edu.nyu.jet

Class MaxEntModel

java.lang.Object

edu.nyu.jet.MaxEntModel

public class MaxEntModel
extends Object

a wrapper for the maximum entropy code provided in the OpenNLP package. modified by Yifan He in 2014 to optionally call Mallet max ent.

Constructor Summary

Constructors

Constructor and Description
MaxEntModel() creates a new maximum entropy model.
MaxEntModel(String featureFileName, String modelFileName) creates a new maximum entropy model, specifying files for both the features and the resulting model.

Method Summary

Methods

Modifier and Type	Method and Description
void	addEvent(Datum d) invoked during training to add one training Datum d to the training set.
String	bestOutcome(Datum d) (for a trained model) returns the most likely outcome for Datum d.
void	buildModel()
int	getNumOutcomes()
String	getOutcome(int i)
double[]	getOutcomeProbabilities(Datum d)
void	initializeForTraining()
void	initializeForTraining(String featureFileName)
boolean	isLoaded()
void	loadModel()
void	loadModel(BufferedReader reader)
void	loadModel(String modelFileName)
double	prob(Datum d, String value) (for a trained model) returns the probability that the Datum d is classified as value.
void	saveModel()
void	saveModel(BufferedWriter writer)
void	saveModel(String modelFileName)
void	setCutoff(int cutoff) sets the feature cutoff.
void	setIterations(int iterations)

Methods inherited from class java.lang.Object

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

Constructor Detail

MaxEntModel

public MaxEntModel()

creates a new maximum entropy model.

MaxEntModel

public MaxEntModel(String featureFileName,
           String modelFileName)

creates a new maximum entropy model, specifying files for both the features and the resulting model.

Parameters:

featureFileName - the name of the file in which features will be stored during training

modelFileName - the name of the file in which the max ent model will be stored

Method Detail

initializeForTraining

public void initializeForTraining(String featureFileName)

initializeForTraining

public void initializeForTraining()

addEvent

public void addEvent(Datum d)

invoked during training to add one training Datum d to the training set.

setCutoff

public void setCutoff(int cutoff)

sets the feature cutoff. Features occurring fewer than cutoff times in the training set are ignored. Default value is 4.

setIterations

public void setIterations(int iterations)

buildModel

public void buildModel()

saveModel

public void saveModel()

saveModel

public void saveModel(String modelFileName)

saveModel

public void saveModel(BufferedWriter writer)

loadModel

public void loadModel()

loadModel

public void loadModel(String modelFileName)

loadModel

public void loadModel(BufferedReader reader)

isLoaded

public boolean isLoaded()

prob

public double prob(Datum d,
          String value)

(for a trained model) returns the probability that the Datum d is classified as value.

bestOutcome

public String bestOutcome(Datum d)

(for a trained model) returns the most likely outcome for Datum d.

getNumOutcomes

public int getNumOutcomes()

getOutcome

public String getOutcome(int i)

getOutcomeProbabilities

public double[] getOutcomeProbabilities(Datum d)