The aim of mis work is to build a water project management tool, which involves me study of expert knowledge transfer to a computer system. The architecture of this Intelligent Workstation is introduced wim the paradigm of Second Generation Expert Systems : different knowledge levels, which need several ways of knowledge acquisition, appear. On one hand, a project simulator is used for deep knowledge acquisition; and, on the omer hand, a telematics system, based on machine learning from examples, is used for shallow knowledge acquisition. This machine learning is done by incrementally constructing a Perceived Dependencies Network. Nevertheless, we don't expect me system to learn only from examples of resolved problems supplied by experts. The system also has to be able to hold a dialogue with experts to validate learned knowledge. This permits to prevent erroneous data introduction, while allowing knowledge base changes due to possible field evolutions (technological innovation, statutory evolution, ... ).
