服装电商评论情感分析研究  

Sentiment analysis on electronic commerce reviews in the field of clothing

作　　者：李宏媛[1] 陶然 

机构地区：[1]东华大学计算机科学与技术学院,上海201600

出　　处：《智能计算机与应用》2017年第1期27-30,34,共5页Intelligent Computer and Applications

摘　　要：本文讨论了电商评论情感分析流程及其在服装电商评论中的应用。首先,提出一种电商评论情感分析流程;其次,对爬取到的服装电商评论进行分词和语义标注;然后,根据通用情感词库设计并实现了服装电商评论情感词典,并通过服装电商评论分词结果进行对比和完善;进而,基于语义计算规则,采用线性回归模型和支持向量机回归模型对服装电商评论情感值进行研究;最后,通过对比平均绝对误差、均方差误差、均方根误差研究评论的情感倾向与销量的关系。实验结果表明,语义情感强度值在支持向量机回归模型下可以以相对小的误差来预测销量,前3页评论对用户购买行为的影响较大。The sentiment analysis process of reviews on electronic commerce sites and its application on electronic commerce in the field of clothing are discussed by this paper. First of all, an sentiment analysis process is presented. Secondly, the reviews on electronic commerce sites in the field of clothing are segmented and annotated with semantics. Then, according to the general sentiment lexicon, the sentiment dictionaries of electronic commerce are designed and implemented. And it is improved through the results of word segmentation. Furthermore, based on the semantic computing rules, the linear regression model and the support vector machine regression model are used to study the sentiment value of the reviews on electronic commerce in the field of clothing. Finally, the relation between the sentiment tendencies and sales are discussed by comparing the mean absolute error, mean square error and root mean square error. The simulation results show that the support vector machine model in the case of small samples with good results. Meanwhile, the first three pages have a great influence on the user purchase behavior.

关 键 词：情感分析 电商评论 情感词典 语义规则 支持向量机回归 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]