I started from an artificial way of communicating: attaching messages on a shared virtual map. My interest was on how to support this kind of communication through an ontological categorisation of the messages. The problems I face, using this approach, are:

– the lack of a specific task, which makes hard to understand what is going on in the interaction;

– the complexity of natural language, which makes it hard to decode specific idiosyncrasy of the communication;

– the lack of the peers hidden context, which makes it hard to reconstruct the meaning of the message.

Another possible approach is to concentrate on the bits of commonsense knowledge that is built by analysing transversally the messages that pile in the environment. The focus is then not on the communication process but on the spatial context reconstruction with the help of the messages left by the people. Only after this context reconstruction is then possible to offer a support to human interaction.