Advancing ELK: Not Only Performance Matters

This paper reports on the recent development of ELK, a consequencebased reasoner for EL+⊥ ontologies. It covers novel reasoning techniques which aim at improving efficiency and providing foundation for new reasoning services. On the former front we present a simple optimization for handling of role composition axioms, such as transitivity, which substantially reduces the number of rule applications. For the latter, we describe a new rule application strategy that takes advantage of concept definitions to avoid many redundant inferences without making rules dependent on derived conclusions. This improvement is not visible to the end user but considerably simplifies implementation for incremental reasoning and proof generation. We also present a rewriting of low-level inferences used by ELK to higher-level proofs that can be defined in the standard DL syntax, and thus be used for automatic verification of reasoning results or (visual) ontology debugging. We demonstrate the latter capability using a new ELK Protégé plugin.