dragon.ir.summarize

Class ClusterLexRankSummarizer

java.lang.Object

dragon.ir.summarize.AbstractSentenceSum

dragon.ir.summarize.LexRankSummarizer

dragon.ir.summarize.ClusterLexRankSummarizer

All Implemented Interfaces:

GenericMultiDocSummarizer

public class ClusterLexRankSummarizer
extends LexRankSummarizer

An extension of LexRank text summarizer

LexRank select sentences purely according to the importance of sentences generated by the power method. A poetential problem is that selected top sentences may belong to the same cluster, i.e. not representative. To handle this problem, ClusterLexRank namely clusters those sentences, and then select one sentences with highest score in each cluster and choose remaining sentences acoording to the ranking score globally if nencessary

Copyright: Copyright (c) 2005

Company: IST, Drexel University

Version:

1.0

Author:

Davis Zhou

Field Summary

Fields

Modifier and Type	Field and Description
protected ClusterModel	clusterModel

Fields inherited from class dragon.ir.summarize.LexRankSummarizer

collectionReader, distanceMetric, indexer, indexReader, powerMethod, threshold, useContinuousValue, useTFIDF

Constructor Summary

Constructors

Constructor and Description
ClusterLexRankSummarizer(OnlineSentenceIndexer indexer, boolean useTFIDF, int clusterNum)
ClusterLexRankSummarizer(OnlineSentenceIndexer indexer, int clusterNum)

Method Summary

Modifier and Type	Method and Description
protected double	computeSimilarity(IRDoc firstSent, IRDoc secondSent) LexRank Summarizer uses cosine similrity.
static ClusterLexRankSummarizer	getClusterLexRankSummarizer(ConfigureNode node)
String	summarize(CollectionReader collectionReader, int maxLength)

Methods inherited from class dragon.ir.summarize.LexRankSummarizer

buildWeightMatrix, getSentenceSet, setContinuousScoreOpiton, setSimilarityThreshold

Methods inherited from class dragon.ir.summarize.AbstractSentenceSum

buildSummary, buildSummary

Methods inherited from class java.lang.Object

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

Field Detail

clusterModel

protected ClusterModel clusterModel

Constructor Detail

ClusterLexRankSummarizer

public ClusterLexRankSummarizer(OnlineSentenceIndexer indexer,
                                int clusterNum)

ClusterLexRankSummarizer

public ClusterLexRankSummarizer(OnlineSentenceIndexer indexer,
                                boolean useTFIDF,
                                int clusterNum)

Method Detail

getClusterLexRankSummarizer

public static ClusterLexRankSummarizer getClusterLexRankSummarizer(ConfigureNode node)

summarize

public String summarize(CollectionReader collectionReader,
                        int maxLength)

Specified by:

summarize in interface GenericMultiDocSummarizer

Overrides:

summarize in class LexRankSummarizer

Parameters:

collectionReader - the collection reader

maxLength - the length of the summary

Returns:

the summary in natural language.

computeSimilarity

protected double computeSimilarity(IRDoc firstSent,
                                   IRDoc secondSent)

Description copied from class: LexRankSummarizer

LexRank Summarizer uses cosine similrity. This method can be overrided.

Overrides:

computeSimilarity in class LexRankSummarizer

Returns:

the similarity between two specified sentences. The similarity value ranges from 0 to 1.