Researchers have developed HERMES, a multi-agent framework designed to extract structured data from lengthy scientific documents, particularly in geoscience. This system utilizes a coordinating large language model to integrate domain constraints, validation rules, and evidence tracing, processing text, tables, figures, and captions. When applied to the 55-volume Treatise on Invertebrate Paleontology, HERMES successfully created a structured database of over 32,000 fossil taxonomic entities and 450,000 attributes, achieving high F1 scores for extraction accuracy and significantly improving efficiency compared to manual methods. The framework also demonstrated its adaptability by performing well in palaeomagnetism and geochemistry without additional training, offering a path to convert historical scientific literature into FAIR-compliant structured data. AI
IMPACT Enables efficient conversion of historical scientific literature into structured, FAIR-compliant data for broader accessibility and integration.
RANK_REASON The item is a research paper detailing a new framework for knowledge extraction. [lever_c_demoted from research: ic=1 ai=1.0]
- alphaXiv
- arXivLabs
- CatalyzeX Code Finder for Papers
- CORE Recommender
- DagsHub
- Gotit.pub
- HERMES
- Hugging Face
- Influence Flower
- large language model
- ScienceCast
- Treatise on Invertebrate Paleontology
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
