Presentation Information
[10a-N304-9]Materials Development with LLM (VII): Autonomous Polymer Entity Linking Using PoLyInfo Molecular Structure Taxonomy and Agent AI
〇Masashi Ishii1, Koichi Sakamoto1 (1.NIMS)
Keywords:
PoLyInfo,Agent AI,Database
By combining the PoLyInfo molecular structure taxonomy with an LLM, we achieved autonomous entity linking between polymer names and PoLyInfo polymer IDs (PIDs). Using GPT-OSS:120B and ReAct, we had the system infer the optimal synthetic polymers that constitute composite materials with spider silk. As the most chemically plausible structures, it identified the PIDs of two types of block copolymers: polyether-based polyurethane and polyether-based polyurea.
