Top-K Linked Data Query Processing

Presented at: 9th Extended Semantic Web Conference (ESWC2012)

by Andreas Wagner, Duc Thanh Tran, G√ľnter Ladwig, Andreas Harth, Rudi Studer

In recent years, top-k query processing has attracted much attention because in large-scale scenarios, computing only the $k$ best solutions is often sufficient and also, the only affordable way to reach acceptable response time. Top-k query processing has been dealt with in different contexts. One line of research targets the so-called join top-k, where the goal is to produce the k best final results through joining partial results. In this paper, we study join top-k in the Linked Data setting, where partial results to be joined come from different sources. Because the only available access pattern in this setting is URI source lookup, processing queries requires entire sources to be retrieved. Targeting this scenario, we show how existing work on join top-$k$ can be adopted to produce top-k results over Linked Data. We elaborate on strategies for book-keeping scores of partial results and to use them for better estimation of candidate result scores, i.e. to obtain tighter bounds for early termination. Based on experiments on real-world Linked Data, we show that the proposed top-$k$ processing technique substantially improves runtime performance.

Keywords: Linked data query processing, Query processing, Top-k query processing


Resource URI on the dog food server: http://data.semanticweb.org/conference/eswc/2012/paper/research/110


Explore this resource elsewhere: