This Friday we'll have two talks followed by drinks. Our industrial speaker is Lucas Bernardi, a Data Scientist at Booking.com. He will talk about recommender systems at Booking.com. Our academic speaker is Anna Sepilarskaia, a PhD student at the University of Amsterdam. She will talk about an approach for addressing the new user cold-start problem in recommender systems.
16:00-16:30 Lucas Bernardi
16:30-17:00 Anna Sepilarskaia
17:00-18:00 Drinks and snacks
Lucas Bernardi - Recommender Systems: A view from the trenches
Booking.com, the world's largest online travel agent applies Recommender Systems to improve customers experience at every step of their journey. We successfully design, build, and deploy dozens of recommender systems that today serve predictions for hundreds of millions of users every day. In this talk we present our business cases, how we solve them, what challenges we found and what we have learnt 'from the trenches'.
Anna Sepilarskaia - Preference Elicitation as an Optimization Problem
The new user cold-start problem arises when a recommender system does not yet have any information about a user. A common solution to this problem is to generate a user profile by asking the user to rate a number of items. Which items are selected will determine the quality of the recommendations made, and thus has been studied extensively. We propose a new elicitation method to generate a static preference questionnaire (SPQ) that asks a new user to make pairwise comparisons between items by posing relative preference questions. Using a latent factor model, we show that SPQ improves personalized recommendations by choosing a minimal and diverse set of static preference questions to ask any new user. We are the first to rigorously prove which optimization task should be solved to select each question in static questionnaires. Our theoretical results are confirmed by extensive experimentation. We test the performance of SPQ on two real world datasets, under two experimental conditions: simulated, when users behave according to latent factor model (LFM), and real, in which only real user judgments are revealed as the system asks questions. We show that SPQ reduces the necessary length of a questionnaire by up to a factor of three compared to state-of-the-art preference elicitation methods. Moreover, solving the right optimization task, SPQ also performs better than baselines with dynamically generated questions.