A Large Language Model-based Framework to Retrieve Life Cycle Inventory and Environmental Impact Data from Scientific Literature

A systematic approach using a retrained large language model to retrieve life cycle inventory and environmental impact category datasets from scientific literature. The proposed framework, named Sustain-LLaMA, comprises three stages, as follows: Stage 1. Classification task: It helps to screen the relevant articles that contain life cycle studies in regard to the product. It is a fine-tuning step; the resulting model processed all extracted abstracts from the Elsevier database using an API key to get relevant abstracts. Stage 2. Pretraining: Inject the domain knowledge into the basic architecture of LLM using the relevant articles' text. Stage 3. Question and Answering Task: It is a fine-tuning step that makes the custom-pretrained LLM tailored to respond to queries for LCI and environmental data. This framework enhances scalability and precision by automating LCI data retrieval, offering a promising tool for guiding the chemical and plastic industries toward net-zero emissions.