A general LLM-powered text mining framework: Applied to extract high entropy alloys

Abstract

In this paper, we present a general framework for automating the information extraction process from materials science literature. Our aim is to meet the increasing demand for large-scale databases in both research and engineering. The text mining part consists of three continuous stages: labeling, extraction, and post-processing, which are all powered by large language models (LLMs). Through these successive stages, the framework enables the extraction of material data from both text and tables. It supports the generation of high-quality databases with only a moderate level of prior knowledge about the extraction targets and minimal coding effort, thereby facilitating the rapid development ..