The integration of data is one of the most complicated tasks which need to be addressed by IT researchers. Despite the critical importance, the current approaches to semantic interoperability of heterogeneous databases have not been sufficiently effective. The authors apply an idea of ontology as the foundation for data integration. In the paper the most common data integration methods and their essential features are discussed; the major problems of integration tasks are highlighted. Requirements for implementing data integration tasks are defined for the model, the architecture, the content, and the representation. In order to specify data source semantics, a meta-model is created, which is used to describe the concepts of source and relations.