User adaptive models based on similarity

In this work, we try to give a more precise meaning to the notion of "context" and to the knowledge associated with it, in order to build models representing a context able to adapt themselves to the user needs. We propose similarities as a tool to represent contexts. To do so, we consider a parametric fanfily of sunilarities and supply some methods for trying to "guess" the value of the parameter that best fits the context where the similarity shall be applied. We describe a general-purpose technique that derives a similarity starting from some sample information. Finally we give some hints on how to use our approach to develop adaptive user interfaces.