In this article we present an approach extracting contextualized user profiles in an enterprise resource sharing platform according to the users' different topics of interest. The system analyses the social annotations of each user's preferred resources and identifies thematic groups. For every group a weighted term vector is derived that represents the respective topic of interest. Each user profile consists of several such vectors that way enabling recommendation lists with a high degree of inter-topic diversity as well as targeted context-sensitive recommendations.

The proposed approach has been tested in our Enterprise 2.0 platform ALOE. A first evaluation has shown that the method is likely to identify reasonable user interest topics and that resource recommendations for these topics are widely appreciated by the users.