Music recommendation in the long tail

Abstract

The online music industry has been growing at a fast pace, especially during the recent years. Even music sales have moved from physical sales to digital sales, paving the way for millions of digital music becoming available for all users. However, this produces information overload, where there are so many items available due to, virtually, no storage limitations, it becomes difficult for users to find what they are looking for. There have been many approaches in recommending music to users to tackle information overload, one successful approach is collaborative filtering, which is currently widely used in commercial services. Although collaborative filtering produces very satisfiable results, it becomes prone to popularity bias, recommending items that are correct recommendations but quite ``obvious``. In this thesis, a new recommendation algorithm is proposed that is based on collaborative filtering and focuses on producing novel recommendations. The algorithm produces novel, yet relevant, recommendations to users based on analyzing the users’ and the entire population’s listening behaviors. An online user test shows that the system is able to produce relevant and novel recommendations and has greater potential with some minor adjustments in parameters.