Abstract: Aho-Corasick automata algorithm is a famous multi-string matching algorithm, which backtracks to the effective subsequence state through the fail pointer when it fails in one pattern matching, where one or more effective subsequent states may exist. According to the above characteristics, this paper proposes an automata algorithm suitable for Chinese segmentation. The algorithm calculates the context matching probability of the current pattern by dynamic programming method, and backtracks to the optimal subsequent state of maximum probability, namely, it can realize the combination of the mechanical Chinese segmentation and statistics and probability model. The experimental result shows that a high accuracy rate in Chinese segmentation can be obtained through this algorithm.