edu.stanford.nlp.ie.crf

Class CRFBiasedClassifier<IN extends CoreMap>

java.lang.Object

edu.stanford.nlp.ie.AbstractSequenceClassifier<IN>

edu.stanford.nlp.ie.crf.CRFClassifier<IN>

edu.stanford.nlp.ie.crf.CRFBiasedClassifier<IN>

All Implemented Interfaces:

java.util.function.Function<String,String>

public class CRFBiasedClassifier<IN extends CoreMap>
extends CRFClassifier<IN>

CRFBiasedClassifier is used to adjust the precision-recall tradeoff of any CRF model implemented using CRFClassifier. This adjustment is performed after CRF training. The method is described in (Minkov, Wang, Tomasic, and Cohen, 2006): "NER Systems that Suit User's Preferences: Adjusting the Recall-Precision Trade-off for Entity Extraction". CRFBiasedClassifier can import any model serialized with CRFClassifier and supports most command-line parameters available in CRFClassifier. In addition to this, CRFBiasedClassifier also interprets the parameter -classBias, as in:

java -server -mx500m edu.stanford.nlp.ie.crf.CRFBiasedClassifier -loadClassifier model.gz -testFile test.txt -classBias A:0.5,B:1.5

The command above sets a bias of 0.5 towards class A and a bias of 1.5 towards class B. These biases (which internally are treated as feature weights in the log-linear model underpinning the CRF classifier) can take any real value. As the weight of A tends to plus infinity, the classifier will only predict A labels, and as it tends towards minus infinity, it will never predict A labels.

Author:

Michel Galley, Sonal Gupta (made the class generic)

Field Summary

Fields inherited from class edu.stanford.nlp.ie.crf.CRFClassifier

DEFAULT_CLASSIFIER

Fields inherited from class edu.stanford.nlp.ie.AbstractSequenceClassifier

classIndex, featureFactories, flags, knownLCWords, pad, windowSize

Constructor Summary

Constructors

Constructor and Description
CRFBiasedClassifier(Properties props)
CRFBiasedClassifier(SeqClassifierFlags flags)

Method Summary

Modifier and Type	Method and Description
void	adjustBias(List<List<IN>> develData, java.util.function.Function<Double,Double> evalFunction, double low, double high) Adjust the bias parameter to optimize some objective function.
List<IN>	classify(List<IN> document) Classify a List of something that extendsCoreMap.
static void	main(String[] args) The main method, which is essentially the same as in CRFClassifier.
CRFDatum<List<String>,CRFLabel>	makeDatum(List<IN> info, int loc, List<FeatureFactory<IN>> featureFactories) Makes a CRFDatum by producing features and a label from input data at a specific position, using the provided factory.
void	setBiasWeight(int cindex, double weight)
void	setBiasWeight(String cname, double weight)

Methods inherited from class edu.stanford.nlp.ie.crf.CRFClassifier

addProcessedData, allLabels, classifyGibbs, classifyGibbs, classifyMaxEnt, classifyWithGlobalInformation, combine, documentsToDataAndLabels, documentsToDataAndLabelsList, documentToDataAndLabels, dropFeaturesBelowThreshold, dumpFeatures, extractDatumSequence, getClassifier, getClassifier, getClassifier, getClassifier, getClassifierNoExceptions, getCliquePotentialFunctionForTest, getCliqueTree, getCliqueTree, getCliqueTrees, getDefaultClassifier, getDefaultClassifier, getJarClassifier, getMinimizer, getMinimizer, getNumWeights, getObjectiveFunction, getSequenceModel, loadAuxiliaryData, loadClassifier, loadClassIndexFromFile, loadDefaultClassifier, loadDefaultClassifier, loadFeatureIndexFromFile, loadProcessedData, loadTagIndex, loadTextClassifier, loadTextClassifier, loadWeightsFromFile, makeAnswerArraysAndTagIndex, printFactorTable, printFactorTableDocument, printFactorTableDocuments, printFeatures, printFirstOrderProbs, printFirstOrderProbsDocument, printFirstOrderProbsDocuments, printLabelInformation, printLabelValue, printProbsDocument, pruneNodeFeatureIndices, saveProcessedData, scaleWeights, serializeClassifier, serializeClassifier, serializeClassIndex, serializeFeatureIndex, serializeTextClassifier, serializeTextClassifier, serializeWeights, to2D, topWeights, train, trainWeights, updateWeightsForTest, writeWeights

