An improved modeling method for profile-based personalized search

Abstract

A Personalized search system aims to provide tailor-made results for each user's query according to his preferences. Building such a system depends mainly on creating profiles that represent real user interests. In this paper, we present a method for modeling users and creating accurate profiles by implicitly tracking and collecting user-browsing data. Furthermore, we examine techniques to enhance profile accuracy through combining multiple browsing data sources, distinguishing important concepts from irrelevant ones in a user profile, and the concept levels required from a reference ontology to describe user's interests.