Guest Talk: Flexible Techniques for Querying RDF Graphs on the Web
Maribel Acosta, PhD
Date/Time: 26.11.2018, 16:00
Linked Data initiatives have encouraged the publication of RDF datasets on the web. To support online querying over RDF graphs, web access interfaces such as SPARQL endpoints or Triple Pattern Fragments (TPFs) have been deployed. However, the efficiency of query engines that consume Linked Data from SPARQL endpoints or TPFs can be negatively impacted by the unpredictable data transfer rates and server workload. The problem is mainly generated because SPARQL engines implement the traditional optimize-then-execute paradigm where querying strategies follow a fixed plan. An orthogonal but equally important aspect of querying RDF graphs on the web is the quality of the retrieved data. Executing SPARQL queries against graphs with quality issues leads to low-quality and even incomplete results.
In this talk, I will present flexible querying techniques to evaluate queries efficiently and effectively over remote sources. Our proposed solution relies on Adaptive Query Processing (AQP) to adjust on-the-fly the execution schedulers according to the current conditions, as well as on Crowdsourcing to complete missing statements in the RDF graphs. Our experimental results indicate that adaptivity speeds up query execution in comparison to fixed plans. In addition, we empirically show that crowdsourcing is a feasible solution for enhancing completeness of RDF graphs in different knowledge domains.
Finally, this talk will provide insights of open research directions to support more comprehensive RDF querying solutions on the web.
Maribel Acosta holds a PhD from the Institute AIFB in the Karlsruhe Institute of Technology (KIT), where she actually works as Postdoc researcher. She is currently focused on researching the enhancement of Semantic Web technologies with human-based computation, particularly applied to SPARQL query execution. Her research interests also include adaptive techniques for Linked Data management and federated query processing.
Back to overview