RDFS Update: From Theory to Practice

Abstract

Abstract. There is a comprehensive body of theory studying updates and schema evolution of knowledge bases, ontologies, and in particular of RDFS. In this paper we turn these ideas into practice by presenting a feasible and practical procedure for updating RDFS. Along the lines of ontology evolution, we treat schema and instance updates separately, showing that RDFS instance updates are not only feasible, but also deterministic. For RDFS schema update, known to be intractable in the general abstract case, we show that it becomes feasible in real world datasets. We present for both, instance and schema update, simple and feasible algorithms.

Updating RDFS ABoxes and TBoxes in SPARQL

“…The next result follow immediately from, e.g., [4,11,19] and shows that query answering under RDF can be done by either query rewriting or materialisation. Various existing triple stores (e.g., BigOWLIM [3]) have the option to perform ABox materialisation directly upon loading data.…”

“…[11] suggest minimal cuts, in the rdfs:subClassOf (or rdfs:subPropertyOf, resp.) graphs as candidates for deletions of rdfs:subClassOf (or rdfs:subPropertyOf, resp.)…”

“…Previous work on updates in the context of tractable ontologies such as RDFS [11] and DL-Lite [5] typically has treated DELETEs and INSERTs in isolation, but not both at the same time nor in combination with templates filled by WHERE clauses, as in SPARQL 1.1; that is, these approaches are not based on BGP matching but rather on a set of ABox assertions to be updated known a priori. Pairing both DELETE and INSERT, as in our case, poses new challenges, which we tried to address by introducing different semantics and taking into account both materialised and reduced stores.…”

“…11 Sem red 0 : Again, the baseline for a red-preserving semantics would be to apply the naïve semantics, followed by (re-)reducing the triple store. …”

Efficient query answering against dynamic RDF databases

“…On average, it is worth maintaining the saturation, though in some (rare) cases, when the updates affect upper-level classes or properties of the schema, it may be more costly than re-saturating (up to five times in our experiments). A distinct yet related problem is finding which triples to delete from an RDF graph, so that an implicit triple no longer holds [32].…”