OntoBook Method for Ontology-Grounded Medical Encoder Pretraining
July 22, 2026
OntoBook converts medical ontology graphs into synthetic textbook-style prose via LLMs to train medical encoders. Training a 149M-parameter French encoder on this data yielded a +8.0 micro-F1 improvement on the Distemist-FR benchmark.
HOW THIS AFFECTS YOU
●
builderYou can use ontology-to-text pipelines to generate high-quality synthetic pretraining data for domain-specific encoders.
●
healthThis improves the accuracy of medical coding and relation prediction in specialized language models.