Research of Feature Weighting Method Based on Document Structure

Abstract

Documents in different categories can be represented as the same vector, relative position relationship among features in the document is considered in the paper on the basis of analyzing commonlyused document feature weighting methods aiming at the problem, and document structure matrix DS is introduced. DS is combined with three commonly-used weight algorithms for conforming three new models. The six models are utilized for classification experiment on actual corpus. Experimental results show that compared with original weight algorithm, DS-based weight algorithm can improve text classification effect.