The advent of the World Wide Web and its constant growing have transformed the search for information into a time-consuming task. Intelligent information agents have emerged as a solution to this problem. These agents learn users’ interests and model them into user profiles in order to assist users by discovering, retrieving and summarizing information on behalf of them. This work is focused on the construction of user profiles for information agents starting from observation of users’ readings and behavior in the Web. Existing approaches have attacked partially this problem, treating the user profiling task either as a classification problem from the machine learning point of view or as a pure keywords analysis problem from the information retrieval point of view.
However, user profiling embrace a number of additional aspects that are not currently addressed in these approaches, such as modeling topic of interest with different levels of abstraction or modeling of contextual information about topics. In this work we propose a user profiling technique to be used in the development of intelligent information agents that deal with these aspects.