About this Abstract |
| Meeting |
MS&T25: Materials Science & Technology
|
| Symposium
|
Additive Manufacturing Modeling, Simulation, and Machine Learning: Microstructure, Mechanics, and Process
|
| Presentation Title |
LLMs for Automated Data Extraction: A Case Study on AI’s Applications to Accelerate Meta Analysis for Cold Spraying |
| Author(s) |
Stephen Price, James Edward Saal, Marco Musto, Kyle Tsaknopolous, Kenneth Kroelein, Danielle L. Cote |
| On-Site Speaker (Planned) |
Stephen Price |
| Abstract Scope |
This work includes an end-to-end study on how large language models (LLMs) can accelerate or, in some cases, automate steps of the data extraction process, here using cold spray additive manufacturing (CSAM) as a case study. This includes accelerating the discovery and acquisition of published works, the generation of a dataset schema to extract, assisting in merging contradictory document annotations, and conducting automated extraction. With an extensive hand-labeled dataset to serve as the ground truth, this work demonstrates how different data types can be stored and effectively compared, as well as how various LLMs perform. For CSAM, this process has been used to create the largest dataset relating experimental parameters (material, system, heat treatment, etc.) to mechanical properties (strength, ductility, etc.), enabling more advanced optimization processes more efficiently. |