Categorization of scientific publications using large language models as applied to atomic spectroscopy

We present an automated framework employing zero-temperature large language models to categorize atomic spectroscopy literature for NIST bibliographic databases. The system processes native PDFs and supplementary materials via instruction-driven reasoning agents to assign structured search keywords encompassing target species, subject codes, and experimental methodologies. Validation against human-curated datasets demonstrates high precision in elemental species identification, while establishing that granular subject code assignment remains the primary source of categorization error.