Beyond clicks

dwell time for personalizationXing Yi, Liangjie Hong, Erheng Zhong, Nanthan Nan Liu, Suju Rajan
Proceedings of the 8th ACM Conference on Recommender systems - RecSys '14

Add Remove

excerpt Traditionally, simplistic user feedback signals, such as click through rate (CTR) on items or user-item ratings, have been used to quantify users’ interest and satisfaction. Based on these readily available signals, most content recommendation systems essentially optimize for CTR or attempt to fill in a sparse user-item rating matrix with missing ratings. Specifically for the latter case, with the success of the Netflix Prize competition, matrix-completion based methods have dominated the field of recommender systems. However, in many content recommendation tasks users rarely provide explicit ratings or direct feedback (such as ‘like’ or ‘dislike’) when consuming frequently updated online content. Thus, explicit user ratings are too sparse to be usable as input for matrix factorization approaches. On the other hand, item CTR as implicit user interest signal does not capture any post-click user engagement. For example, users may have clicked on an item by mistake or because of link bait, but are truly not engaged with the content being presented. Thus, it is arguable that leveraging the noisy click-based user engagement signal for recommendation can achieve the best long term user experience. In fact, a recommender system needs to have different strategies to optimize short term metrics like CTR and long term metrics like how many visits a user would pay in several months. Thus, it becomes critical to identify signals and metrics that truly capture user satisfaction and optimize these accordingly. 1 on 2/29/2020, 11:23:22 PM


Deep Neural Networks for YouTube Recommendations cites Beyond clicks on 2/29/2020, 11:22:38 PM