Waste-to-Resource Database Construction for Industrial Symbiosis Using Large Language Models

Industrial Symbiosis (IS) is a subfield in Circular Economy (CE), which aims to optimize resource utilization by transforming wastes into resources among industries, hence reducing waste generation and lowering raw material costs. Identifying waste-to-resource (W2R) pathways is crucial for establishing IS relationships. Existing W2R databases are constructed from IS cases happening around the world, which fail to capture new W2R pathways that are published as papers or patents. Furthermore, extracting W2R information previously requires extensive manual efforts, demanding both time and specialized knowledge. To address these problems, we propose a novel framework that automatically constructs the W2R database with knowledge extracted from published papers. Specifically, we leverage Large Language Models (LLMs) and Retrieval Augmented Generation (RAG) technique, enabling automatic W2R knowledge extraction and self-evaluation. We further demonstrate its application in recommending potential IS opportunities through a case study. Compared to existing static W2R databases, our method provides a dynamic approach to encode emerging W2R knowledge and enable the identification of novel IS opportunities.