Research on Tibetan Automatic Word Segmentation

This paper researches on Tibetan automatic word segmentation. We focus on three key technologies of Tibetan automatic word segmentation: (1) a Tibetan automatic word segmentation approach is proposed, which is taking the advantage of case-auxiliary words and continuous feature. (2) a resolution method of overlapping ambiguity in Tibetan word segmentation is proposed, which is based on forward-backward scanning identification method and improved maximum probability algorithm. (3) through analyzing Tibetan name characteristics, an automatic recognition method of Tibetan name based on multi-features is proposed, which uses the internal features of names, contextual features and boundary features of names, and establishes the dictionary and feature base of Tibetan names. Finally, an experiment is conducted, and the results prove the methods are effective.