Iterative Chinese Bi-gram Term Extraction Using Machine- learning Classification Approach

This paper presents an iterative approach to extracting Chinese terms. Unlike the traditional approach to extracting Chinese terms, which requires the assistance of a dictionary, the proposed approach exploits the Support Vector Machine classifier which learns the extraction rules from the occurrences of a single popular term in the corpus. Additionally, we have designed a very effective feature set and a systematic approach for selecting the positive and negative samples as the source of training. An ancient Chinese corpus, Chinese Buddhist Texts, was taken as the experiment corpus. According to our experiment results, the proposed approach can achieve a very competitive result in comparison with the Chinese Knowledge and Information Processing (CKIP) system from Academia Sinica