Utility-aware Semantics for Alternative Service Expressions in Federated SPARQL Queries
Lars Heling, Maribel Acosta · 2022
SPARQL endpoints are web services that allow for querying RDF knowledge graphs online using the SPARQL query language. In SPARQL 1.1, the federation extension allows for explicitly evaluating expressions of a query at certain SPARQL endpoints using the SERVICE keyword. However, the current foundations of federated SPARQL query processing rely on two central assumptions: (i) members of the federation are assumed to be equal in terms of quality, and (ii) a strict definition of answer completeness, where a SPARQL expression is evaluated over all relevant endpoints. These assumptions hinder the scalability of federated query processing approaches, as the number of contacted sources is potentially high. To address this challenge, in this work, we propose a solution that supports quality descriptions about federation members, and further exploit these descriptions to prune low-quality sources and to speed up query execution. To this end, we present a definition of quality-aware federations of SPARQL endpoints. Moreover, we define and investigate the semantics of recurring ALTERNATIVE SERVICE expressions which are the union of SERVICE requests around the same expression. We then devise a utility-aware semantics for them, in which quality information about the federation is leveraged to potentially reduce the number of endpoints to be contacted. We evaluate the proposed semantics in an experimental study and investigate the utility, execution time, and answer completeness of the proposed semantics for queries in over 60 quality-aware federations of varying size and quality.