Automatic word segmentation technology is an important component part of modern Chinese information processing. It is the key technology of the Chinese full-text retrieval. This paper presents a Chinese word segmentation algorithm based on maximum entropy. It uses of part-of-speech tagging and word frequency tagging of corpus to establish maximum entropy model based on mutual information as a word segmentation language model to make word segmentation. At last, the binary model was used to test whether the expansion of the training corpus may impact the word segmentation accuracy, and the relationship curves between the expansion of training corpus and the word segmentation accuracy was obtained.
展开▼