Optimizing Horn-$$\mathcal {SHIQ}$$ Reasoning for OBDA

The ontology-based data access (OBDA) paradigm can ease access to heterogeneous and incomplete data sources in many application domains. However, state-of-the-art tools are still based on the DL-Lite family of description logics (DLs) that underlies OWL 2 QL, which despite its usefulness is not sufficiently expressive for many domains. Accommodating more expressive ontology languages remains an open challenge, and the consensus is that Horn DLs like Horn-\(\mathcal {SHIQ}\) are particularly promising. Query answering in Horn-\(\mathcal {SHIQ}\), a prerequisite for OBDA, is supported in existing reasoners, but many ontologies cannot be handled. This is largely because algorithms build on an ABox-independent approach to ontological reasoning that easily incurs in an exponential behaviour. As an alternative to full ABox-independence, in this paper we advocate taking into account general information about the structure of the ABoxes of interest. This is especially natural in the setting of OBDA, where ABoxes are generated via mappings, and thus have a predictable structure. We present a simple yet effective approach that guides ontological reasoning using the possible combinations of concepts that may occur in the ABox, which can be obtained from the mappings of an OBDA specification. We implemented and tested our optimization in the Clipper reasoner with encouraging results.