On the Learning of Functional Dependencies in Deductive Databases

The contribution of deduction mechanisms to (relational) databases is today well-studied. These methods open the way to an enrichment of databases by rules that govern these very data: these rules formalize some laws of the application under modelling. To express such rules we often need probabilistic (or modal) logics. We give hereafter a short abstract of some ideas concerning data analysis techniques used to obtain rules describing functional dependencies. For this we use a probabilistic logic and an elaborate framework for a database theory. All these topics will be treated at greater length in a forthcoming article.Original report number R86008. (A more comprehensive version, in French, appears in the Proceedings of the 1st Spanish Congress on Artificial Intelligence and Databases, Blanes, 1985.)