Integrating Graph Retrieval-Augmented Generation With Large Language Models for Supplier Discovery

Abstract

As supply chain complexity and dynamism challenge traditional management approaches, integrating large language models (LLMs) and knowledge graphs (KGs) emerges as a promising method for advancing supply chain analytics. This article presents a methodology crafted to harness the synergies between LLMs and KGs, with a particular focus on enhancing supplier discovery practices. The primary goal is to transform and integrate a vast body of unstructured supplier capability data into a harmonized KG, thus improving the supplier discovery process and enhancing the accessibility and findability of manufacturing suppliers. Through an ontology-driven graph construction process, the presented methodology integrates KGs and retrieval-augmented generation with advanced LLM-based natural language processing techniques. With the aid of a detailed case study, we showcase how this integrated approach not only enhances the quality of answers and increases visibility for small- and medium-sized manufacturers but also amplifies agility and provides strategic insights into supply chain management.