edu.washington.cs.knowitall.extractor.conf
Class WekaClassifierConfFunction<T>
java.lang.Object
edu.washington.cs.knowitall.extractor.conf.WekaClassifierConfFunction<T>
- Type Parameters:
T
-
public class WekaClassifierConfFunction<T>
- extends java.lang.Object
A confidence function that takes an object of type T
as input, and
returns a real valued number. This function operates in two steps.
First, it computes a boolean feature representation of the object using a
BooleanFeatureSet
object.
Second, it passes the featurized object to a binary classifier and returns the confidence
of the object belonging to the positive class. The binary classifier is represented
by a Weka Classifier
object.
The caller is responsible for making sure that the features defined by the
BooleanFeatureSet
object are consistent with the feature representation used
by the Classifier
object. If the BooleanFeatureSet
object has
n features, then the Classifier
object must have n numeric attributes and one
nominal attribute (corresponding to the class value). The order of the features in Classifier
must be the same as the order defined by BooleanFeatureSet
, with the class attribute
appearing last.
- Author:
- afader
Method Summary |
double |
getConf(T object)
Computes the confidence that the given object belongs to the positive class
according to the classifier. |
Methods inherited from class java.lang.Object |
clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait |
WekaClassifierConfFunction
public WekaClassifierConfFunction(BooleanFeatureSet<T> featureSet,
weka.classifiers.Classifier classifier)
- Constructs a new classifier confidence function that uses the given feature set and classifier.
- Parameters:
featureSet
- classifier
-
getConf
public double getConf(T object)
throws java.lang.Exception
- Computes the confidence that the given object belongs to the positive class
according to the classifier.
- Parameters:
object
-
- Returns:
- the confidence score
- Throws:
java.lang.Exception
- if the classifier is unable to compute the confidence