Abstract: Studies on product design improvement of online reviews have attracted considerable attention.However,existing studies still have several limitations such as ignoring the competitive products analysis and unable to acquire fine-grained customers` concerns.Therefore,an approach for conducting Importance Performance competitor Analysis (IPCA) based on product online reviews was proposed.Product topics and their document-topic probability distribution were determined with Latent Dirichlet Allocation (LDA) topic model and the importance levels of product features were calculated.Sentiment labels were extracted through dependency syntax analysis and were quantitatively expressed in performance levels.On this basis,multiple IPCA models were constructed to find target product features to be improved.The use cases were extracted from the original negative reviews to discover the fine-grained customer requirements and the product problems,and then the product design improvement strategies were developed.A case study of the word-of-mouth data in Autohome Website was provided to verify the effectiveness of the proposed method.

Key words: online reviews, importance performance competitor analysis, product design improvement, latent Dirichlet allocation topic model, sentiment analysis

摘要： 基于在线评论的产品设计改进研究,受到越来越多学者的关注。然而,现有研究仍有很多局限,如忽略竞争产品分析以及无法获取细粒度顾客需求等。为解决上述问题,提出一种基于在线评论的重要度绩效竞争对手分析方法(IPCA)。首先,应用隐含狄利克雷分布(LDA)主题模型识别产品主题,基于文档—主题概率分布计算产品特征重要度;其次,通过依存句法分析提取情感标签,并将其量化为产品特征绩效值;然后,构建多个IPCA模型找到目标产品的待改进特征;最后,从原始负面评论中提取用例,获取顾客细粒度需求并发现产品问题,进而提出产品设计改进策略。以汽车之家口碑数据为例,验证了所提方法的有效性。

关键词: 在线评论, 重要度绩效竞争对手分析, 产品设计改进, LDA主题模型, 情感分析