In the Vitalas project, we had the opportunity to work with search logs provided by user partner Belga News Agency, collected from their picture search portal. In this presentation, I will discuss the opportunities for context adaptation based on these search logs. The goal is to present different photos to a sports journalist who queries for Beckham, than the glossy magazine editor issuing the same query. We first (manually) classified 813 users into IPTC categories to represent their organisations main interest. Technically, we adapt the cluster-based re-ranking approach introduced by Kurland and Lee, to modify scores based on user context rather than document context. I will also discuss the application of the potential for personalization analysis method introduced by Teevan, Dumais and Horvitz in this setting, to estimate the potential for context adaptation, comparing IPTC-based user groups rather than individuals.
Presentation at SSR