The article discusses how large language models (LLMs), despite their exceptional applicative potential, have not been effectively used in mining relationships from graph data. The authors propose a framework that combines the capabilities of LLMs with the relationship extraction and analysis functions of graph neural networks. This new approach aims to improve the understanding of graph relationships by LLMs, thereby providing more accurate, comprehensive, and personalized recommendations. The proposed framework has been evaluated on real-world datasets, demonstrating its ability to understand connectivity information in graph data and improve the quality of recommendation results.
Publication date: 16 Feb 2024
Project Page: https://github.com/anord-wang/LLM4REC.git
Paper: https://arxiv.org/pdf/2402.09617
