In this paper, we present an approach to refine user profiles that were derived from Web server logs in an automated procedure. In most application scenarios, such automatically derived profiles can only deliver a preliminary result and require human interaction for further refinement. We describe the individual steps to enhance and refine derived user profiles which can be used for personalization purposes (e.g. information filtering). In particular, the user can choose to refine the profile manually or use supporting techniques, such as ontologies, that assist him in the refinement process. In addition to information included in automatically derived profiles, the user thus explicitly provides information to refine his profile.