In Ontology-Based Data Access (OBDA), a domain ontology is linked to the data sources of an organization in order to query, integrate and manage data through the concepts and relations of the domain of interest, thus abstracting from the structure and the implementation details of the data layer. While the great majority of contributions in OBDA in the last decade have been concerned with the issue of computing the answers of queries expressed over the ontology, recent papers address a different problem, namely the one of providing suitable abstractions of data services, i.e., characterizing or explaining the semantics of queries over the sources in terms of queries over the domain ontology. Current works on this subject are based on expressing abstractions in terms of unions of conjunctive queries (UCQs). In this paper we advocate the use of a non-monotonic language for this task. As a first contribution, we present a simple extension of UCQs with nonmonotonic features, and show that non-monotonicity provides more expressive power in characterizing the semantics of data services. A second contribution is to prove that, similarly to the case of monotonic abstractions, depending on the expressive power of the languages used to specify the various components of the OBDA system, there are cases where neither perfect nor approximated abstractions exist for a given data service. As a third contribution, we single out interesting special cases where the existence of abstractions is guaranteed, and we present algorithms for computing such abstractions in these cases.
2020, Proceedings of the Seventeenth International Conference on Principles of Knowledge Representation and Reasoning, Pages 243-252
Non-Monotonic Ontology-based Abstractions of Data Services (04b Atto di convegno in volume)
Cima Gianluca, Lenzerini Maurizio, Poggi Antonella
Gruppo di ricerca: Artificial Intelligence and Knowledge Representation