Q.3. a) Recommender systems are used to generate recommendations for users on unseen items. Collaborative filtering is one such approach. Explain the main stages of collaborative filtering and illustrate this approach by generating a recommendation for item Oh Mercy for user Jack. (10 marks) Dylan Desire Saved Tempest Oh Mercy 3 3 Jack 1 Lily 1 Rosemary_4 Rita 1 5 4 4 4 4 5 3 b) In many domains, in addition to collaborative ratings, extra information is often present and may prove useful in the provision of useful recommendations. For example, in the domain of music recommendations, we may have information on each album being rated. So, in addition to each rating present, we have for each album the following set of attributes: year, producers, musicians, genre, sub-genre, songs. Outline an approach where this extra information can be used to hopefully improve the performance of the system. (8 marks) c) In your words, describe how you would measure the performance of your system designed in part b). (7 marks)