edu.washington.cs.knowitall.extractor.conf
Class ReVerbConfFunction

java.lang.Object
  extended by edu.washington.cs.knowitall.extractor.conf.ReVerbConfFunction

public class ReVerbConfFunction
extends java.lang.Object

An extraction confidence function that is backed by a logistic regression classifier. This function will assign an extraction a real valued number between 0 and 1 according to the logistic regression model. It represents an extraction using the boolean features defined by the ReVerbFeatures class. See that documentation for details. It uses the model returned by DefaultObjects.getDefaultConfClassifier, which searches the classpath for a file called "conf.weka".

Author:
afader

Constructor Summary
ReVerbConfFunction()
          Constructs a new instance of the confidence function.
 
Method Summary
 double getConf(ChunkedBinaryExtraction extr)
           
 
Methods inherited from class java.lang.Object
clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait
 

Constructor Detail

ReVerbConfFunction

public ReVerbConfFunction()
                   throws java.io.IOException
Constructs a new instance of the confidence function.

Throws:
java.io.IOException - if unable to load the classifier model
Method Detail

getConf

public double getConf(ChunkedBinaryExtraction extr)
               throws ConfidenceFunctionException
Parameters:
extr -
Returns:
the probability that the given extraction belongs to the positive class
Throws:
ConfidenceFunctionException - if unable to compute the confidence score