汉语口语语段边界的界定

UTTERANCE SEGMENTATION OF SPOKEN CHINESE

没有明确的句子边界导致口语分析无法像书面语那样以句子为单位进行．为此，文中在汉语口语分析中引入话轮和语段的概念，并提出正反向精简循环网络和规则相结合的汉语口语段边界界定方法．该方法全面利用语法、语义和词的个性特征，综合上文和下文信息。

Owing to the absence of clearly marked sentence boundaries, spoken language cannot be analyzed by sentence as in written language analysis. To solve this problem, this paper introduces “dialog turn” and “utterance” into spoken Chinese analysis and further proposes a method of integrating FBSRN with the rules proposed for Chinese utterance segmentation. The fine grained rule is used first to tag a segmented or unsegmented sign. Then, FBSRN is used to further segment according to syntactic and semantic dependent information in context. The overall performance is satisfying.