Track Recommender System Based on Lyrics, Audio and Popularity Features

In this paper we propose an experimental content-based track recommendation system, which relies on audio feature values from Spotify Data Catalogue track lyrics and popularity ratings. In order to assess the utility of the system, we considered evaluating its accuracy on users’ list of listened tracks, by computing the precision as performance measure. However, before evaluating this performance measure, we also had to take into account the assessment of a few parameters related to feature engineering required for the functioning of both the experimental system and, ultimately, the recommendation algorithm.