Automatic Chinese Term Extraction based on Cognition Theory

Term extraction is a basic research topic to establish knowledge bases. This paper puts forward a new automatic Chinese term extraction based on cognition theory. Supervised by both linguistic knowledge and statistics information of research papers, we improve the traditional fair SCP and C-Value measures originally developed for multi-words, and then present a new comprehensive metric called MC-SCP measure, which also fit for uni-words, to examine "unithood" and "termhood" of candidate terms. The experimental results show that our method is expected even for terms with low frequency.