Résumé en anglais

Information in real-world applications is often vague, imprecise and uncertain. Ignoring the inherent imperfect nature of real-world will undoubtedly introduce some deformation of human perception of real-world and may eliminate several substantial information, which may be very useful in several data-intensive applications. In database context, several fuzzy database models have been proposed. In these works, fuzziness is introduced at different levels. Common to all these proposals is the support of fuzziness at the attribute level. This paper proposes first a rich set of data types devoted to model the different kinds of imperfect information. The paper then proposes a formal approach to implement these data types. The proposed approach was implemented within a relational object database model but it is generic enough to be incorporated into other database models.