Chinese word classification based on statistics

Chinese words classification based on statistics plays an important role in natural language processing, such as speech recognition, intelligent Chinese input method, and so on. We first do statistics and calculation work on the large-scale corpus text, and then use the average mutual information as the global cost function for clustering all Chinese words into a predefined number of classes with a hybrid top-down splitting and bottom-up merging approach. The result of classification is encouraging and can be used in the class-based language model