基于多层次融合的语句相似度计算模型  

The Computation Model of Sentence-Similarity Based on Multi-level Fusion

作　　者：南铉国[1] 崔荣一[1] 

机构地区：[1]延边大学工学院计算机科学与技术系智能信息处理研究室,吉林延吉133002

出　　处：《延边大学学报（自然科学版）》2007年第3期191-194,共4页Journal of Yanbian University（Natural Science Edition）

摘　　要：句子相似度的计算在自然语言处理的各个领域都占有很重要的地位,且在主观题自动批改技术中,句子间相似度的计算是一个关键的问题.本文提出一种基于多层次融合的语句相似度计算模型,在对句子的词形、词序、词义三个层次进行相似度计算的基础上,融合这三个层次,从而使句子相似度计算更加准确,并通过实验验证了此方法的有效性.The sentence-similarity holds the very important status in natural language disposal. And computing sentence-similarity between sentences is the key problem in marking subjective topic automatically. A model based on the multi-level sentence-similarity is proposed in this paper, to fuse those three levels which based on the sentence morphology, the word order, and the word meaning, so that the sentence-similarity to be more accurate, the reliability of the method is validated by the results of experiment.

关 键 词：语句相似度 自然语言处理 多层次融合 知网 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]