一种基于向量词序的句子相似度算法研究

A Sentences Similarity Algorithm Based on Word Order of Vectors Distance

针对传统的句子相似度算法在句法结构等方面存在匹配率低的问题。为提高相似度计算的准确性,提出了一种改进的基于向量距离词序的句子相似度算法,从特征领域权重和词序方面进行改进,通过建立相应的领域特征集,对于相关领域的特征项给予更高的权重,同时,在向量空间模型的基础上,引入词序因子,计算句子词序相似度。使用包含6个领域的2651个句子作为语料库,实验结果表明,改进方法使特征领域内句子相似度计算的准确度得到提高。

Traditional sentence similarity algorithms cannot achieve high accuracy of similarity calculation due to their low matching rate in the syntactic structure. This paper proposes an improved sentence similarity algorithm based on the word order of vector distance to improve the accuracy of similarity calculation. It improves from the text feature -weight and Chinese word order calculation. With the establishment of corresponding domain feature set, the feature item in the domain will be given a higher weight. At the same time, on the basis of the vector space model, word order factor is introduced for the similarity calculation of Chinese word order. With the corpus containing six areas of 2651 sentences, the experimental result shows that the proposed algorithm can increase the accuracy of similarity cal- culation within the domain.