Exploiting Implicit User Activity for Media Recommendation
Michele Trevisiol · ACM SIGIR Forum · 2015
This thesis explores in depth how to exploit the user browsing behavior, and in particular the referrer URL, to understand the interest of the users. The aim is, first, to understand the preferences of the users from their navigation patterns, i.e., from the implicit actions of the users. Then, to exploit this information to personalize the content offered by the service provider. The key findings from our studies allowed us to propose innovative solutions to perform recommendation and ranking of media content. We show how the browsing logs are extremely meaningful also for cold-start problem -- estimating the preferences of newcomers.