Package edu.stanford.nlp.parser.metrics

Class EvaluationMetric

java.lang.Object

edu.stanford.nlp.parser.metrics.EvaluationMetric

public class EvaluationMetric
extends Object

A general class that scores precision, recall, and F1. The package supports incremental computation of these statistics via the update(double, double, double, double) method. Formatted output reflecting the current values of the statistics may be obtained via the toString() method.

Author:

Spence Green

See Also:

Evalb

Constructor Summary

Constructors

Constructor	Description
EvaluationMetric()

Method Summary

Modifier and Type	Method	Description
Pair<Double,Double>	getPFractionals()	Returns the components of the precision.
Pair<Double,Double>	getRFractionals()	Returns the components of the recall.
double	getTestInstances()	Returns the number of test instances (e.g., parse trees or sentences) used in the calculation of the statistics.
double	numRelevantExamples()	A convenience method that returns the number of true positive examples from among the test instances.
String	toString()
void	update(double curP, double curPnum, double curR, double curRnum)	Updates the evaluation statistics.

Methods inherited from class java.lang.Object

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

Constructor Detail

EvaluationMetric

public EvaluationMetric()

Method Detail

update

public void update(double curP,
                   double curPnum,
                   double curR,
                   double curRnum)

Updates the evaluation statistics. Should be called once for each test example (e.g., sentence, parse tree, etc.).

Parameters:

curP - Precision of the current test example

curPnum - The denominator used to calculate the current precision

curR - Recall of the current test example

curRnum - The denominator used to calculate the current recall

getPFractionals

public Pair<Double,Double> getPFractionals()

Returns the components of the precision.

Returns:

A Pair with the numerator of the precision in the first element and the denominator of the precision in the second element.

getRFractionals

public Pair<Double,Double> getRFractionals()

Returns the components of the recall.

Returns:

A Pair with the numerator of the recall in the first element and the denominator of the recall in the second element.

getTestInstances

public double getTestInstances()

Returns the number of test instances (e.g., parse trees or sentences) used in the calculation of the statistics. This value corresponds to the number of calls to update(double, double, double, double).

Returns:

The number of test instances

numRelevantExamples

public double numRelevantExamples()

A convenience method that returns the number of true positive examples from among the test instances. Mathematically, this value is the denominator of the recall.

Returns:

Number of true positive examples

toString

public String toString()

Overrides:

toString in class Object