[3R5-OS-13c] OS-13

[3R5-OS-13c-02] Supporting User Review Creation with a Dialogue System

〇Yoshiki Tanaka¹, Michimasa Inaba¹ (1. The University of Electro-Communications)

Keywords:dialogue system, natural language processing, interview

User reviews posted on e-commerce and review sites help in making purchase decisions. However, it takes time and effort to create detailed reviews. In this study, we propose a novel utilization of dialogue systems for supporting user review creation. In the proposed method, the dialogue system initially interviews the user to elicit their impressions and requests for the product. Subsequently, the system generates a review based on the dialogue history. The proposed method can reduce the burden on the user in creating a review, in that the user only needs to interact with the system. Moreover, the system could generate helpful reviews by eliciting the appropriate balance of positive and negative aspects. We implemented the system using GPT-4. The results show that the subjects positively rated their enjoyment of interacting with the system, its quality, and their satisfaction with the generated reviews.