Evaluating model outputs by comparing them against a structured knowledge base of medical concepts and relationships.