gov.llnl.ontology.wordnet.wsd
Class GraphConnectivityDisambiguation

java.lang.Object
  gov.llnl.ontology.wordnet.wsd.GraphConnectivityDisambiguation

All Implemented Interfaces:

WordSenseDisambiguation

Direct Known Subclasses:

DegreeCentralityDisambiguation, PageRankCentralityDisambiguation

public abstract class GraphConnectivityDisambiguation
extends Object
implements WordSenseDisambiguation

An abstract base class for any of the graph centrality Word Sense Disambiguation algorithms described in the following paper:

• Navigli, R.; Lapata, M.; , "An Experimental Study of Graph Connectivity for Unsupervised Word Sense Disambiguation," Pattern Analysis and Machine Intelligence, IEEE Transactions on , vol.32, no.4, pp.678-692, April 2010. Available here

This base class extracts a small connected graph from the wordnet heirarchy that is centered around content words in a sentence that needs to be disambiguated. The target sense for each content word will be included in the graph, along with any GraphConnectivityDisambiguation.AnnotationSynsets in the shortest path connecting these target senses. The extracted graph structure for each focus word to be disambiguated will passed to subclasses as an affinity Matrix that records the known edges.

Author:

Keith Stevens

Nested Class Summary

class

GraphConnectivityDisambiguation.AnnotationSynset A structure class that represents a Annotation that needs to be disambiguated and it's possible target Synsets.

Constructor Summary

GraphConnectivityDisambiguation()

Method Summary

Sentence	disambiguate(Sentence sentence) Returns a new List of s which have word sense labels for each noun found in sentences.
Sentence	disambiguate(Sentence sentence, Set<Integer> focusIndices) Returns a new List of Sentences which have word sense labels for each annotation index stored in targetWords.
protected Synset[]	getSynsets(edu.stanford.nlp.pipeline.Annotation annot) Returns all of the GraphConnectivityDisambiguation.AnnotationSynsets found given the word and part of speech information, if present, in annot.
protected abstract void	processSentenceGraph(List<GraphConnectivityDisambiguation.AnnotationSynset> targetWords, Set<Synset> synsets, edu.ucla.sspace.basis.StringBasisMapping synsetBasis, edu.ucla.sspace.matrix.Matrix adjacencyMatrix) Disambiguates the Annotations in targetWords by using the Set GraphConnectivityDisambiguation.AnnotationSynsets carved out of the ontology for a given setnence.
void	setup(OntologyReader reader) Initializes the WordSenseDisambiguation algorithm with the given OntologyReader.

Methods inherited from class java.lang.Object

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

Constructor Detail

GraphConnectivityDisambiguation

public GraphConnectivityDisambiguation()

Method Detail

processSentenceGraph

protected abstract void processSentenceGraph(List<GraphConnectivityDisambiguation.AnnotationSynset> targetWords,
                                             Set<Synset> synsets,
                                             edu.ucla.sspace.basis.StringBasisMapping synsetBasis,
                                             edu.ucla.sspace.matrix.Matrix adjacencyMatrix)

Disambiguates the Annotations in targetWords by using the Set GraphConnectivityDisambiguation.AnnotationSynsets carved out of the ontology for a given setnence.

Parameters:

targetWords - A list of Annotations and their possible senses that need to be disambiguated. The Annotation in each GraphConnectivityDisambiguation.AnnotationSynset should be updated with the selected word sense.

synsets - The set of GraphConnectivityDisambiguation.AnnotationSynsets to consider for any graph traversal.

synsetBasis - A mapping from GraphConnectivityDisambiguation.AnnotationSynset names to indices in a graph.

adjacencyMatrix - An adjaceny matrix detailing how GraphConnectivityDisambiguation.AnnotationSynsets in synsets are connected.

setup

public void setup(OntologyReader reader)

Initializes the WordSenseDisambiguation algorithm with the given OntologyReader. Any other configuration values should be set via the global system properties object.

Specified by:

setup in interface WordSenseDisambiguation

disambiguate

public Sentence disambiguate(Sentence sentence)

Returns a new List of s which have word sense labels for each noun found in sentences.

Specified by:

disambiguate in interface WordSenseDisambiguation

disambiguate

public Sentence disambiguate(Sentence sentence,
                             Set<Integer> focusIndices)

Returns a new List of Sentences which have word sense labels for each annotation index stored in targetWords.

Specified by:

disambiguate in interface WordSenseDisambiguation

getSynsets

protected Synset[] getSynsets(edu.stanford.nlp.pipeline.Annotation annot)

Returns all of the GraphConnectivityDisambiguation.AnnotationSynsets found given the word and part of speech information, if present, in annot. If the part of speech is available, but provides no synsets, all possible synsets are returned for the word, under the assumption that the tag may be incorrect.