In the settings of collaborative work or concurrent engineering, ontologies can be developed and shared between the different actors. This thesis studies how a context-based approach could improve interoperability between heterogeneous independent ontologies. The objective is to rank various ontology concepts according to their similarity to a given concept, based on existing ontology mappings, and on the context of a user or agent. A request being made on a concept, a prototype based on the approach should (1) discover concepts from other ontologies that are semantically related to the concept, (2) rank these concepts according to their relevance to a particular context, (3) in a timeframe of 10s.

We propose to place resources and users in a same reference system constituted of a restricted set of concepts that should cover most of the corpus, so as to associate to every concept a vector of coordinates. The context is made of a selection of concepts among the ontology concepts, and tasks among a hierarchy of task specific to the organisation and to the domains modelled. The ranking is done by comparing concepts based on their coordinates, and filtering them by the vector of coordinates built from the user's context.