Presented at: First International Workshop on Ordering and Reasoning (OrdRing2011)
Data on the Semantic Web is ordered, for instance, through orderingsover literals (e.g., age, heights, location, rating etc), resources(e.g., popularity) or triples (e.g., provenance, confidence, time-stamps).Recently, orderings have been established as first class citizensin an extended SPARQL algebra, with an efficient calculus for finding, e.g.,top-k answers. In this paper we advocate a similar initiative for thesemantic (or reasoning) part of such querying: given an expressiveknowledge base with orderings, can we make use of these orderings toefficiently reason over large datasets where top-k answers are required.The basic idea is to define the notion of a top-k closure, which isa subset of the full semantic closure sufficient to answertop-k queries. We argue that this top-k closure can be efficiently calculated by applying top-k joins needed when applying entailment rules of the standard calculus.This paper introduces the top-k reasoning problem, and advocates a potential solution.
Resource URI on the dog food server: http://data.semanticweb.org/workshop/ordring/2011/paper/9
Explore this resource elsewhere: