Learning to Rank (LTR), once the domain of academic researchers in machine learning and information retrieval, has begun to make great headway in practical applications on the web. Its tools and techniques offer a new way to think about challenges in relevance ranking, personalization, localization, ad targeting and multimedia search, but it shares enough conceptual foundations with traditional search relevance that it’s easy for hands-on engineers to get started with.

In this talk, Andrew will introduce the field and its key concepts, before diving into one of its best-known algorithms, Ranking SVM, which adapts Support Vector Machines to ranking tasks instead of classification problems.

With real examples taken from work at Etsy and elsewhere, he’ll talk about how you can use this algorithm and others like it to incorporate a huge variety of features into your search ranking model: query-specific term weights, implicit user feedback, temporal and geographic data, and even image features. And he’ll touch on applications of LTR beyond traditional search, including ad click prediction and content-based recommendations.

No past experience of machine learning will be required. Attendees can expect to leave this talk with an understanding of LTR and its applications, and enough insight into Ranking SVM to enable them to experiment with ranking models on their own data.