World Wide Web is a source of information, and searches on the Web can be analyzed to detect patterns in Web users’ search behaviors and information needs to effectively handle the users’ subsequent needs. The rationale is that the information need of a user at a particular time point occurs in a particular context, and queries are derived from that need. In this paper, we discuss an extension of our personalization approach that was originally developed for a traditional bibliographic retrieval system but has been adapted and extended with a collaborative model for the Web retrieval environment. We start with a brief introduction of our personalization approach in a traditional information retrieval system. Then, based on the differences in the nature of documents, users and search tasks between traditional and Web retrieval environments, we describe our extensions of integrating collaboration in personalization in the Web retrieval environment. The architecture for the extension integrates machine learning techniques for the purpose of better modeling users' search tasks. Finally, a user-oriented evaluation of Web-based adaptive retrieval systems is presented as an important aspect of the overall strategy for personalization.