User models for browsing digital libraries have to reflect each individual user’s interests in sufficient detail while still being of manageable dimensionality for efficiently comparing different users. Using a collection of computer science publications we explore ways to derive topic-based profiles and show how to compare them.