Abstract
The inheritance of knowledge from Ancient Chinese Medicine Classics (ACMC) confronts challenges including fragmented literature, terminological heterogeneity, and reliance on traditional apprenticeship. Knowledge Graphs (KG) have become one of the tools for the digitalization and intelligentization of ACMC, playing a vital role in unifying terminology, standardizing data, and structuring and linking knowledge. However, due to the complexity of the ancient Chinese language in ACMC texts and the diversity of syndrome differentiation systems, current KG construction techniques still rely on manual input or traditional Natural Language Processing, with applications primarily limited to basic question-answering (Q&A) systems. Although large language models (LLMs) in the field of traditional Chinese medicine have incorporated ACMC corpora, automated extraction and intelligent integration within KG remain underdeveloped. This paper proposes an innovative approach that combines Multi-Agent Systems (MAS) with KG for advancing the intelligent application of ACMC. The technical approach involves using KG as the knowledge foundation, while leveraging MAS's LLM-based semantic understanding and collaborative task distribution to enable breakthroughs in triple extraction technology and to advance the intelligent applications of ACMC, including context-aware Q&A, herbal formula innovation, dynamic diagnosis and treatment, and personalized education. Additionally, the integration of Retrieval-Augmented Generation technology enables the dynamic synthesis of multi-source knowledge, resolves semantic ambiguities, and optimizes MAS decision-making. These discussions aim to inform the design of a high-fidelity, adaptive, and perception-driven autonomous system for the intelligent inheritance and innovation of ACMC.