Presentation Information

[18a-PA1-7]Automatic Extraction and Structuring of Equations and Descriptors in Materials Science Papers

Chie Suematsu1, 〇Toshihisa Anazawa1, Takuya Kadohira1, Satoshi Minamoto1 (1.NIMS)

Keywords:

Equation extraction,Materials literature,Graph database

Technical documents in the field of materials science contain many equations. However, information such as the meanings and units of variables is often described in a non-structured manner, making cross-literature utilization difficult. In this study, equations and descriptors were automatically extracted and structured from adsorption materials papers using large language models, and the equations were represented in a numerically computable SymPy format and organized into a graph database, thereby establishing a basis for the cross-literature reference and reuse of knowledge derived from equations.