Mining online reviews by deep learning-based UIE-ERNIE for AI-empowered live streaming product selection

In recent years, live streaming selling has experienced rapid growth, and become a widely adopted online sales way. The selection of products for live streaming is critical to attracting and retaining customers, thereby enhancing the reputation of live streaming rooms. However, given the frequent concerns about product quality in live streaming, a product selection approach that incorporates consumer preferences and the characteristics of live streaming room is essential. To address this challenge, this study introduces an AI-empowered large-scale group decision making (LSGDM) approach for live streaming product selection. Firstly, online reviews about live streaming products from e-commerce platforms, such as “JD.com†and “Taobao†, are collected by utilizing the Octopus crawler software. Then, a deep learning-based Universal Information Extraction with ERNIE (UIE-ERNIE) is firstly introduced to mine online reviews, which can automatically identify product attributes that consumers care about, and can clearly classify sentiments in the reviews. Furthermore, linguistic hesitant-Z-numbers (LHZNs) are employed to concisely represent evaluation information. Finally, an online reviews-driven case study is formulated to illustrate the applicability of the proposed method. Compared with the LDA and TF-IDF, mining reviews by UIE-ERNIE offers higher accuracy and efficiency without complex preprocessing. Sensitivity analysis is performed to explore the effects of the number of experts, online reviews and consensus threshold for the decision process. Comparative analysis indicates that the LHZNs significantly improve consensus degree. Overall, this study proposes an AI-empowered approach for live streaming product selection, combining real customer online reviews with expert evaluations to support decision-making.