我中心研究员吕晓玲、王菲菲及博士生郭昱璇（第一作者）在《Electronic Commerce Research and Applications》发表论文。该研究提出了多品牌用户评论联合主题模型。该模型能够同时分析多品牌评论语料，提取所有品牌的共有主题以及各个品牌的特有主题，并在每个主题中区分出评论特征及用户观点。该模型能够更好地分析竞争品牌的优势及劣势，在两个领域的评论数据集上实现了较好的挖掘效果。

论文题目

Mining multi-brand characteristics from online reviews for competitive analysis: A brand joint model using latent Dirichlet allocation

文章摘要

Online reviews reflect customers’opinions of the products or services they have bought. Analysing online reviews provides a reliable way for e-commerce platforms to understand users’ needs and attitudes. To understand the strengths and weaknesses of several competitive brands, we propose a brand joint latent Dirichlet allocation model to analyse multi corpora simultaneously, particularly online reviews for multi-brands. Our model is the first to capture the general aspects of the online opinions of users concerning multi-brands and specific aspects of user opinions within individual brands. The results can assist the analysis of competitive brands and make meaningful suggestions for brand managers and marketers. We present two case studies to prove the efficiency of the proposed model.

作者介绍

郭昱璇，中国人民大学统计学院在读博士生。研究方向：评论文本挖掘、统计学习等。

王菲菲，中国人民大学统计学院副教授。研究方向：文本挖掘及其商业应用、社交网络分析、大数据建模等。

邢晨，本硕就读于中国人民大学统计学院，现就职于中国银行总行。

吕晓玲，中国人民大学统计学院教授、应用统计科学研究中心研究员。研究方向：统计学习、消费者行为分析、文本分析等。

论文发表截图