Characterizing certain answers via compactness

In relational data integration, some approaches rely on logic programming to get answers for one query from several sources. More precisely, in the mediation setting where a virtual global schema linked to all local source schemas is assumed, some algorithms are based on SLD-resolution to generate rewritings of a query expressed on the global schema, so that it can be evaluated on the many sources. The aim of this evaluation is to obtain the certain answers [1]. These answers are the answers that are always obtained when a user evaluates his query on any materialization of the global schema which reflects the source data [9]. Of course, it has been shown that the generated rewritings, especially via SLD-resolution, could compute the certain answers. However, from a more general point of view, and up to our knowledge, no certain answers characterization w.r.t. answers of a definite program formalizing a mediation setting has been proposed yet. This is the main contribution of this paper, which is based on a new compactness result that characterizes certain answers w.r.t. an extension to infinite models of certain answers.