Using Large Language Models for Extracting Structured Information From Scientific Texts

Extracting structured information from scientificworks is challenging as sought parameters or properties areoften scattered across lengthy texts. We introduce a novel it-erative approach using Large Language Models (LLMs) toautomate this process. Our method first condenses scientificliterature, preserving essential information in a dense format,then retrieves predefined attributes. As a biomedical applica-tion example, our concept is employed to extract experimentalparameters for preparing Metal-Organic Frameworks (MOFs)from scientific work to enable complex and information-richapplications in the biotechnology-oriented life sciences. Ouropen-source method automates extracting information fromverbose texts, converting them into structured and easily navi-gable data. This considerably improves scientific literature re-search by utilizing the power of LLMs and paves the way forenhanced and faster information extraction from extensive sci-entific texts.