The team responsible for BioCypher accomplished precisely this by taking a big corpus of medical research papers, building a large language model around them, and then deriving a knowledge graph ...