The digital and intelligent transformation presents significant opportunities for the modernization and intelligent servicing of Traditional Chinese Medicine (TCM) knowledge. However, the inherent characteristics of the TCM knowledge system—such as ambiguous terminology, multi-source heterogeneity, and semantic complexity—pose considerable challenges to its structured organization and intelligent application. This paper constructs a human-in-the-loop TCM semantic knowledge organization and service framework based on a "model generation—human correction—feedback optimization" paradigm. By integrating the automated extraction capabilities of large language models with the manual verification mechanisms of domain experts, the framework achieves precise identification and dynamic optimization of entities and relationships within TCM literature. Using the prevention and treatment of gastric diseases in TCM as a case study, the framework's application potential is validated in areas such as assisting clinical diagnosis and treatment, enabling personalized medication recommendations, and supporting proactive health management. This work contributes to the systematic transformation of TCM knowledge, its intelligent service delivery, and its deeper integration with modern medicine.