语义检索中的词语相似度计算研究

Research of Word Similarity Computing in Semantic Retrieval

词语相似度计算是信息处理技术中的一个关键问题,主要研究用什么样的方法来计算或比较两个词语的相似性。词语相似度计算可以从广义与狭义两方面考虑,文中从狭义的角度对词语相似度计算进行了研究,通过对语素相似度、字序相似度和词长相似度的分析,提出了一种基于语素、字序和词长的综合词语相似度计算方法。通过词语相似度计算可以修正用户提问中错字、漏字、多字等情况,并且通过实验给出了词语相似度计算中参数的参考值。实验结果验证了该计算方法的有效性。

Word similarity computing is a crucial question in information processing technology.It mainly researches what kind of method to calculate or compare the similarity between two words.The word similarity computing can be considered from broad aspect and narrow aspect.Researched word similarity computing from narrow aspect,an integrated word similarity calculation method is proposed by analyzing morpheme＇s similarity,word order＇s similarity and word length＇s similarity.It can revise the following conditions： incorrect word,missing word and extra word.The parameters of the method were decided by experiments.Experiments show that this method has high efficiency.