A Combining Approach for Chinese Word Segmentation

In Chinese and many other Asian languages which are based on non-ASCII alphabet, words are not delimited with whitespace (space, tab etc.), and word boundaries must therefore be reconstructed. Further syntactic analysis is based on the output of word segmentation result. Ambiguity and unregistered words are the most important problems in Chinese word segmentation. In this paper we analyzed the ambiguous reasons and presented a one-pass scan method for the detection and modification of ambiguous cases. To deal with the unregistered words and special words (such as names), we proposed a combination method that can recognize new words, hence the accuracy can be increased. In the realization, we used the bisection search method to look up words in a large dictionary (more than 40,000 items), and the average search cost for a word is less than 16 operations, so the speed is satisfactory if the system is embedded into Chinese understanding systems or Chinese speech processing systems.