Generating product reviews from aspect-based ratings using large language models

Abstract

The rapid growth of e-commerce has made textual reviews and product ratings crucial for consumer purchase decisions. However, the overall Likert scale rating of the product does not convey any information about major aspects of a product. In contrast, many textual reviews often lack detailing of various aspects of the product, leading to incomplete feedback. This paper proposes a framework that generates detailed textual reviews from user-provided ratings on various aspects of a product using large language models (LLMs). Our approach enhances the online product review system by integrating specific feedback from structured ratings, resulting in more detailed and reliable product reviews. Our results show that AI-generated reviews exhibit high readability, coherence, relevance, and informativeness, rivaling human-written reviews to the extent that distinguishing between the two proves challenging, even for human evaluators. This research contributes to develop more accurate and comprehensive review systems, enhancing the overall quality and usefulness of e-commerce reviews and empowering consumers to make informed purchasing decisions. The proposed framework offers a valuable tool for businesses and e-commerce platforms to improve product reviews, enhance customer satisfaction, and increase sales.