A Diversity Adjusting Strategy with Personality for Music Recommendation

Diversity-based recommender systems aim to select a wide rangeof relevant content for users, but diversity needs for users withdifferent personalities are rarely studied. Similarly, research onpersonality-based recommender systems has primarily focused onthe ‘cold-start problem’; few previous works have investigated howpersonality influences users’ diversity needs. This paper combinesthese two branches of research together: re-ranking for diversifica-tion, and improving accuracy using personality traits. Anchoredin the music domain, we investigate how personality informationcan be used to adjust the diversity degrees for people with differentpersonalities. We proposed a personality-based diversification algo-rithm to help enhance the diversity adjusting strategy according topeople’s personality information in music recommendations. Ouroffline and online evaluation results demonstrate that our proposedmethod is an effective solution to generate personalized recommen-dation lists that not only have relatively higher diversity as well asaccuracy, but which also lead to increased user satisfaction.