Incremental Maintenance of dynamic Datalog programs

This paper is an extended abstract of [5], which discusses the incremental maintenance of materialized ontologies in a rule-enabled Semantic Web. The reader may wonder how this relates to Datalog, a language that has been proposed in the deductive database context. However, the semantics underlying Web ontology languages can often be realized by translating the ontology into appropriate Datalog programs. Therefore, our solution for incrementally maintaining materialized ontologies actually builds on the incremental maintenance of dynamic Datalog programs. Materialization generally allows to speed up query processing and inferencing by explicating the implicit entailments which are sanctioned by the rules of the program. The complexity of reasoning with Datalog is thereby shifted from query time to update time. We assume that materialization techniques will frequently be important for the Semantic Web to achieve a scalable solutions, since read access to is predominant in a Web setting. Central to materialization are maintenance techniques that allow to incrementally update a materialization when changes occur. We present a novel solution that allows to cope with changes in rules and facts for Datalog programs. To achieve this we extend a known approach for the incremental maintenance of views (intentional predicates) in deductive databases. We show how our technique can be employed for a broad range of existing Web ontology languages, such as RDF/S and subsets of OWL. Our technique can be applied to a wide range of ontology languages, namely those that can be axiomatized by a set of rules in Datalog with stratified negation. The challenge that has not been tackled before is dealing with updates and new definitions of rules. However, our solution extends a declarative algorithm for the incremental maintenance of views [4] that was developed in the deductive database context.