Package edu.upf.taln.dri.module.summary.lexrank.summ

Class LexRanker

java.lang.Object

edu.upf.taln.dri.module.summary.lexrank.summ.LexRanker

public class LexRanker
extends Object

An Implementation of the LexRank algorithm described in the paper "LexRank: Graph-based Centrality as Salience in Text Summarization", Erkan & Radev '04, with some of our own modifications.

Constructor Summary

Constructors

Constructor	Description
LexRanker()

Method Summary

Modifier and Type	Method	Description
static <T extends Similar<T>> LexRankResults<T>	rank(List<T> data, double similarityThreshold, boolean continuous)	Runs the LexRank algorithm over a set of data.

Methods inherited from class java.lang.Object

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

Constructor Detail

LexRanker

public LexRanker()

Method Detail

rank

public static <T extends Similar<T>> LexRankResults<T> rank(List<T> data,
                                                            double similarityThreshold,
                                                            boolean continuous)

Runs the LexRank algorithm over a set of data. The data must have a similarity function, and it is assumed that the similarity function is symmetric. (If this is not the case, the similarity matrix will not be computed correctly.)

Parameters:

data - the data to rank.

similarityThreshold - how similar two items must be to be considered "connected". The LexRank paper suggests a value of 0.1.

continuous - whether or not to use a continuous version of the LexRank algorithm, If set to false, all similarity links above the similarity threshold will be considered equal; otherwise, the similarity scores are used. The paper authors note that non-continuous LexRank seems to perform better.