Mapping relational database schema to ontology is a critical task for ontology-based heterogeneous database integration. This paper proposes a hybrid approach combining element-level and structure-level techniques to discover the correspondences between relational database schema and ontology. In the process of element-level matching, string-based and Word Net sense-based techniques are combined to calculate the name similarity between elements. This paper analyzes different characteristics exhibited by matching element pairs of various types and describes how to use these characteristics to calculate the structure similarity. A series of experiments indicate that our approach performs better than the previous ones.
展开▼
