A tool based on concepts used in knowledge representation, to facilitate application design

This work is at the boundary between knowledge representation and application design. We want to define a method to design an object schema from examples. We had to improve clustering algorithms used in knowledge representation (they build a hierarchy of classes from examples), so that they can manage additional problems encountered in application design. We have carried out four main improvements. First, to take semantics into account, we have built a thesaurus. Second, we must take into account not only attributes, but also methods. We must manage these methods while comparing examples and classes and while merging and extending classes. Third we induce multiple inheritance, which is necessary for real world applications. Fourth, in the case of multi-expertise (the examples are given by several users), we build a schema for each user, because we consider each user to have her/his own perspective. These schemata are then merged using an integration method. Our method is a tool based on knowledge representation that helps application design.