Handling owl:sameAs in RDFox via Rewriting

RDFox is a scalable, centralised, main-memory, multi-core RDF system that supports materialisation-based reasoning in OWL 2 RL and SWRL. Reasoning with the owl:sameAs property using the OWL 2 RL/RDF rules can be very inefficient, so in this paper we extend RDFox to handle owl:sameAs via rewriting—replacing equal resources with a single representative. The main challenge in applying this well-known idea is to effectively parallelise processing. We also show that answering SPARQL queries over the rewritten triples is not straightforward due to the multiset semantics and builtin functions. Finally, we empirically show that, on data sets commonly used in practice, rewriting can reduce materialisation times and memory consumption by orders of magnitude.