Chinese word automatic segmentation is the foundation of Chinese Information Processing, and it has widely application in many fields. In this paper, a new dictionary mechanism is presented: According to the Chinese characteristic of the high frequency of one word and two words we put forward such an idea that we can build up index table by the first two words as the keywords, and this index table is a two-dimensional array. This algorithm directly locates data by establishing a corresponding relationship between the first two Chinese characters' internal code. In this way, we can directly find out the two-word words by using the two-dimensional array. This approach can significantly reduce the times of queries, so as to further accelerate the speed of segmentation.