Unifying Star Ratings and Text Reviews in Linguistic Terms for Product Competitiveness Analysis Based on Stochastic Dominance
Huchang Liao, Jiayi Wang, Zeshui Xu · IEEE Transactions on Computational Social Systems · 2023
Online reviews, mainly consisted of star ratings and text reviews, provide enterprises with consumer feedback on product usage. From the perspective of a product designer, exploring the values of online reviews to help enterprises analyze product competitiveness and grasp market trends becomes necessary. However, heterogeneous and unstructured online reviews can confuse them. Previous research on online reviews mainly focused on comparing the different usefulness between star ratings and text reviews, whereas the research on combining these two types of information to compare product performances was relatively limited. In addition, the unified expression of heterogeneous star ratings and text reviews is the difficulty in information aggregation, and simple expressions may ignore differences in consumers’ psychological cognition. To solve these challenges, this study introduces the prospect theory to unify star ratings and text reviews of products and then takes two heterogeneous information as the basis for product designers analyzing competitiveness. First, we propose a unification model to express different grades of star ratings and text reviews as a unified evaluation system according to value consistency. Then, we use evidential reasoning (ER) theory to aggregate unified product evaluation results. According to the confidence distribution of evaluation results, stochastic dominance (SD) rules are used to determine the dominance relations of products under each criterion. Finally, comprehensive comparisons of products are carried out based on the characteristics of dominance relations of products. The applicability of the proposed method is illustrated by a case study of online review-based automobiles comparison on Autohome.com.cn.