Abstract: Based on the phenomenon of the full- and half-width forms of lettered words in Chinese language corpus, we investigated the influence of this phenomenon on the consistency and accuracy of lettered words in word segmentation results, and suggested how to improve the consistency and accuracy of word segmentation. We also introduced our work on constructing the word segmentation system sponsored by Communication University of China.

Key words: lettered words, term extraction, word segmentation, full- and half-width