Quest: Effcient SPARQL-to-SQL for RDF and OWL

Motivation. One of the most important uses of semantic technology is that of Ontology Based Data Access (OBDA), where the objective is to use shared vocabularies and ontologies as means to access data living in possibly disperse and heterogenous data sources (e.g., relational DBMS, XML databases, spreadsheets, etc.) Today this task often involves an ETL process in which the data is (E)xtracted from the source, (T)ransformed into RDF or OWL datasets in the target vocabulary and (L)oaded into a SPARQL endpoint or an OWL reasoner. This process carries all the issues that come with e.g., the need for synchronisation mechanisms to keep data up-to-date, the extra cost in time and space due to the duplication process, the additional software complexity at the client side, etc. Often it would be better to have live access to the original sources to avoid these issues and to be able to exploit any kind of optimisations that the original source can offer. In the context of relational DBMS and SPARQL queries, there exist several systems that allow for this on-the-fly approach, e.g., the D2RQ engine, Triplify, Spyder, Virtuoso RDF views, etc. However, often these systems fall short either in support for semantics (entailment regimes) and/or in query answering performance, e.g., the systems may send multiple queries to the sources and perform operations in-memory or they may generate complex SQL queries that cannot be planned and executed efficiently by the DBMS. This reality forces the use of the ETL approach, sometimes even in use cases in which an on-the-fly approach would be evidently possible.