredObject instances. If false, features are to be specified as a list of numerical values in consistent order. (default false)

isNormalize

public boolean isNormalize()

Returns:

if true, normalize features to the range 0-1

setNormalize

public void setNormalize(boolean normalize)

Parameters:

normalize - if true, normalize features to the range 0-1