摘要： 本文在分析了现有切词方法和汉语特点的基础上,提出一种具有多知识支持的分词方法SSK。SSK方法采用一种分层结构的词典,使每个词在匹配过程中能自动生成其所有可能的重切词,使切词失败时的歧义处理变得十分简单、有效。SSK方法不但得到字、词层次上知识的支持,且得到语法、语义知识的支持。该方法通过语法、语义检查可及时剔除一些切分错误,减少了歧义切分,且SSK方法具有简单的词汇学习功能,提高了切词正确率。

关键词: 切分, 自然语言, 语法, 歧义

Abstract: Having analysed the present segmenting methods and the peculiarity of Chinese, this paper presents a new segmenting method supported by multi-knowledge, SSX, which uses a dictionary of hierarchical structure to get all the re-segmented words of a sentence in case of unexpected ambiguity. Main attributes of this method are easy and effective ambiguity processing, less wrong segmentation by checking grammar and semantics, and elementary capability of learning new words.