Dynamic User Modeling for Personalized Advertisement Delivery on Mobile Devices

With the advent of broadband connections, the Internet service providers have begun to charge users with fees in order to cover the higher infrastructural costs due to the management/rental of broadband networks. However, the user is often left the chance for a free Internet connection, provided that he/she accepts to be invaded by banner ads, which arbitrarily, in terms of time and position, appear on the screen. In order to make effective the latter approach, the Internet service providers have been studying the most appropriate advertisement (ad, for short) supply policy, in order to satisfy companies who wish to advertise their products without annoying the user while he/she is interacting with the service. Personalized delivery of ads is recognized to be a promising approach to capture users’ interests in certain domains. However, further attention still deserves the issue of finding an adequate thread-off between the requirements elicited from advertising companies and the degree to which a user may accept to be bothered while performing some task. This problem is even more complex when ads are presented on mobile devices, where the reduced screen size considerably limits user’s capability to continue his/her task, so increasing the frustration caused by the presence of the ads. In this paper we present an approach for the presentation of personalized ads on mobile devices, which is based on a user model that takes into account user’s in-terests over time. The approach has been adopted within the LUNA wireless net-work project, which is targeted at realizing a business model such that the services provided in the area of Trento, in the North of Italy, are accessible and usable by everybody, at a very low cost. The client-side of the mobile LUNA application re-lies on a dynamic user modeling technique. We describe the personalization com-ponent of the advertising management system employed and we explain how a user model is dynamically created, saved and updated on the basis of the latest interaction history and of the delivered contents