Research on the Construction of an E-Commerce Discipline Knowledge Graph Based on Multi-Source Heterogeneous Data Fusion

Addressing the fragmentation and unstructured nature of e-commerce discipline knowledge, this study proposes a framework for constructing an E-commerce Discipline Knowledge Graph (EDKG) based on multi-source heterogeneous data fusion. The framework establishes a schema layer grounded in professional and industry standards and employs the DeepKE model for entity recognition, relationship extraction, and data fusion to form the data layer. The final EDKG consists of 627 entities and 2,577 relationships. Experimental results demonstrate that the model excels in prediction accuracy and correctness.