About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Microstructure-Sensitive Modeling Across Length Scales: An MPMD/SMD Symposium in Honor of David L. McDowell
|
| Presentation Title |
Leveraging Large Language Models to Extract Composition, Processing, Microstructure, and Property Data of Metallic Materials From Literature |
| Author(s) |
Shuozhi Xu, Xin Wang, Anshu Raj, Haiming Wen, Kun Lu |
| On-Site Speaker (Planned) |
Shuozhi Xu |
| Abstract Scope |
Within the materials science domain, interest in large language models (LLMs) is growing at an unprecedented pace, largely due to the fact that the majority of materials-related knowledge exists in textual form, which aligns naturally with the capabilities of LLMs. To date, much of the work in this emerging subfield has centered on extracting compositional and property-related data from the literature, primarily in the context of functional materials like catalysts or semiconductors. However, in the case of structural materials such as metals, the situation is more complex because microstructure plays a central role in determining mechanical property. Accurately capturing this information is essential for understanding and designing advanced alloys. In this work, we explore the abilities of LLMs in extracting information including chemical compositions, processing conditions, microstructures, and mechanical properties, directly from published literature. Our findings demonstrate that, following effective prompt engineering, LLMs can well handle these diverse data types. |
| Proceedings Inclusion? |
Planned: |