We present a logic-based approach for query refinement in ontology-based information portal. The approach enables a user to navigate through the information content incrementally and interactively. In each refinement step a user is provided with a complete but minimal set of refinements, which enables him to develop/express his information need in a step-by-step fashion. The approach is based on the model-theoretic interpretation of the refinement problem, so that the query refinement process can be considered as the process of inferring all queries which are subsumed by a given query. Moreover, query refinements are ranked according to their relevance to user's needs, whereas these needs are online discovered by analysing a user's behaviour. The approach is very suitable for modelling information retrieval tasks based on the database repositories, like searching product catalogues.