以事件为特征的文本聚类方法  

A METHOD OF TEXT CLUSTERING CHARACTERISED BY EVENT

作　　者：王崇国[1] 

机构地区：[1]新疆大学信息科学与工程学院,新疆乌鲁木齐830046

出　　处：《计算机应用与软件》2010年第8期195-197,共3页Computer Applications and Software

摘　　要：文本聚类是聚类的一个重要的研究分支,是聚类方法在自然语言处理领域的具体应用。研究表明事件在大量的文本中是客观存在的,诸多文本是由事件组成的,事件是文本表示的最小语义单位。探讨了以事件为特征的文本聚类方法,首先给出了基于事件的文本概念格表示模型,在此模型的基础上,给出了文本相似度的计算方法。最后,使用K-means聚类算法对所提方法进行了实验验证。Text clustering, one of the important research branches in clustering, is the application of clustering algorithm in natural language processing. Some research results have shown that the events are the objective existence in many texts. The texts are composed of events, and the event is the basic semantic units for text ＇ s representation. In the paper we discuss the text clustering method characterised by events. First- ly the event-based text concept lattice representation model is given, and then the text similarity computing method based on this model is pres- ented. Finally, the method proposed is validated by an experiment with K-means clustering algorithm.

关 键 词：文本聚类 事件 文本相似度 概念格 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]