Motivation
Music has been a core part of human entertainment for thousands of years. By having diverse and rich music taste, we make ourselves less ignorant toward other cultures, we can derive more enjoyment and appreciation for music as a whole, and we open our minds to new ideas.
Spotify, as the world’s leading streaming service, is the overwhelmingly the most common way for people to listen to music. By looking at genre and artist networks we can better understand what music people wish to create and listen to.
Analysis
Artist Collaboration Network
Via the Spotify API, South takes a collection of artists (n = 1250065), and adds an edge between two artists if they’ve collaborated on a track. These do not necessarily include only direct collaborations, but sampling, or covering another artist, for example would count as an edge. For purposes of the analysis all edges will be treated as undirected edges.
Most artists tend to have few collaborators, with a very small few having many, as expected. It may be worth looking at which artists have the most collaborators, as that would suggest a strong reverence in the music community and popularity. To get the artist with the most collaborations, all we have to do is find the nodes with the most amount of neighbours, or the node with the highest degree. We can see that classical composers like Bach and Mozart end up have the highest amount of neighbours.
I found this result to be simultaneously surprising and unsurprising. While classical music is very far removed from popular music in the last century, and thus has few listeners compared to pop music, popularity among listeners doesn’t necessarily result in more collaborations. Mozart and Bach’s works have become legendary and their works are some of the first you learn when studying music. This, combined with the fact that their work is copyright free and how easy it is to upload your own music to Spotify leads to the overwhelming amount of collaborators we see.
The friendship paradox as discussed in class, is also something that applies to artists. This means, on average, an artist that you have collaborated with has more collaborations with you.
This makes sense, as an artist that you’ve collaborated with is more likely to have more collaborations with other artists. For example, imagine you’re an indie artist and you sample a big artist like Kanye West. There are likely many others in the same scenario, as people want to collaborate with people bigger than them, due to that being the music they know and because of popularity.
Artist Genre Network
Hariharan’s article focuses on the connection of genres by looking at the most popular artists (n = 85283) on Spotify. She classifies each of their genres as a node, and if an artist is involved with more than one genre, she connects the genres together. These are undirected weighted edges, with weight being increased if multiple artists are classified as the same genres.
Looking at the degrees of each node, most nodes have small degrees, with very few having large degrees. This makes sense, many niche subgenres exist which only connect to their parent genre, which acts as a hub.
Looking the full network, we can immediately identify genre clusters within the network. For example, the dark green nodes are all avant-garde pop genres, orange nodes are all latin pop genres, etc. This is unsurprising as most artists like to stay in their own lane and make what they enjoy.
However, some of the connections were surprising. I found it surprising that latin pop was so condensed into its own corner, as currently it is extremely popular, with many western musicians seeking collaborations with latin pop stars. But it does make sense given how this trend is very recent and this network takes into consideration artists of all eras.
But at the same time, looking at the full network suggests a close connection between pop and rap music, despite the fact that hip hop’s dominance in popular music only really began in the mid 2010s. So it was surprising to see that when pop music for the majority of time has been disassociated with hip hop. In fact, this network suggests pop is closer to hip-hop than it is to any art pop subgenre or international pop, which is very unexpected.
If I had to guess a reasoning for this, it would be that Spotify classifies genres on modern standards. So while The Beatles were considered pop in the 60s, Spotify does not class them as pop because they are not pop music right now.
Sources
Hariharan, M. (2020, June 15). What kind of music do you listen to? - exploring the network of Spotify’s genres. Medium. Retrieved September 29, 2022, from https://medium.com/analytics-vidhya/what-kind-of-music-do-you-listen-to-exploring-the-network-of-spotifys-genres-56d188201a07
Network analysis of the Spotify artist collaboration graph – VRS 2023. (n.d.). Retrieved September 29, 2022, from https://vrs.amsi.org.au/wp-content/uploads/sites/78/2018/04/tobin_south_vrs-report.pdf