Request Correction

User Profiling for Specification-Sensitive Recommendations with Large Language Model Prompting

Recently, there has been an increasing focus in research on the potential applications of large language models (LLMs) for personalized recommendations. Previous studies utilize LLMs to analyze the interaction between users and products to establish various personalized recommendation systems. However, recommendation becomes particularly challenging when items are associated with varied attributes, influenced by personal preferences, and described primarily through unstructured data. Moreover, analyzing implicit user preferences with product specifications for specification-sensitive recommendations remains largely unexplored. In this paper, we propose a framework that fully leverages prompting-based strategies to analyze user reviews and item attributes for the generation of user and product profiles, respectively. These profiles capture users’ implicit preferences and enable rating prediction or product recommendation, which are crucial for personalized recommendations. Experimental results show that our proposed framework effectively handles complex item attributes and user preferences to achieve promising performances in rating prediction.