摘要： 为更全面准确地从字词搭配中获取语义搭配信息，提出一种新的语义搭配知识提取模型和知识库的表示方法。利用特殊度度量词语搭配的相关程度，在此基础上，基于知网的语义信息，实现从42万条记录的词语搭配中定量地抽取语义搭配信息。实验结果表明，该方法的语义搭配准确率为92.1%，且较大地扩充了字词搭配的规模。

关键词: 词语搭配, 语义搭配, 特殊度, 知识获取, 知识表示, 搭配知识库

Abstract: In order to achieve semantic collocation from word collocation, this paper proposes a new model of extracting semantic collocation and a new representation method of semantic collocation. By introducing Special Degree(SD) as a tool and utilizing knowledge of the HowNet, 420 000 records of words collocation are converted to semantic collocation in a quantitative way. Experimental results show that the accuracy of semantic collocation reaches 92.1%, and the semantic collocation expands the scale of the words collocation effectively.

Key words: word collocation, semantic collocation, Special Degree(SD), knowledge acquisition, knowledge representation, collocation knowledgebase

中图分类号: 

• TP391