The Landscape of Music(www2.research.att.com)
www2.research.att.com
The Landscape of Music
http://www2.research.att.com/~yifanhu/MusicMap/index.html
4 comments
Looks awesome. Seems to be sourced from Last.fm.
Also, don't miss:
map of movies: http://www2.research.att.com/~yifanhu/MovieMap/index.html
map of tv: http://www2.research.att.com/~yifanhu/TVMap/index.html
and map of books: http://www2.research.att.com/~yifanhu/BookMap/index.html
Also, don't miss:
map of movies: http://www2.research.att.com/~yifanhu/MovieMap/index.html
map of tv: http://www2.research.att.com/~yifanhu/TVMap/index.html
and map of books: http://www2.research.att.com/~yifanhu/BookMap/index.html
any idea how the relationships are inferred?
[i've used last fm's artist tags to construct relationships between artists (artists with tags in common are related) and so generate playlists that of related tracks - the last fm api is really easy to use.]
[i've used last fm's artist tags to construct relationships between artists (artists with tags in common are related) and so generate playlists that of related tracks - the last fm api is really easy to use.]
My first guess was that they used the artist.getSimilar API method. I dug around a little and found this:
"In April 2009, we crawled the website [last.fm] by starting with Beethoven, and the top 20 musicians most similar to Beethoven, provided that each has at least 100,000 listeners. We then found the top 20 most similar musicians to each of those with at least 100,000 listeners and proceed recursively. Our crawl yielded a graph with 2782 musicians, with edge weights the similarity between musicians."
- Emden Gansner, Yifan Hu, Stephen Kobourov and Chris Volinsky, Putting Recommendations on the Map - Visualizing Clusters and Relations: http://arxiv.org/abs/0906.5286
On an semi-related note, there is a wealth of information on visualizing music in the slide deck here: http://visualizingmusic.com/2009/10/22/using-visualizations-...
"In April 2009, we crawled the website [last.fm] by starting with Beethoven, and the top 20 musicians most similar to Beethoven, provided that each has at least 100,000 listeners. We then found the top 20 most similar musicians to each of those with at least 100,000 listeners and proceed recursively. Our crawl yielded a graph with 2782 musicians, with edge weights the similarity between musicians."
- Emden Gansner, Yifan Hu, Stephen Kobourov and Chris Volinsky, Putting Recommendations on the Map - Visualizing Clusters and Relations: http://arxiv.org/abs/0906.5286
On an semi-related note, there is a wealth of information on visualizing music in the slide deck here: http://visualizingmusic.com/2009/10/22/using-visualizations-...
thanks!
The Blues Brothers are between Frank Sinatra and The Beatles?
Pretty much the only thing I don't like about Hacker News: the drive-by shootings. This was a serious comment, although clearly I need to expand on it.
Somehow this map has put The Blues Brothers between Frank Sinatra and The Beatles. To me that's just ridiculous, and a clear indication that their "similarity" algorithm is seriously broken.
They don't seem to explain it either. They talk about their recursive technique for starting from one point and fanning out, but they magically talk about "similarity" without giving any definition or indication of how they compute it.
Personally, I write dozens of programs every week to explore data, and I look at the result for stand-out stupidities. To me, this is one, and it says that the algorithms need to be checked.
Somehow this map has put The Blues Brothers between Frank Sinatra and The Beatles. To me that's just ridiculous, and a clear indication that their "similarity" algorithm is seriously broken.
They don't seem to explain it either. They talk about their recursive technique for starting from one point and fanning out, but they magically talk about "similarity" without giving any definition or indication of how they compute it.
Personally, I write dozens of programs every week to explore data, and I look at the result for stand-out stupidities. To me, this is one, and it says that the algorithms need to be checked.
I don't really have any intuition as to where the The Blues Brothers should lie on this type of map, so maybe you could explain why you think their placement is so ridiculous?
The similarity algorithm appears to be taken directly from last.fm, which uses some kind of collaborative filtering method on user listening stats to determine artist similarity.
In general, I think this map does a good job of grouping similar artists. Mapping a high dimensional similarity space down to two dimensions is undoubtedly going to result in a few data points that seem out of place, particularly when dealing with content like artist similarity where individual perceptions can vary drastically.
The similarity algorithm appears to be taken directly from last.fm, which uses some kind of collaborative filtering method on user listening stats to determine artist similarity.
In general, I think this map does a good job of grouping similar artists. Mapping a high dimensional similarity space down to two dimensions is undoubtedly going to result in a few data points that seem out of place, particularly when dealing with content like artist similarity where individual perceptions can vary drastically.
I don't know where they should go on this type of map, but I do know that there's not much in common between The Blues Brothers and Frank Sinatra, or The Blues Brothers and The Beatles. Just listening to the music suggests there's not much in common. Any algorithm that suggests there is is, to my mind, suspect.
I don't know much, indeed, anything, about most of the bands/artists on this map, so I'm not going to suggest any rearrangement, but I'd be interested to see this concept of "similarity".
Actually, I probably wouldn't be interested, but I hereby announce my position as a skeptic.
I don't know much, indeed, anything, about most of the bands/artists on this map, so I'm not going to suggest any rearrangement, but I'd be interested to see this concept of "similarity".
Actually, I probably wouldn't be interested, but I hereby announce my position as a skeptic.
As a musician (one who, admittedly, wound up owing money when the bar tab was subtracted from the gate) I can't think of a better place to put them. They actually live on a line just south of the Sinatra-Beatles meridian. If you look just southwest of Sinatra, you'll see that Cab Calloway lives on the true line, with Count Basie just to the left. This would be the "dirty", less polished, and less, well, white version of the Sinatra-Beatles line (which would probably extend westward to Paul Whiteman if he appeared on the map at all).
No, the one that puzzles me is Sarah Brightman. I can understand them trying to keep her away from the cool kids, but she's got to be feeling like she got off the bus in the wrong neighborhood.
The style of music the Blues Brothers played, the Stax sound, cuts the continuum running from jazz-influenced big band risqué pop to straight-ahead risqué rock and roll. The rhythm section is from the rock world -- electric bass, guitar and piano, with foursquare beat-keeping drumming (as opposed to the sort of thing that, say, a Buddy Rich would do) -- the vocal performances are clearly in the Calloway-Prima-Charles lineage, and the horns are pure big band pop.
No, the one that puzzles me is Sarah Brightman. I can understand them trying to keep her away from the cool kids, but she's got to be feeling like she got off the bus in the wrong neighborhood.
The style of music the Blues Brothers played, the Stax sound, cuts the continuum running from jazz-influenced big band risqué pop to straight-ahead risqué rock and roll. The rhythm section is from the rock world -- electric bass, guitar and piano, with foursquare beat-keeping drumming (as opposed to the sort of thing that, say, a Buddy Rich would do) -- the vocal performances are clearly in the Calloway-Prima-Charles lineage, and the horns are pure big band pop.
Hmm.
OK, I'll go and listen to some more and try to hear the similarities and connections you make. Thanks for taking the time and effort to explain - 'preciate it.
OK, I'll go and listen to some more and try to hear the similarities and connections you make. Thanks for taking the time and effort to explain - 'preciate it.
It may help to understand the difference between lastfm and pandora. It is the latter that deals in musical similarity. The former is based on social network, as in people who listen to X are likely to listen to Y. This could explain nearness of disimilar artists.
What's more interesting is the bands missing from the map.
Do the missing bands (NIN, for example) form an isolated "island" in the last.fm similarity graph?
http://img16.imageshack.us/img16/2796/tvmap.png