In this paper, we present an ontology-based approach for the improvement of searching in an information portal. The approach is based on incremental refinement of user's queries, according to the ambiguity of a query's interpretation. The so-called Librarian Agent plays the role of the human librarian in the traditional library - it uses information, about the domain vocabulary, the capacity of the knowledge repository and the behaviour of previous users in order to help users find the resources they are interested in. Moreover, the agent analyses the users requests off-line and compares the users' interests with the capacity of the information repository, in order to find which new topics should be introduced or which topics users are no more interested in. We partially implemented the approach in the Web Portal of our Institute and some initial evaluation results are shown.

information retrieval, ontology, query refinement