Interactively Eliciting Database Constraints and Dependencies

When designing the conceptual schema of a future information system, it is crucial to define a set of constraints that will guarantee that once the subsequent database is implemented and operational, any changes made to its content by authorised users will maintain its consistency. Eliciting and expressing such constraints and dependencies is far from trivial, especially when end-users are involved and when there is no directly usable data to play with. In this paper, we present an interactive process aimed to elicit hidden constraints such as value domains, functional dependencies, attribute and role optionality and existence constraints. Inspired by the principles of Armstrong relations, it attempts to acquire minimal data samples in order to validate declared constraints, to elicit hidden constraints and to reject irrelevant constraints in conceptual schemas. This process is part of the RAINBOW approach, a collaborative interactive user-oriented approach to develop the data model of an information system based, among others, on the reverse engineering of user-drawn form-based interfaces.