Methods inherited from class edu.stanford.nlp.ie.AbstractSequenceClassifier

apply, backgroundSymbol, classify, classifyAndWriteAnswers, classifyAndWriteAnswers, classifyAndWriteAnswers, classifyAndWriteAnswers, classifyAndWriteAnswers, classifyAndWriteAnswers, classifyAndWriteAnswersKBest, classifyAndWriteAnswersKBest, classifyAndWriteViterbiSearchGraph, classifyFile, classifyFilesAndWriteAnswers, classifyFilesAndWriteAnswers, classifyKBest, classifyRaw, classifySentence, classifySentenceWithGlobalInformation, classifyStdin, classifyStdin, classifyToCharacterOffsets, classifyToString, classifyToString, classifyWithInlineXML, countResults, countResultsSegmenter, defaultReaderAndWriter, finalizeClassification, getKnownLCWords, getSampler, getViterbiSearchGraph, labels, loadClassifier, loadClassifier, loadClassifier, loadClassifier, loadClassifier, loadClassifier, loadClassifierNoExceptions, loadClassifierNoExceptions, loadClassifierNoExceptions, loadClassifierNoExceptions, loadClassifierNoExceptions, loadJarClassifier, makeObjectBankFromFile, makeObjectBankFromFile, makeObjectBankFromFiles, makeObjectBankFromFiles, makeObjectBankFromFiles, makeObjectBankFromReader, makeObjectBankFromString, makePlainTextReaderAndWriter, makeReaderAndWriter, plainTextReaderAndWriter, printFeatureLists, printFeatures, printProbs, printProbsDocuments, printResults, reinit, segmentString, segmentString, train, train, train, train, train, train, windowSize, writeAnswers

Methods inherited from class java.lang.Object

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

Methods inherited from interface java.util.function.Function

andThen, compose, identity

Constructor Detail

CRFBiasedClassifier

public CRFBiasedClassifier(Properties props)

CRFBiasedClassifier

public CRFBiasedClassifier(SeqClassifierFlags flags)

Method Detail

makeDatum

public CRFDatum<List<String>,CRFLabel> makeDatum(List<IN> info,
                                                 int loc,
                                                 List<FeatureFactory<IN>> featureFactories)

Description copied from class: CRFClassifier

Makes a CRFDatum by producing features and a label from input data at a specific position, using the provided factory.

Overrides:

makeDatum in class CRFClassifier<IN extends CoreMap>

Parameters:

info - The input data

loc - The position to build a datum at

featureFactories - The FeatureFactories to use to extract features

Returns:

The constructed CRFDatum

setBiasWeight

public void setBiasWeight(String cname,
                          double weight)

setBiasWeight

public void setBiasWeight(int cindex,
                          double weight)

classify

public List<IN> classify(List<IN> document)

Description copied from class: AbstractSequenceClassifier

Classify a List of something that extendsCoreMap. The classifications are added in place to the items of the document, which is also returned by this method

Overrides:

classify in class CRFClassifier<IN extends CoreMap>

Parameters:

document - A List of something that extends CoreMap.

Returns:

The same List, but with the elements annotated with their answers (stored under the CoreAnnotations.AnswerAnnotation key).

adjustBias

public void adjustBias(List<List<IN>> develData,
                       java.util.function.Function<Double,Double> evalFunction,
                       double low,
                       double high)

Adjust the bias parameter to optimize some objective function. Note that this function only tunes the bias parameter of one class (class of index 0), and is thus only useful for binary classification problems.

main

public static void main(String[] args)
                 throws Exception

The main method, which is essentially the same as in CRFClassifier. See the class documentation.

Throws:

Exception