In experimental sciences such as food science, data play an essential role, since domain theories are based on experimental data, their exploitation and their analysis. However, the state of the art shows that available experimental data are often partial, scattered on various supports, or without an established underlying mathematical model. Another information source is also available: expert knowledge, however not always formalized on written supports. Expert knowledge may express different viewpoints, possibly conflictual since they pursue divergent objectives. A main challenge is thus to integrate these data and knowledge and to develop ways of supporting decision from them. This research report presents a set of complementary strategies and methods defined and developed in order to, together, face this issue. It addresses three research topics: integration of heterogeneous formalisms, predictive methods, and argumentation for decision support.
