Abstract
The idea of using ontologies as a conceptual view over data repositories is becoming more and more popular. In these contexts, data are typically very large (much larger than the intentional level of the ontologies), and query answering becomes the basic reasoning services. In these contexts query answering should be very efficient on the data, and currently the only technology that is available to deal with large amounts of data is the one provided by relational data management systems (RDBMS). In this paper we advocate that for such contexts a suitable fragment of OWL-DL should be devised. Such a fragment must allow forms of query answering that exploit RDBMS when reasoning on the data, while it must include the main modeling features of conceptual models like UML class diagrams and ER diagrams. In particular it must include cyclic assertions, ISA on concepts, inverses of roles, role typing, mandatory participation to roles, and functional restrictions on roles. Also the query language should go beyond the expressive capabilities of concept expressions in description logics, and include at least conjunctive queries (corresponding to the select-project-join fragment of SQL). We discuss this issues by exhibiting a fragment of OWL-DL that includes all such features, namely DL-Lite, and showing that such a fragment is essentially maximal.