Musings on the Semantics of SPARQL

Vortrag von Stephan Mennicke

Veranstaltungsort: APB 3027
Beginn: 9. Januar 2020 um 13:00
Ende: 9. Januar 2020 um 14:30
Forschungsgruppe: Wissensbasierte Systeme
Event series: KBS Seminar
iCal

Graph simulations have found their way into different graph database management (GDBM) tasks, e.g., in the shape of Offline indexing structures, as theoretical models for graph schemas, or as viable alternatives to matching patterns up to graph homomorphisms. Among other advantages, it is often the tractability of the simulation problem being exploited in emerging applications. However, when it comes to evaluating the approaches, only basic graph patterns (BGPs) and rather small data instances, compared to today's large data instances like Wikidata or DBpedia, are considered. In the first part of this talk, I give some insights on how far graph simulations may be incorporated into full-fledged graph query processing. Therefore, we analyze different semantic interpretations of SPARQL, based on graph simulation, w.r.t. correctness, complexity, and effectiveness. Second, I briefly sketch why state-of-the-art simulation algorithms do not scale well in the graph query/data setting. I further show the effects of a devised solution that even integrates well with the SPARQL semantics we envisioned in the first part.