Summary: AI algorithms used by music streaming services are better at providing accurate recommendations for those who enjoy mainstream music. However, the algorithms often miss the mark when it comes to recommendations for those who listen to non-mainstream musical genres like hip-hop or heavy metal.
Source: TU Graz
Correction:The authors of the research paper referenced are Dominik Kowald et al.An earlier version cited Elisabeth Lex et al in error.
A team of researchers from Graz University of Technology, Know-Center GmbH, Johannes Kepler University Linz, University of Innsbruck, Austria and University of Utrecht, the Netherlands, compared how accurate algorithm-generated music recommendations were for mainstream and non-mainstream music listeners.
They used a dataset containing the listening histories of 4,148 users of the music streaming platform Last.fm who either listened to mostly non-mainstream music or mostly mainstream music (2,074 users in each group).
Based on the artists music users’ listened to most frequently, the authors used a computational model to predict how likely music users were to like the music recommended to them by four common music recommendation algorithms.
They found that listeners of mainstream music appeared to receive more accurate music recommendations than listeners of non-mainstream music.
Algorithm to classify music listeners
The authors then used an algorithm to categorise the non-mainstream music listeners in their sample based on the features of the music they most frequently listened to. These groups were: listeners of music genres containing only acoustic instruments such as folk, listeners of high-energy music such as hard rock and hip-hop, listeners of music with acoustic instruments and no vocals such as ambient, and listeners of high-energy music with no vocals such as electronica.
The authors compared the listening histories of each group and identified which users were the most likely to listen to music outside of their preferred genres and the diversity of music genres listened to within each group.
Those who mostly listened to music such as ambient were found to be most likely to also listen to music preferred by hard rock, folk or electronica listeners. Those who mostly listened to high-energy music were least likely to also listen to music preferred by folk, electronica or ambient listeners, but they listened to the widest variety of genres, for example hard rock, punk, singer/songwriter and hip-hop,
The authors then used users’ listening histories and a computational model to predict how likely the different groups of non-mainstream music listeners were to like the music recommendations generated by the four common music recommendation algorithms.
They found that those who listened to mostly high-energy music appeared to receive the least accurate music recommendations and those who mostly listened to music such as ambient appeared to receive the most accurate recommendations.
Biased music recommendation algorithms
Elisabeth Lex, the corresponding author, said: “As increasing amounts of music have become available via music streaming services, music recommendation systems have become essential to helping users search, sort and filter extensive music collections.
Our findings suggest that many state-of-the-art music recommendation techniques may not provide quality recommendations for non-mainstream music listeners. This could be because music recommendation algorithms are biased towards more popular music, resulting in non-mainstream music being less likely to be recommended by algorithms.”
“Further,” added Elisabeth Lex, “our results indicate that the music preferences of those who mostly listen to music such as ambient can be more easily predicted by music recommendation algorithms than the preferences of those who listen to music such as hard rock and hip-hop. This means that they may receive better music recommendations.
The authors suggest that their findings could inform the creation of music recommendation systems that provide more accurate recommendations to non-mainstream music listeners. However, they caution that as their analyses are based on a sample of Last.fm users their findings may not be representative of all Last.fm users or users of other music streaming platforms.
This research area is anchored in the Field of Expertise “Information, Communication & Computing”, one of five strategic foci of TU Graz.
About this AI research news
Source: TU Graz Contact: Elisabeth Lex – TU Graz Image: The image is credited to TU Graz
Support the underground: characteristics of beyond-mainstream music listeners
Music recommender systems have become an integral part of music streaming services such as Spotify and Last.fm to assist users navigating the extensive music collections offered by them. However, while music listeners interested in mainstream music are traditionally served well by music recommender systems, users interested in music beyond the mainstream (i.e., non-popular music) rarely receive relevant recommendations.
In this paper, we study the characteristics of beyond-mainstream music and music listeners and analyze to what extent these characteristics impact the quality of music recommendations provided. Therefore, we create a novel dataset consisting of Last.fm listening histories of several thousand beyond-mainstream music listeners, which we enrich with additional metadata describing music tracks and music listeners. Our analysis of this dataset shows four subgroups within the group of beyond-mainstream music listeners that differ not only with respect to their preferred music but also with their demographic characteristics.
Furthermore, we evaluate the quality of music recommendations that these subgroups are provided with four different recommendation algorithms where we find significant differences between the groups. Specifically, our results show a positive correlation between a subgroup’s openness towards music listened to by members of other subgroups and recommendation accuracy.
We believe that our findings provide valuable insights for developing improved user models and recommendation approaches to better serve beyond-mainstream music listeners.