This paper analyzed the "non-standardization" phenomenon of syndrome differentiation in traditional Chinese medicine （TCM） from the diagnosis process. It is proposed that the symptoms and signs collected by the four examinations of inspection， listening/smelling， inquiry， and palpation naturally have a certain "ambiguity"， which can be reduced by the comparison and comprehensive condensation （comprehensive analysis of the four examinations） of a large amount of multi-dimensional clinical data， thereby realizing the sublimation of TCM diagnosis from "ambiguity" of four examinations to "accuracy" of diagnostic conclusion. Based on the above assumptions， this paper further proposed that a research idea of intelligent syndrome differentiation in TCM， that is， by taking the clinical thinking ability of TCM physicians as the core， adopting artificial intelligence technology based on knowledge graph visualization， integrating the complex network association and reasoning method of "symptom-pathogenesis-syndrome" linked by pathogenesis， and through the automatic analysis and reasoning process of a large amount of multi-dimensional and ambiguous clinical data， the intelligent TCM diagnosis based on syndrome differentiation can be realized.