Second-Order Queries for Rule-Based Data Access

Rules and ontologies can be used to enrich a database system with advanced data access capabilities. The success of this paradigm has led to a number of languages such as DL-Lite, Datalog+/and OWL RL. The two major approaches to answering queries under constraints expressed in such languages are forward-chaining (materialization) and backward-chaining (query rewriting). The latter is typically focused on first-order queries that have only limited expressivity. We propose a querying formalism based on monadic second-order logic which subsumes and goes beyond conjunctive queries and regular path queries, but still has a decidable query subsumption problem. We devise methods for rewriting rule sets to queries in this new formalism and we show that query entailment in most of the established rule-based approaches can be decided by combining two methods: (i) bottom-up forward-chaining computation w.r.t. a rule set with the bounded treewidth model property and (ii) top-down second-order query rewriting w.r.t. a rewritable rule set.