Agent-based learning of materials datasets from the scientific literature

Abstract

Advancements in machine learning and artificial intelligence are transforming the discovery of materials. While the vast corpus of scientific literature presents a valuable and rich resource of experimental data that can be used for training machine learning models, the availability and accessibility of these data remains a bottleneck. Accessing these data by manual dataset creation is limited due to issues in maintaining quality and consistency, scalability limitations, and the risk of human error and bias. Therefore, in this work, we develop a chemist AI agent, powered by large language models (LLMs), to overcome these limitations by autonomously creating structured datasets from natural language text, ranging from sentences and paragraphs to extensive scientific research articles and extract guidelines for designing materials with desired properties. Our chemist AI agent, Eunomia, can plan and execute actions by leveraging the existing knowledge from decades of scientific research articles, scientists, the Internet and other tools altogether. We benchmark the performance of our approach in three different information extraction tasks with various levels of complexity, including solid-state impurity doping, metal–organic framework (MOF) chemical formula, and property relationships. Our results demonstrate that our zero-shot agent, with the appropriate tools, is capable of attaining performance that is either superior or comparable to the state-of-the-art fine-tuned material information extraction methods. This approach simplifies compilation of machine learning-ready datasets for the applications of discovery of various materials, and significantly eases the accessibility of advanced natural language processing tools for novice users in natural language. The methodology in this work is developed as open-source software on https://github.com/AI4ChemS/Eunomia.