一种提高文本聚类算法质量的方法

针对基于VSM(vector space model)的文本聚类算法存在的主要问题,即忽略了词之间的语义信息、忽略了各维度之间的联系而导致文本的相似度计算不够精确,提出基于语义距离计算文档间相似度及两阶段聚类方案来提高文本聚类算法的质量.首先,从语义上分析文档,采用最近邻算法进行第一次聚类;其次,根据相似度权重,对类特征词进行优胜劣汰;然后进行类合并;最后,进行第二次聚类,解决最近邻算法对输入次序敏感的问题.实验结果表明,提出的方法在聚类精度和召回率上均有显著的提高,较好解决了基于VSM的文本聚类算法存在的问题.

A Method to Improve Text Clustering Algorithm Quality

The main problem with the text clustering algorithm based on vector space model(VSM) is that semantic information between words and the link between the various dimensions are overlooked,resulting in inaccuracy in the text similarity calculation.A method based on computing the text similarity using semantic distance and two-phrase clustering is proposed to improve the text clustering algorithm.First,the text analyzed according to its semantic,with nearest neighbor algorithm used for the first cluster.Some feature words are chosen according to the similarity weight to represent the cluster with the remaining feature words similar to the main themes of the cluster,and then class combination is carried out.Finally,the second clustering is carried out to improve the nearest neighbor clustering which is sensitive to the input order of the document.Simulation experiments indicate that the proposed algorithm can solve these problems and performs better than the text clustering algorithm based on VSM in the clustering precision and recall rate.