Towards Zero-Input Personalization: Referrer-Based Page Prediction

Most web services take a “one size fits all” approach: all visitors see the same generic content, formatted in the same generic manner. But of course each visitor has her own information needs and preferences. In contrast to most personalization systems, we are interested in how effective personalization can be with zero additional user input or feedback. This paper describes PWW, an extensible suite of tools for personalizing web sites, and introduces RBPR, a novel zero-input recommendation technique. RBPR uses information about a visitor’s browsing context (specifically, the referrer URL provided by HTTP) to suggest pages that might be relevant to the visitor’s underlying information need. Empirical results for an actual web site demonstrate that RBPR makes useful suggestions even though it places no additional burden on web visitors.