In order to personalize Web-based tasks, personal agents rely on representations of user interests and preferences contained in user profiles. In consequence, a critical component for these agents is their capacity to acquire and model user interest categories as well as adapt them to changes in user interests over time. In this chapter, we address the problem of modeling the information preferences of Web users and its distinctive characteristics. We discuss the limitations of current profiling approaches and present a novel user profiling technique, named WebProfiler, developed to support incremental learning and adaptation of user profiles in agents assisting users with Web-based tasks. This technique aims at acquiring comprehensible user profiles that accurately capture user interests starting from observation of user behavior on the Web.