基于褒贬倾向的句子相似度计算

Sentence Similarity Computation Based on Appraisable Orientation

首先简单介绍了基于《知网》的词语相似度的计算,指出不足,并对其进行改进,在其计算过程中增加词语褒贬倾向因素.接着给出句子相似度计算步骤:①先分词;②采用匈牙利算法求解最优匹配词语;③用改进的方法计算词语相似度,进而求解句子相似度.最后给出实验结果.结果表明:在词语相似度的计算上考虑其褒贬倾向,将会使计算结果更加合理.

Word similarity computation based on How-Net is introduced briefly.Its shortcoming is pointed out and improved by adding word appraisable orientation in the computing procedure.Secondly,the calculation steps of sentence similarity are discussed： ①split Chinese words;②match the best words by using Hungarian algorithm;③use the improved method to calculate word similarity,and then sentence similarity can be calculated according to the result.At last,the experimental results are given out.It shows that it will be more logical to take appraisable orientation into consideration while calculating the word similarity.