p1 Why l like it: Multi-task Learning for Recommendation and Explanation