Media retrieval is still an open problem, due to many problems: the semantic gap, the difficulty in scaling according to dimension and diversity of media archives, the computational burden. The use of contextual information is considered among the best options to solve such problems, but no ultimate solutions are available to jointly exploit content and context. In this domain, the Glocal project proposes the notion of “event” as a way of organizing and indexing media for browsing and retrieval purposes. Glocal has already defined models able to capture the complexity of events as well as their relationships. Software instruments are being designed to create new event models and make them evolve in time through the social interaction among peers. Furthermore, a set of tools will allow users to generate instances of events and associating them to media collections in an automatic or semi-automatic way. The interaction with the user is also treated with special care, allowing personalization and adaptation to different user categories and applications through customized interfaces. To demonstrate and validate the Glocal approach, the project consortium is continuously working at the construction of media corpora, spanning different domains and targeting diverse communities, from professionals to common users.