A New Method of Computing Chinese Word Similarity Based on Statistics

Word semantic similarity is a very subjective concept and it is very difficult to get a similarity value close to human judgment. Chinese word semantic similarity research is relatively scarce due to its inherent complexity. This paper presents an approach to compute Chinese word semantic similarity based on statistical methods with word frequency contrast introduced (WFC-WS). Word semantic vectors are first obtained using co-occurrence and then extended with HIT-IR Tongyici Cilin (Extended). Word frequency contrast is introduced to filter the semantic vectors. Experiments show that the results of WFC-WS are closer to artificial standard compared with some similar methods.