Magic Rewritings for Efficiently Processing Reactivity on Web Ontologies
- Cite this paper as:
- Tovar E.L., Vidal ME. (2008) Magic Rewritings for Efficiently Processing Reactivity on Web Ontologies. In: Meersman R., Tari Z. (eds) On the Move to Meaningful Internet Systems: OTM 2008. OTM 2008. Lecture Notes in Computer Science, vol 5332. Springer, Berlin, Heidelberg
In this paper, we describe an approach aiming at enriching the Semantic Web with active information. We propose ACTION, an ACTIve ONtology formalism to express reactive behavior. In ACTION, events are categorized as concepts of an ontology and, in conjunction with classes, properties and instances, are considered during the query answering and reasoning tasks. We hypothesize that ACTION provides a more expressive solution to the problem of representing and querying active knowledge than existing ECA-based approaches. However, this expressivity power can negatively impact on the complexity of the query processing and reasoning tasks because the number of derived data depends on the number and relationships of the events. The main source of complexity is produced because the number of the derived facts is polynomial with respect to the size of the events, and the same evaluations may be fired by different events. To overcome this problem, we propose optimization strategies to identify Magic Set rewritings where the number of duplicate evaluations is minimized. We present the query rewriting technique called Intersection of Magic Rewritings (IMR), which is based on Magic Sets rewritings that annotate the minimal set of rules that need to be evaluated to process reactive behavior on an ontology. We have conducted an experimental study and have observed that the proposed strategies are able to speed up the tasks of reasoning and query evaluation in two orders of magnitude for small ontologies, and in four orders of magnitude for medium and large ontologies, with respect to the bottom-up strategy.
KeywordsOntology Active Knowledge Magic Set rewritings
Unable to display preview. Download preview PDF.