WAF-based document clustering algorithm

This paper proposes a novel document cluster algorithm based on Word Activation Forces (WAFs), a type of newly presented statistics.. A matrix of WAFs captures the information of terms occurrence and co-occurrence in a document, reflecting the underlying semantics that have not ever been considered from the current document representations. Its main consideration is that the same word in different documents may form disparate relation net which can be used to gain the similarities of the documents. Experimental evaluations on the dataset of the CLP2010 show that our proposed method is efficient and accurate for documents clustering.