摘要： 提出了一种基于框架的词语搭配抽取方法,可以同时获取词语搭配以及搭配结构信息。引入相对词序比(RRWR)的方法对候选搭配词语进行筛选,应用语言学中词语搭配组合规律对候选搭配的词性进行限定,利用互信息等统计学模型在大规模语料中进行词语搭配的自动抽取,抽取的搭配平均准确率为84.73%,较Xtract系统高4.7%,较国内同类工作结果高50.79%。并且在获得搭配的同时得到了词语搭配的结构信息。

关键词: 词语搭配, 搭配框架, 互信息, 相对词序比

Abstract: This paper presents a frame-based method to extract Chinese collocations and their structure information simultaneously. It uses ratio of relative word rank(RRWR) to pre-process the potential collocations, applies linguistic knowledge to restrict the part of speech of the candidate collocations, makes use of statistical language model such as mutual information to extract collocations in large scale corpus. The result shows that the average accuracy is 84.73%, 4.7% higher than Xtract, 50.79% higher than the work of same kind in Chinese collocation extraction. And the structure information is acquired at the same time.

中图分类号: 

• H085