Abstract: In order to improve text clustering performance, this paper proposes a text clustering method based on word hyperclique. It evaluates document similarity with word relationship between documents, works with word hyperclique as assistance of the document’s vector and uses a corresponding clustering algorithm by graph to partition the document sets. Experimental results validate the effectiveness of the algorithm for improving clustering performance.

Key words: text clustering, word hyperclique, clustering mode, feature selection

摘要： 为优化文本聚类效果，提出一种基于单词超团理论的文本聚类方法。利用文档中单词的关联模式来评估文档间的相似度，将单词超团作为文档向量辅助信息，以图划分的方式进行聚类分析。对不同聚类方法的结果进行比较，证明基于单词超团的文本聚类方法能提高文本聚类的准确性。

关键词: 文本聚类, 单词超团, 聚类模式, 特征选择

CLC Number: 

• TP391.1