Very cool. Although I wonder if the analysis really answered the question it set out to ask. The author's hypothesis was that the "entirety of contemporary Italian music rests on the shoulders of Gianni Maroccolo." He then tries to show this by showing who played music with who.
I would imagine that influence could be transferred even without artists playing in each other's bands. I can think of plenty of extremely influential bands that defined the start of a whole genre, whose members never played in another band.
But perhaps the original argument that the author was making was that Maroccolo is important because he played with everyone, in which case this analysis makes sense.
This and the other top comments are right that sharing an artist is only the crudest (and very noisy) way to estimate actual influence. I suppose one way to improve it would be to actually looking at the songs themselves, which might be possible with some deep learning? But that would require to actually have all the songs, which might not be feasible.
The graphs here show "significant sharing of artists" as lines between nodes (bands at an average point in album releases). I like this, but I think it leaves a lot out of the overall story.
It's much, much harder to compile a dataset that captures the path of music _influence_ as a way to measure it's impact. Yet, I wonder if that would look at all different? Consider the impact that one Giorgio Moroder had on electronic music globally, both in terms of "Italo Disco" and more. He's clearly in the article's illustrations _somewhere_, but as a solo artist and producer, may have few if any connections (collaborations & credits) to other Italian groups.
I have followed this guy for some years now, and he keeps doing marvelous stuff on complex networks and related analysis. I worked on the two parts of a similar idea regarding post-rock music; the first part became a Medium post [1], while I never finished the second one (although the idea is to publish it, I also do research on complex networks and their analysis).
In my misspent high school/college years, I was in search of very obscure, unpopular music, on independent and import labels. I began to find that much of the music that appealed to me was the fruit of collaboration between certain artists, producers, and labels. So I began to trace them out and, at that point, new music discovery consisted of finding brief and obscure collabs among the circle of artists that I was targeting.
For example, The Cure was exclusively signed to Fiction Records, which was more or less a vanity press for them, but they indeed had labelmates who were really, really obscure, and could always be connected back to Cure personnel.
It was sort of an amazing feeling, that everything was really interconnected in unexpected ways. In hindsight, all that music was a terrible influence and I was wasting time and money, but I also learned quite a bit about the record industry and collectibles, such as how to appraise the value of a piece, detect counterfeits vs. authentic pieces, and methods for archiving and preservation. My parents had collected postcards, stamps and other ephemera, and it sort of rubbed off on me!
It contains both the original artist-band bipartite network, and the actual projections used in the paper (and blog post).
The page on my website (https://www.michelecoscia.com/?page_id=2336) is slightly better because it also contains the code to make the projections yourself (and to verify the claims in the paper).
You can't have a correct vision of Italian music (staying only on that subsubsubcategory) of the 80s, 90s, 00s, 10s and 20s like that, at very best you'd end up ignoring significant artists and whole genres.
Wow it's the first time that I don't know part of what i'm talking about and i'm still right...
MySpace is mostly erased (prominent Italian musicians for some music waves of the 00s and 10s were there), some sites are erased, what pertains to radio and television is mostly not indexed on the public internet and there's significant digital divide between genres. In the Italian context, you'd end with a weird pantheon of rock, rap and popstars and not much else left, in conclusion you would... expunge significant artists and whole genres from results.
All correct, it was also one point raised by the reviewers. They had some suggestions about temporal centrality, but in the end I didn't followup on that because this centrality analysis was just a descriptive part that didn't really support any of the primary nor secondary claims in the paper.
Very cool. Although I wonder if the analysis really answered the question it set out to ask. The author's hypothesis was that the "entirety of contemporary Italian music rests on the shoulders of Gianni Maroccolo." He then tries to show this by showing who played music with who.
I would imagine that influence could be transferred even without artists playing in each other's bands. I can think of plenty of extremely influential bands that defined the start of a whole genre, whose members never played in another band.
But perhaps the original argument that the author was making was that Maroccolo is important because he played with everyone, in which case this analysis makes sense.
Anyway, cool networks.
This and the other top comments are right that sharing an artist is only the crudest (and very noisy) way to estimate actual influence. I suppose one way to improve it would be to actually looking at the songs themselves, which might be possible with some deep learning? But that would require to actually have all the songs, which might not be feasible.
vector embeddings of the songs plus coop network
The graphs here show "significant sharing of artists" as lines between nodes (bands at an average point in album releases). I like this, but I think it leaves a lot out of the overall story.
