LLM-Assisted Extraction of Agronomic and Ecological Experiments into Structured Data

An LLM-assisted, human-supervised pipeline to extract, interpret, and reconstruct agronomic and ecological experiment data from scientific papers into structured, BETYdb-compatible records. The system uses an intermediate representation (IR) with field-level provenance (extracted/inferred/unresolved), deterministic validation, and experimental reconstruction logic to ensure accuracy, consistency, and traceability, significantly reducing manual effort and improving data quality and scalability.