Scalable Reasoning by Abstraction Beyond DL-Lite

Recently, it has been shown that ontologies with large datasets can be efficiently materialized by a so-called abstraction refinement technique. The technique consists of the abstraction phase, which partitions individuals into equivalence classes, and the refinement phase, which re-partitions individuals based on entailments for the representative individual of each equivalence class. In this paper, we present an abstraction-based approach for materialization in DL-Lite, i.e. we show that materialization for DL-Lite does not require the refinement phase. We further show that the approach is sound and complete even when adding disjunctions and nominals to the language. The proposed technique allows not only for faster materialization and classification of the ontologies, but also for efficient consistency checking; a step that is often omitted by practical approaches based on query rewriting. A preliminary empirical evaluation on both real-life and benchmark ontologies demonstrates that the approach can handle ontologies with large datasets efficiently.