Abstract: In this paper, a highly efficient data structure for Chinese thesaurus is introduced, which supports standard binary search and hashing operation by means of the first Chinese character in a string, while the length of every word is not limited. Then an improved fast algorithm for Chinese word segmentation is suggested. Based on searching a word composed of two characters quickly, the word including multiple Chinese characters can be found by utilizing the algorithm, which achieves high performance in Chinese word segmentation by invoking neighborhood matching. In theory, its time complexity is 1.66, which is superior to that of other algorithms for Chinese word segmentation.