Shlomo Berkovsky, NICTA, Australia.
The quantity of accessible information on the Web continues to grow rapidly and has far exceeded human processing capabilities. The sheer abundance of information often prevents users from discovering desired services, information, and products, or aggravates making informed and correct choices. This problem highlights the pressing need for intelligent personalized Web applications that simplify information access and content discovery by taking into account users’ preferences and needs (as represented by their user models) and delivering services in a way most valuable and convenient to users. Such applications are referred to as personalized applications.
One particular type of personalized Web-based applications that has recently become tremendously popular is recommender systems. These systems provide to users personalized recommendations about services, products, and information that they may be interested to examine or purchase. The generation of recommendations for users typically exploits information collected during their past interactions with the system, as well as the available domain knowledge. Extensive research into recommender systems has yielded in the last decade a wide variety of techniques, such as content-based filtering, collaborative, filtering, matrix factorization, critique- and knowledge-based recommendations, and their numerous hybridizations.
The main objective of this tutorial is to provide the participants with broad overview and thorough understanding of algorithms and practical applications for Web personalization and recommender systems. The tutorial consists of three principal components: (1) modeling users through observing their interactions in order to elicit their preferences and needs, (2) widely-used algorithmic techniques for Web personalization and recommender systems, and (3) user aspects in Web personalization and recommender systems, and practically deployed personalized Web applications. Each of these three components will initially provide a theoretical/algorithmic background, which will then be instantiated with practical examples and/or deployed systems. Thus, the participants will not only get exposed to a variety of approaches exploited by the state-of-the-art personalization and recommender systems, but will also consider practical challenges related to the deployment of these approaches in a Web-scale system. The tutorial will be delivered as an interactive talk rather than a lecture, such that the presenter will challenge the participants with mini-tasks and the participants will be able to ask questions and discuss their ideas.