Reasoning on web data: Algorithms and performance

Techniques for efficiently managing Semantic Web data have attracted significant interest from the data management and knowledge representation communities. A great deal of effort has been invested, especially in the database community, into algorithms and tools for efficient RDF query evaluation. However, the main interest of RDF lies in its blending of heterogeneous data and semantics. Simple RDF graphs can be seen as collections of facts, which may be further enriched with ontological schemas, or semantic constraints, based on which reasoning can be applied to infer new information. Taking into account this implicit information is crucial for answering queries. The literature provides two classes of techniques for implementing RDF reasoning, namely query reformulation and saturation. Both are based on the idea of decoupling RDF entailment - the reasoning mechanism based on which query answers are defined - from query evaluation; the performance of the respective algorithms depends on the expressive power of the ontological schema language, as well as on the subset of features from the RDF standard which is supported. Our tutorial introduces the RDF ontological schema language for enhancing the RDF graphs' semantics, formalizes the query answering problem relying on reasoning, and provides a principled classification and analysis of the two techniques, with a particular focus on their performance trade-offs.