Predicting SPARQL Query Dynamics

The SPARQL language is the recommendation for querying Linked Data, but querying SPARQL endpoints has problems with performance, particularly when clients remotely query SPARQL endpoints over the Web. Traditionally, caching techniques have been used to deal with performance issues by allowing the reuse of intermediate data and results across different queries. However, the resources in Linked Data represent real-world things which change over time. The resources described by these datasets are thus continuously created, moved, deleted, linked, and unlinked, which may lead to stale data in caches. This situation is more critical in the case of applications that consume or interact intensively with Linked Data through SPARQL, including query engines and browsers that constantly send expensive and repetitive queries. Applications that leverage Linked Data could benefit from knowledge about the dynamics of changing query results to efficiently deliver accurate services, since they could refresh at least the dynamic part of the queries. Along these lines, we want to address open questions in terms of assessing the dynamics of SPARQL query results in order to improve the way applications access dynamic Linked Data, making queries more efficient and ensuring fresher results.