The overall problem addressed in this paper is to improve semantic interoperability in heterogeneous systems. Normally, the semantics of data is carried by ontology (concept model). Reconciling data semantics therefore boils down to reconciling relevant ontologies. A candidate solution is to use extensional information, i.e. the instance information of the ontology to enrich the ontology and further, based on the enrichment structure to calculate similarities between concepts in two ontologies.