In collaborative environment, providing flexible interoperation between heterogeneous knowledge-based software applications is critical for efficiency reasons, particularly in the context of unanticipated business collaborations. It is an objective pursued by ongoing research efforts on semantic interoperability: one expects that describing information within ontologies and reconciling these ontologies is sufficient to reach seamless information exchange. However, limits of ontology model and development result in lack of reliability of agreement between ontologies, thus reconciliation of ontologies is often pragmatically inconsistent. This paper proposes a methodology to determine what is contextual information, how to model and use it. Then, this methodology is applied to the reconciliation of ontologies, in the situation of unanticipated collaborations across organizations and between collaborators. The outcome is a context-based system that provides an evaluation of the pertinence of data associated with a concept, based on three distinct kinds of contextual information: (1) user’s domains and tasks, to personalize the search interface according to user’s needs; (2) perspectives used to annotate ontology concepts, as means for disambiguation of the pragmatic meaning of concepts; (3) the task for which the data is intended to, as a way to evaluate data relevance to fill the interoperability need.