edu.nyu.jet.hmm

Class XNameTagger

java.lang.Object

edu.nyu.jet.hmm.XNameTagger

public class XNameTagger
extends Object

A modified version of HMMNameTagger for experiments with N-best tagging. This skeleton does rescoring using a combination of - the probability from the baseline HMM - a coref measure (the total number of name coreferences) - a relation measure (the total number of relations)

Field Summary

Fields

Modifier and Type	Field and Description
HMMannotator	annotator
HMM	nameHMM

Constructor Summary

Constructors

Constructor and Description
XNameTagger(Class emitterClass) creates a new XNameTagger (with an empty HMM).

Method Summary

Methods

Modifier and Type	Method and Description
void	load(String fileName) load an HMM from file fileName.
static void	main(String[] args) procedures for training and testing the named entity tagger with rescoring.
void	NbestTrainCollection(String testCollection, String keyCollection) generate reranking model training data from documents in 'testCollection', where corresponding NE keys are in 'keyCollection'.
void	scoreCollection(String testCollection, String keyCollection) computes the recall/precision of 'testCollection' with respect to 'keyCollection' (which should have the same documents) with respect to the name annotations in 'tagsToScore'.
void	scoreCollectionWithOracle(String testCollection, String keyCollection) computes the recall/precision of 'testCollection' with respect to 'keyCollection' (which should have the same documents) with respect to the name annotations in 'tagsToScore'.
ArrayList	tag(Document doc, Span span, String sentno) tag span 'span' of Document 'doc' with N-best Named Entity annotations.

Methods inherited from class java.lang.Object

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

Field Detail

nameHMM

public HMM nameHMM

annotator

public HMMannotator annotator

Constructor Detail

XNameTagger

public XNameTagger(Class emitterClass)

creates a new XNameTagger (with an empty HMM).

Parameters:

emitterClass - the class of the emitter associated with each state of the HMM; must be a subclass of HMMemitter.

Method Detail

load

public void load(String fileName)
          throws IOException

load an HMM from file fileName.

Throws:

IOException

tag

public ArrayList tag(Document doc,
            Span span,
            String sentno)

tag span 'span' of Document 'doc' with N-best Named Entity annotations.

Returns:

a list of the hypothesis id-s

scoreCollection

public void scoreCollection(String testCollection,
                   String keyCollection)

computes the recall/precision of 'testCollection' with respect to 'keyCollection' (which should have the same documents) with respect to the name annotations in 'tagsToScore'. Reports both per-document and total scores to System.out.

scoreCollectionWithOracle

public void scoreCollectionWithOracle(String testCollection,
                             String keyCollection)

computes the recall/precision of 'testCollection' with respect to 'keyCollection' (which should have the same documents) with respect to the name annotations in 'tagsToScore'. Reports both per-document and total scores to System.out. This modified version computes both the score for the first hypothesis (the baseline score) and the score for the best of the N hypotheses (the ceiling for N-best rescoring), where 'best' is defined in terms of hypothesis overlap.

main

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

procedures for training and testing the named entity tagger with rescoring. At present it is set up to train and test on the same documents, a 20-document set of ACE 2004 documents.

Throws:

IOException

NbestTrainCollection

public void NbestTrainCollection(String testCollection,
                        String keyCollection)
                          throws IOException

generate reranking model training data from documents in 'testCollection', where corresponding NE keys are in 'keyCollection'.

Throws:

IOException