基于多特征融合的产品评论语句相似度计算  

Similarity computation of product comment statements based on multi-feature fusion

作　　者：谭文斌[1,2] 张太华[1,2] 何二宝[1,2] 

机构地区：[1]贵州师范大学制造服务与知识工程技术工程中心,贵州贵阳550025 [2]贵阳市机械控制与仿真重点实验室,贵州贵阳550014

出　　处：《贵州师范大学学报（自然科学版）》2016年第1期81-87,120,共8页Journal of Guizhou Normal University：Natural Sciences

基　　金：国家自然科学基金资助项目(71061004;71161006);贵州省优秀青年人才培养对象专项项目(2013(37));贵州省教育厅人才团队项目(黔教合人才团队字[2015]58)

摘　　要：为了提高产品评论语句相似度计算的准确率,提出了一种新的基于多特征融合的相似度计算模型。在构建相似度计算模型时,抽取句子的名词、动词和形容词作为关键词构成句子的向量表示,并将关键词的频次信息及其修饰成分信息、句子的主谓结构和动宾结构等特征信息都纳入到向量模型当中,通过各特征要素的重要度来表征这些信息在句子相似度计算中的重要程度。实验结果表明,该计算模型提升了产品评论语句相似度计算的准确率,对提升产品评论语句的聚类分析能力具有重要意义。In order to improve the accuracy of the similarity calculation of product reviews statements,a new similarity computing model based on multi-feature fusion is proposed. In constructing the similarity computing model,the noun,verb,and adjective of the sentence are expressed as the keywords to constitute a sentence vector,and put the keywords frequency information and its modifier,the sentence of the subject-predicate structure,verb object structure and other feature information included in the vector model,the importance of this information in sentence similarity computation is characterized by the importance of the feature elements. The experimental results show that the proposed model can improve the accuracy rate of the similarity calculation of the product reviews statements,and it has important significance to improve the clustering analysis ability of the product reviews statements.

关 键 词：相似度计算 多特征融合 产品评论语句 句子向量模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]