Fast ABox Consistency Checking Using Incomplete Reasoning and Caching

Reasoning with complex ontologies can be a resource-intensive task, which can be an obstacle, e.g., for real-time applications. Hence, weakening the constraints of soundness and/or completeness is often an approach to practical solutions. In this paper, we propose an extension of incomplete reasoning methods for checking the consistency of a large number of ABoxes against a given TBox. In particular, we use and extend the clash queries proposed by Lembo et al. [9] for DL-Lite to compute inconsistent patterns of ABox assertions. By caching instantiations of these patterns, we are able to reduce the amount of reasoning required to determine the inconsistency of an ABox with every previously processed ABox. We present experimental results of our approach in terms of runtime and accuracy and compare it against complete reasoning techniques, the reasoning approach for DL-Lite \(_{\mathcal {A}}\), and an approximate reasoning approach based on machine learning proposed in [15].