A Framework for Ontology-Based Data Integration

This paper presents an ontology-based data integration framework that is capable of deriving an ontology from a collection of XML schemas in a semiautomatic manner and integrating heterogeneous XML sources at the semantic level. The ontology in our system is constructed following a layered approach where an intermediate model is introduced to explicate the underlying semantics of XML schemas and to reduce the complexity of ontology derivation. This two-phase approach is performed semi-automatically by applying a set of heuristic rules and by interpreting mapping information defined by users. The resulting ontology serves as a global semantic view over a set of data sources to be integrated. Moreover, we also adopt a data warehousing approach to populate this ontology with data from XML instance documents automatically. The proposed framework has been implemented and evaluated using off the-Web XML schemas.