融合依存句法与产品特征库的用户观点识别研究

Research on User Opinion Recognition Based on Dependency Syntax and Product Feature Thesaurus

[目的/意义]有效分析并利用电子商务网站用户评论数据,发掘海量用户评论数据的价值,识别用户关注的商品属性,在丰富用户行为学理论研究内容的同时,也为相关实践提供参考。[方法/过程]基于依存句法分析技术以及Word2Vec词向量技术构建的产品特征库进行在线评论用户观点的抽取,并通过引入依存词对的词性特征、依存关系组合特征和词汇距离约束等方法,提升用户观点抽取的精度和质量。[结果/结论]文章所提的基于依存句法和产品特征库的用户观点抽取方法相较于最近距离法和SBV极性传递法有更优的实验效果,在准确率、召回率和F1值上相较于两种基准方法均有较大的提升,证明了所提方法的有效性。

[Purpose/significance]Effectively analyzing and using the user review data of e-commerce website,exploring the value of massive user review data,identifying the commodity attributes of users’concern,enriching the theoretical research content of user behavior,and providing reference for related practice.[Method/process]The product feature thesaurus based on dependency parsing technology and word2vec technology is used to extract the user’s opinion.The precision and quality of user’s opinion extraction are improved by introducing the part of speech feature of dependency word pair,dependency combination feature and words distance constraint.[Result/conclusion]Compared with the nearest distance method and SBV polarity transfer method,the proposed user opinion extraction method based on dependency syntax and product feature thesaurus has better experimental results.Compared with the two benchmark methods in accuracy,recall rate and F1 value,the proposed method is effective.