Abstract: In the Chinese natural language processing,aimming at the problem that ordinary dictionary cannot be used for reverse maximum matching method and it is difficult to maintain a reverse dictionary,we put forward a new kind of dictionary structure and corresponding bi-direction maximum matching method,and added mutual information ambiguity processing block in the algorithm.Compared with the previous maximum matching method,this algorithm can increase the segmentation accuracy significantly.It is applicable to some Chinese natural language processing systems which have high segmentation accuracy requirement.

Key words: Segmentation dictionary,Bi-direction maximum matching method,Single word index based on Hash structure,Mutual information ambiguity processing