Abstract
Ontology-based data access (OBDA), also known as virtual knowledge graphs (VKG), is a well-established approach to information management that facilitates the access to a (single) relational data source through the mediation of a high-level ontology, and the use of a declarative mapping linking the data layer to the ontology. In order to integrate multiple, possibly distributed and heterogeneous, data sources, in this work we formally introduce an extension of OBDA, called ontology-based data federation (OBDF), by combining OBDA with a data federation layer, which can expose multiple data sources as a single relational database. We discuss opportunities and challenges of OBDF, and provide techniques to deliver efficient query answering in OBDF by exploiting inter-source relations (called data hints) in the federated sources. Such techniques are validated through an extensive experimental evaluation based on the Berlin SPARQL Benchmark.