par Picalausa, Francois 
;Vansummeren, Stijn
Référence Proceedings of the International Workshop on Semantic Web Information Management, ACM Press, page (7:1-7:6)
Publication Publié, 2011
;Vansummeren, Stijn Référence Proceedings of the International Workshop on Semantic Web Information Management, ACM Press, page (7:1-7:6)
Publication Publié, 2011
Publication dans des actes
| Résumé : | We present some statistics on real world SPARQL queries that may be of interest to researchers building SPARQL query processing engines and benchmarks. In particular, we analyze a log of SPARQL queries, harvested from the DBPedia SPARQL Endpoint from April to July 2010. We show that although a sizable portion of the log consists of so-called conjunctive SPARQL queries, non-conjunctive queries that use SPARQL's UNION and OPTIONAL operators are more than substantial.Current proposals for native RDF query engines however, focus their attention on providing innovative optimization algorithms for conjunctive SPARQL queries only, under the assumption that conjunctive queries are more commonly used than the others. By invalidating this assumption, our analysis hence motivates further research on practical heuristics for processing non-conjunctive SPARQL queries.Unfortunately, recent results into the evaluation complexity of SPARQL show that query evaluation quickly becomes hard for non-conjunctive queries. We therefore drill deeper into the syntactical structure of the non-conjunctive queries in the log and show that in 50% of these cases, these patterns satisfy certain structural restrictions that imply tractable evaluation in theory. We hope that the identification of these restrictions can aid in the future development of practical heuristics for processing non-conjunctive SPARQL queries. |
