This paper discribes a generic approach for data reverse engineering. This approach considers a data reverse engineering procedure to be composed of a source language, a data model and an executable process. It takes into account application domain knowledge. The GENRETRO tool, supporting this approach, is process-centered and considers two levels of interaction : one for experts and the other for analysts. The first level can be used to introduce data reverse engineering procedures and application domain knowledge. The second is used to guide the data reverse engineering process. The extracted information is stored in the data dictionary of a Software Engineering Environment.