Resolution to Combinational Ambiguity of Chinese Word Segmentation

摘要

Chinese word segmentation ambiguity can be divided into two categories: overlapped ambiguity and combinational ambiguity. This paper only focuses on the resolution to combinational ambiguity of Chinese word segmentation. We select 36 typical combinational ambiguity strings, and make use of transformation-based learning methods to learn the rules of combinational ambiguity. Using these rules to test "People''s Daily" Corpus of 1996, we find that the average precision rate is improved from 79.08% to 94.35%.