In this paper, we present an approach which uses users' assessments on documents to group them into communities of interests. An open source Internet/Intranet prototype was developed for GNU/Linux in the framework of the GALILEI project. The users of the system are described in terms of profiles, with each profile corresponding to one area of interest. While browsing on a collection of documents, users' profiles are computed on the basis of both the content of the consulted documents and the assessments from the profiles. These profiles are then grouped into Internet communities, which allows the individuals to collaborate. Currently, documents of interest are shared between members of a same community. This approach was validated on several document collections according to a well-defined methodology and provides promising results.