A New Method to Compute Chinese Text Concept

A new method to compute Chinese text concept is proposed in this paper. In this method, we construct sentence vectors from the text by extracting and quantifying some syntax and semantic features such as concept elements, dependent relations and correlative relations. Then, we combine these sentence vectors to the text vector to represent the text concept. Experimental results show that, in the application of text classification, the precision and the recall of this method achieve 91.5% and 91.3%. Contrast with TF-IDF and LSA, the proposed method is more accurate, less affected by text class and more stable on text concept approaching.