A Novel Vector Representation Model for Text Mining Based on Enhancing Features

Abstract

Semantics relationships such as polysemy and synonymy etc. contained in text documents can be used to enhance the features in the vector representation. Previous feature representation methods rarely consider such semantic relations to improve mining precision, because taking semantic relations into account greatly depends on the learning process. In this paper, we present a novel document representation method which utilizes semantic relations to enhance the features. To obtain semantic relations, we use the external resource of Wikipedia. Moreover, a term-concept mapping structure is constructed to describe features. When concepts are extracted from document, Markov random walk model is used to handle the semantic relationships between the concepts and related concepts are voted to enhance the features. We conduct comprehensive experiment using 20-newsgroup corpus and Reuters corpus. Results show that the proposed method has good performance in text mining.