This thesis exploits latent information in personalised recommendation, and investigates how this information can be used to improve recommender systems. The investigations span three directions: scalar rating-based ...