Rush: repeated recommendations on mobile devices
Proceedings of the 15th international conference on Intelligent user interfaces
| Hi-index | 0.00 |
The problem treated in this study is how to estimate the quality of playlists produced by an automatic playlist generator. Although a number of solutions has been offered on the topic of automatic playlist generation, very little work was published that regards evaluation of these systems. We are presenting here the experiments we did in order to establish whether a playlist can be evaluated by sight only, without listening, given that the songs are already known to the participants in the experiment. We also wanted to examine whether groupings of songs contribute to the playlist quality. The experiment was based on rating playlists of various origin. The result of the experiment was striking: the manually assembled playlists got the same ratings as the random ones. The answer to both questions seem to be negative. But this result revealed also some new important problems with playlist evaluation.