Incremental Reasoning on RDF Streams

The introduction of stream processing methods in the Semantic Web enables the management of data streams on the Web. Chapter 6 introduced models for RDF stream and several extensions of SPARQL engines with windows for stream processing. The chapter assumes the absence of a TBox, so it is possible to compute the query answer without considering the ontology entailment defined through a TBox described in an ontological language. In this chapter, we relax this constraint and we consider the case of query answering over RDF streams when the TBox is not empty. In particular, we focus on Stream Reasoning [544], the topic that studies how to compute and incrementally maintain the ontological entailments in RDF streams. In traditional Semantic Web reasoning data are usually static or quasistatic1, so the whole computation of the ontological entailment can be executed every time the data change. When we consider RDF streams the static hypothesis is not valid anymore: RDF stream engines work with highly dynamic data and they need to process them faster than new data arrives to avoid congestion states. In this scenario, traditional materialization techniques could fail; a possible solution is the incremental maintenance of the materialized entailment using adaptations of the classical DRed algorithm [128, 506]: when new triples are added, the deducible data is added to the materialization; similarly, when triples are deleted the triples that cannot be deducted anymore are removed from the entailment. The idea of incremental maintenance was previously delivered in the context of deductive databases, where logic programming was used for the incremental maintenance of such entailments. The idea of incrementally maintaining an ontological entailment was proposed