Semantic conceptual primitives computing in text classification

Abstract

This paper presents a method for enhancing text classification performance with semantic computing. It adopts conceptual primitives with semantic relations as knowledge expression. Based on the semantic expression, it mined the association relation of primitives among different text classification, and these association rules take association relation as text classification feature. The presented method not only considers what kind of the semantic primitives that a text contains, but also takes account of the association relation of the semantic primitives. Moreover, we test the method with public text classification text set. The experiment result shows that, comparing with the commonly used methods, this method prompts text classification performance.