翻译记忆中语句相似度计算方法的研究  

Research of calculating sentential similarity in translation memory

作　　者：周文[1] 徐国梁[1] 

机构地区：[1]湖南大学计算机与通信学院,湖南长沙410082

出　　处：《计算机应用》2007年第5期1210-1213,共4页journal of Computer Applications

摘　　要：对翻译记忆的关键技术——语句相似度的几种常用计算方法进行了分析,针对英汉翻译记忆系统的要求,提出了基于块的方法计算英语句子相似度,通过将英语句子按照句型分类存储,组合句子各个组成块的语义相似度与权重来得到整个句子的相似度。结果证明,它能很好的反映句子在语义和结构上的相似程度。Several commonly-used calculating methods of the sentential similarity, one of the key technologies of Translation Memory, were analyzed. According to the requirements of Engllsh-Chlnese translation memory system, a new method based on sentential form and semantics was proposed. The method stored the sentences on the basis of sentential forms, and obtained the sentential similarity with combining the semantic similarity and the weight of the component of the sentences. It can display the degree of semantic and structure between two English sentences commendably.

关 键 词：编辑距离 词语相似度 句型 权重 翻译记忆 

分 类 号：TP391.2[自动化与计算机技术—计算机应用技术]