Abstract: Customer satisfaction is the essential driven force of product research and development.The information of competitive products is rarely involved in the existing research of customer satisfaction.By incorporating the online user-generated content such as “online comparative Questions & Answers (Q&A)” and “online reviews”,a comprehensive customer satisfaction measurement method considering competitive products was proposed.The competitive products were identified by mining comparative questions in Q&A,the comparative score between products was calculated based on sentiment analysis of comparative answers,according to which,the comparative network was constructed,and the PageRank algorithm was performed to calculate the Competitive Customer Satisfaction (CCS).The Latent Dirichlet Allocation (LDA) topic elicitation method and sentiment analysis were combined to induce the Basic Customer Satisfaction (BCS) from online reviews.Finally,the Integrated Customer Satisfaction (ICS) was obtained by merging the predetermined CCS and BCS.The customer satisfaction measurement of Redmi K50 was established as the case study.The results showed that there was a relatively high consistency between the obtained ICS and sales of Redmi K50,which could effectively support decision-making for product improvement.

Key words: customer satisfaction, competitive products, online user-generated content, sentiment analysis, Latent Dirichlet allocation topic elicitation

摘要： 产品研发的根本驱动力是实现顾客满意,少有研究结合竞品信息探究产品的顾客满意度。综合“比较问答”和“在线评论”等在线用户生成信息,提出一种考虑竞品信息的综合顾客满意度建模方法。首先,通过挖掘在线问答中的比较问题识别竞品,基于比较回答的情感分析计算产品间的比较得分,构建比较网络,利用PageRank算法计算竞争顾客满意度。其次,运用LDA主题提取和情感分析从在线评论中提取基础顾客满意度。最后,结合竞争满意度和基础满意度获得综合顾客满意度。以红米K50的顾客满意度测度为例进行了案例分析,结果表明,综合顾客满意度结果与销量数据高度一致,可为产品改进提供决策依据。

关键词: 顾客满意度, 竞品, 在线用户生成信息, 情感分析, LDA主题提取