Knowledge representation is the premise and foundation of knowledge organization, since any method of knowledge organization should be based on knowledge representation. Traditional knowledge representation methods are mainly symbol representation, including logic representation, generative rule representation, structure representation, semantic network, ontology representation, etc, which are not suitable for medical knowledge characterized by various terms, scattered knowledge, complex knowledge and uneven quality. iFLYHealth adopts the super-large pre-training model to discover the deep semantic information in the massive high-quality medical texts, such as clinical paths, clinical guides, medical textbooks, drug instructions, etc., laying a good foundation for knowledge automatic construction and rule reasoning.