fr.inrialpes.exmo.align.impl.eval
Class PRGraphEvaluator

java.lang.Object
  fr.inrialpes.exmo.align.impl.eval.GraphEvaluator
      fr.inrialpes.exmo.align.impl.eval.PRGraphEvaluator

public class PRGraphEvaluator
extends GraphEvaluator

Compute the precision recall graph on 11 points

Version:

$Id: PRGraphEvaluator.java 1356 2010-03-25 14:19:41Z euzenat $ The computation is remotely inspired from the sample programme of Raymond J. Mooney available under GPL from http://www.cs.utexas.edu/users/mooney/ir-course/ Mooney also provides the averaging of these graphs over several queries: unfortunatelly, the resulting graph is not anymore a Precision/Recall graph This works perfectly correctly. I mention below the point which are mentionned as design points in a forecoming Exmotto entry: [R=0%] What should be P when R is 0% (obviously 100%) [R=100%] What should be P when R=100% is unreachable [Interp.] How is a chaotic curve interpolated Note: a very interesting measure is the MAP (mean average precision) which is figuratively the area under the curve and more precisely the average precision obtained for each correspondence in the reference alignment. The problem is that it can only be valid if the compared alignment has provided all the correspondences in the reference. Otherwise, it would basically be: SUM_c\in correct( P( c ) ) / nbexpected NOTE (JE:2010): This should be adapated for other notions of Precision and Recall This is fully possible and will be done with

Author:

Jerome Euzenat

Field Summary

private double	map
private int	STEP
private double	sumprecisions

Fields inherited from class fr.inrialpes.exmo.align.impl.eval.GraphEvaluator

cellSet, nbexpected, points

Constructor Summary

PRGraphEvaluator()

Method Summary

Vector<Pair>	eval() Compute precision and recall graphs.
Vector<Pair>	eval(Properties params) Compute precision and recall graphs.
Vector<Pair>	evalOldStyle() Returns a list of Precision at recall points (Pairs) From an ordered vector of cells with their correctness status
Vector<Pair>	evalOpenEnded() Returns a list of Precision at recall points (Pairs) From an ordered vector of cells with their correctness status
double	getGlobalResult() Retuns a simple global evaluation measure if any
double	getMap()
void	write(PrintWriter writer) This output the result

Methods inherited from class fr.inrialpes.exmo.align.impl.eval.GraphEvaluator

getStep, ingest, initCellSet, isCorrect, nbCells, setStep, writePlot, writeXMLMap

Methods inherited from class java.lang.Object

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

Field Detail

STEP

private int STEP

map

private double map

sumprecisions

private double sumprecisions

Constructor Detail

PRGraphEvaluator

public PRGraphEvaluator()

Method Detail

eval

public Vector<Pair> eval()

Compute precision and recall graphs.

Specified by:

eval in class GraphEvaluator

eval

public Vector<Pair> eval(Properties params)

Compute precision and recall graphs.

Specified by:

eval in class GraphEvaluator

evalOpenEnded

public Vector<Pair> evalOpenEnded()

Returns a list of Precision at recall points (Pairs) From an ordered vector of cells with their correctness status

evalOldStyle

public Vector<Pair> evalOldStyle()

Returns a list of Precision at recall points (Pairs) From an ordered vector of cells with their correctness status

write

public void write(PrintWriter writer)
           throws IOException

This output the result

Throws:

IOException

getMap

public double getMap()

getGlobalResult

public double getGlobalResult()

Description copied from class: GraphEvaluator

Retuns a simple global evaluation measure if any

Specified by:

getGlobalResult in class GraphEvaluator