Database Schema Matching Model Based on Multimodal Fusion Analysis

Automated schema matching for multi-source heterogeneous databases can effectively promote data integration and interoperability, enhance data quality, support data migration and transformation, and improve data management and governance. To address the schema matching problem in databases and further improve matching accuracy, this paper proposes a database schema matching model based on multimodal fusion analysis (MFA). A data relationship graph is constructed based on database metadata information. For the nodes in the graph, text embedding features and structural embedding features are constructed separately, and a comprehensive representation feature of the nodes is obtained through fusion. Based on the node representation features, a multi-round matching model based on Jaccard similarity is constructed to calculate the similarity between two nodes, achieving automatic matching of the correspondence between source attributes and target attributes in database schema matching. Experimental results show that the proposed model has higher matching accuracy compared to existing methods.