Ontology Graph Based Approach for Automatic Chinese Text Classification

Abstract

Automatic classification of Chinese text documents requires a machine to process and analyze the meaning of Chinese terms. We propose an Ontology Graph based approach to measure the relations between Chinese terms for the text classification purpose. The method improves traditional high dimensional termbased text classification approach, in that the new method selects very small number of semantically related concepts to create Ontology Graphs. The Ontology Graphs can be used to represent different classes (domains). It enhances text classification performance by using its small-size but high semantically associated concepts. Our experiments show that the proposed method has classified a Chinese document set with 92% accuracy in f-measure by using Ontology Graphs containing only 80 concepts for each class. The high accuracy result shows that the Ontology Graphs used in the process are enable to represent the knowledge of a domain and also the Ontology Graph based approach of text classification is effective and accurate.