Cold start solutions for recommender systems
12th February 2015 at 14:00
Speaker: Amin Mantrach, research scientist at Yahoo Research, Barcelona, Spain
Abstract: Recommender systems are facing cold start problems usually referred as the item cold start and the user cold start. To solve both (1) the item cold-start problem and (2) the problem of recommending to the weakly engaged user, I will introduce my contribution to the topic, a novel framework based on non-negative matrix factorization (NMF). The framework learns collective representations from the item's features and the users' feedback. To illustrate the effectiveness of the framework in recommending cold items, I will present how it is implemented and deployed to personalize fresh videos. Afterwards, in order to solve the problem of recommending to the weakly-engaged user I will introduce two novel techniques:
- By relying on endogenous information, I will show how our framework improves the recommendations for the weakly-engaged user, this is around 75% of the users (i.e. large coverage);
- By relying on exogenous information extracted from search queries, I will show how the recommendations can be improved for 200K users.