All reviewers are positive of this papers contributions in both methodology and experiments. Hence I am happy to suggest acceptance. Couple of comments about the paper that I hope authors will address in the final version. 1) Some of the drawbacks/ideas about user modeling have been explored thoroughly in the collaborative filtering/matrix completion setup. However this paper currently lacks such references. Drawing connections to this field of related work is important. 2) The paper can benefit from including more experiments on publicly available datasets.