Speaker: Efthymia Tsamoura, Alan Turing Institute and Oxford University, UK
In this talk, I will present my work on a problem that has attracted significant attention over the last decades: answering queries over incomplete datasets in the presence of TGDs and EGDs. Many techniques have been proposed in the literature for the query answering problem. However, it is difficult to apply them in practical scenarios. The “bottom-up” approaches involve computing all possible consequences under all dependencies, incurring, thus, an intolerable computational and storage cost. The “top-down” approaches based on query-rewriting aim at decreasing the overhead of the bottom-up ones, however, they only support restricted subclasses of TGDs and no EGDs. My contribution to this problem is a proposal of a goal-driven query answering algorithm that supports terminating classes of TGDs and EGDs. Myalgorithm involves a novel static rule analysis component and an extension of the well-known magic sets algorithm for rules with function symbols and equalities. The empirical evaluation over a challenging benchmark shows that the algorithm can drastically improve the performance of query answering.
This is a joint work with Michael Benedikt and Boris Motik from University of Oxford.
About me: I got my PhD in Computer Science in 2013. From June 2013 to September 2016 I have been a postdoctoral researcher in the department of Computer Science of University of Oxford working on problems related to querying data under constraints. On March 2016, I was awarded a fellowship for young researchers from the Alan Turing Institute, the UK’s National Institute for Data Science, and since October 2016 I am a Research Fellow in the Alan Turing Institute affiliated with University of Oxford. My main research interests include databases, knowledge representation and reasoning.