It's much, much harder to compile a dataset that captures the path of music _influence_ as a way to measure it's impact. Yet, I wonder if that would look at all different? Consider the impact that one Giorgio Moroder had on electronic music globally, both in terms of "Italo Disco" and more. He's clearly in the article's illustrations _somewhere_, but as a solo artist and producer, may have few if any connections (collaborations & credits) to other Italian groups.
https://en.wikipedia.org/wiki/Giorgio_Moroder
I have followed this guy for some years now, and he keeps doing marvelous stuff on complex networks and related analysis. I worked on the two parts of a similar idea regarding post-rock music; the first part became a Medium post [1], while I never finished the second one (although the idea is to publish it, I also do research on complex networks and their analysis).
[1] https://medium.com/festival-peak/exploring-the-post-rock-wor...
Thank you for the kind words! I'll take a look at your post. And feel free to get in touch in case you'd like to collaborate on this.
In my misspent high school/college years, I was in search of very obscure, unpopular music, on independent and import labels. I began to find that much of the music that appealed to me was the fruit of collaboration between certain artists, producers, and labels. So I began to trace them out and, at that point, new music discovery consisted of finding brief and obscure collabs among the circle of artists that I was targeting.
For example, The Cure was exclusively signed to Fiction Records, which was more or less a vanity press for them, but they indeed had labelmates who were really, really obscure, and could always be connected back to Cure personnel.
It was sort of an amazing feeling, that everything was really interconnected in unexpected ways. In hindsight, all that music was a terrible influence and I was wasting time and money, but I also learned quite a bit about the record industry and collectibles, such as how to appraise the value of a piece, detect counterfeits vs. authentic pieces, and methods for archiving and preservation. My parents had collected postcards, stamps and other ephemera, and it sort of rubbed off on me!
I have been a big Litfiba fan during their early career and loved this cover of Bowie's Yassassin by them. Maroccolo was their bassist.
https://www.youtube.com/watch?v=Q3vrp5t63K4
Great stuff. Network Backboning with Noisy Data paper that this is based on is even more interesting IMO. https://www.michelecoscia.com/wp-content/uploads/2017/01/201...
What a fascinating dataset they've constructed. Is the network available for download somewhere? I didn't see anything obvious.
Actually yes! It's on Zenodo: https://zenodo.org/records/13309793
It contains both the original artist-band bipartite network, and the actual projections used in the paper (and blog post).
The page on my website (https://www.michelecoscia.com/?page_id=2336) is slightly better because it also contains the code to make the projections yourself (and to verify the claims in the paper).
Thank you! I love analyzing graphics data like this.
Lovely analysis, some new music for me, and love to see work from former colleagues get recognition - hi Michele!
Hey Mali! Thank you so much! Do you want to do the network of Turkish music? :-)
Oh that does sound like a fun project!
Degree, closeness, and betweenness all seem like poor choices to make such an important decision. Why not HITS?
You can't have a correct vision of Italian music (staying only on that subsubsubcategory) of the 80s, 90s, 00s, 10s and 20s like that, at very best you'd end up ignoring significant artists and whole genres.
HITS doesn't refer to musical hit records. It's a network analysis technique.
https://en.wikipedia.org/wiki/HITS_algorithm
Wow it's the first time that I don't know part of what i'm talking about and i'm still right...
MySpace is mostly erased (prominent Italian musicians for some music waves of the 00s and 10s were there), some sites are erased, what pertains to radio and television is mostly not indexed on the public internet and there's significant digital divide between genres. In the Italian context, you'd end with a weird pantheon of rock, rap and popstars and not much else left, in conclusion you would... expunge significant artists and whole genres from results.
I'm not certain HITS could capture what appears to be the temporal component of the network the author incorporates.
I agree that it's difficult, I don't know much about centrality measures for dynamic networks.
However, degree sounds like a bad measure, since the value can be drastically increased by participating in a (few) big project(s).
All correct, it was also one point raised by the reviewers. They had some suggestions about temporal centrality, but in the end I didn't followup on that because this centrality analysis was just a descriptive part that didn't really support any of the primary nor secondary claims in the paper.
As Italian I never tought about it, but it makes lot of sense
What a great story this project tells.
... and Fabrizio De André [1][2]?
[1] https://www.youtube.com/watch?v=iKSkLw7YKwo
[2] https://en.wikipedia.org/wiki/Fabrizio_De_Andr%C3%A9
In the context OP evokes when he says "Maroccolo" (90s and 00s) I would have said Bubola in place of Maroccolo!
[flagged]