Segmentation Ambiguity is an important factor influencing accuracy of Chinese auto-segmentation system. Time words include expressions both indicating exact time positions and those scattering in a treriod of time. On the foundations of modern Chinese corpus processing principles and certain type time word segmentation ambiguity, this paper proposes problem, a statistical language model and corresponding approach based on maximum likelihood to solve the ambiguous and it reaches a 90% accuracy which shows the effectiveness of the algorithm.