edu.illinois.cs.cogcomp.lbj.coref.util.stats
Class Correctness
java.lang.Object
edu.illinois.cs.cogcomp.lbj.coref.util.stats.Correctness
public class Correctness
- extends java.lang.Object
Method Summary |
static double |
calcF1(LBJ2.classify.Classifier tested,
LBJ2.classify.Classifier correct,
java.util.List<? extends java.lang.Object> examples)
Since labels are always positive for their class (even in the two
class (e.g. |
static double |
calcF1(double p,
double r)
|
static double |
calcF1(int truePosPredicts,
int truePos,
int posPredicts)
|
static double |
calcF1(java.util.List<java.lang.String> gold,
java.util.List<java.lang.String> predicted)
|
static double |
calcF1Multi(LBJ2.classify.Classifier tested,
LBJ2.classify.Classifier correct,
java.util.List<? extends java.lang.Object> examples)
|
static double |
calcF1Multi(java.util.Map<java.lang.String,java.lang.Integer> truePosPredicts,
java.util.Map<java.lang.String,java.lang.Integer> truePos,
java.util.Map<java.lang.String,java.lang.Integer> posPredicts)
Currently calculates based on the harmonic mean of the individual
labels. |
static double |
calcMultiNoneF1(LBJ2.classify.Classifier predictor,
LBJ2.classify.Classifier oracle,
java.util.List<? extends java.lang.Object> examples)
Calculates the F1 for Classifiers that return a multi-class label,
all of which labels are considered positive except the "NONE",
"negative", and "false" labels. |
static double |
calcPrecision(int truePosPredicts,
int posPredicts)
|
static double |
calcRecall(int truePosPredicts,
int truePos)
|
Methods inherited from class java.lang.Object |
clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait |
Correctness
public Correctness()
calcMultiNoneF1
public static double calcMultiNoneF1(LBJ2.classify.Classifier predictor,
LBJ2.classify.Classifier oracle,
java.util.List<? extends java.lang.Object> examples)
- Calculates the F1 for Classifiers that return a multi-class label,
all of which labels are considered positive except the "NONE",
"negative", and "false" labels.
- Parameters:
predictor
- The predicting classifier.oracle
- The true classifier.examples
- The examples over which to evaulate the predictor.
- Returns:
- The F-Score.
calcF1Multi
public static double calcF1Multi(LBJ2.classify.Classifier tested,
LBJ2.classify.Classifier correct,
java.util.List<? extends java.lang.Object> examples)
- Parameters:
tested
- The Classifier producing labels to be tested.correct
- The classifier producing known correct labels.examples
- A list of examples to test.
- Returns:
- A combined F (Beta=1) Score.
calcF1Multi
public static double calcF1Multi(java.util.Map<java.lang.String,java.lang.Integer> truePosPredicts,
java.util.Map<java.lang.String,java.lang.Integer> truePos,
java.util.Map<java.lang.String,java.lang.Integer> posPredicts)
- Currently calculates based on the harmonic mean of the individual
labels.
calcF1
public static double calcF1(LBJ2.classify.Classifier tested,
LBJ2.classify.Classifier correct,
java.util.List<? extends java.lang.Object> examples)
- Since labels are always positive for their class (even in the two
class (e.g. binary) case, we assume that positive labels start with
T, t, P, or p and negative start with anything else (e.g. F, f, N, or
n)
- Parameters:
tested
- Classifier to be testedcorrect
- Classifier producing correct labelsexamples
- List of examples to classify
- Returns:
- F measure with Beta=1.
calcF1
public static double calcF1(java.util.List<java.lang.String> gold,
java.util.List<java.lang.String> predicted)
calcF1
public static double calcF1(int truePosPredicts,
int truePos,
int posPredicts)
calcF1
public static double calcF1(double p,
double r)
calcPrecision
public static double calcPrecision(int truePosPredicts,
int posPredicts)
calcRecall
public static double calcRecall(int truePosPredicts,
int truePos)