Building Knowledge Bases with situations to help the cooperative revision

This paper describes REVINOS, an incremental modeling and cooperative revision tool for Knowledge Bases (KB) expressed with situation nodules. Situation nodules are simple and understandable objects, which allow to specify step by step interactions between the knowledge-based system and his environment. Knowledge representation by situation nodules has been applied for defining the chaining of French Minitel dialogues. This representation can be useful for building "evolutive" KB in an incremental and operational approach. Representations and tools that facilitate the comprehension and revision of the KB are very helpful in this context. The goal of our cooperative revision tool is to let the user concentrate on the modeling of some objects of the KB (situations, actions, facts) while the updating of the chaining rules of situation nodules are made automatically by machine learning techniques.