last.fm
the tracks found in my recent listening section are from my last.fm rss feed. last.fm is one of those damn-why-didn't-i-think-of-that web pages. you get a last.fm account and install their plugin into your mp3 playing software. the plugin tracks what you listen to and sends it to your account. the folks at last.fm use statistical-clustering magic on the data from all the user accounts to find interesting patterns in music listening habits.
based on your neighbors (people whose tastes are similar to yours) they give you recommendations for artists you have not listened to yet. i've found the recommendations to be quite good so far. when i first starting playing around with last.fm i didn't play my cds on my computer, which meant that the only tracks i was sending it is stuff that i own in digital form, mostly purchased from emusic. and yes, i own it, not like you itunes/napster/yahoo/etc drm chumps that are basically renting your music... but that's a whole other topic.
once i had sent a fair number of my listens to last.fm - the more data the better in statistical magic land - my top ten recommended artists were all ones whose cds i owned. this had me completely sold. i've since started ripping all my cds (slowly, since i have close to 1,000 and a slow as hell cdrom), which has lead to a lot of music rediscovery.
last.fm also provides tons of charts, which is great for an info junkie like myself. it is interesting to see my own listening habits over time. i can spy on other people's listening habits too. they also produce charts for top tracks and artists and artist similarity charts that are another good source for recommendations. it is very community oriented - anyone can start a group with any theme and other people can join and discuss stuff in refreshingly un-moderated forums. there are also the last.fm radio stations. these stations select the tracks that play based on the data they have collected and some initial starting point you give it. you can say "play me stuff like this" and give it an artist and it will play songs from similar artists. just another step in the internet killing off the bullshit mainstream music industry. such a good idea.
what i love about last.fm is that it is based on behavioral data. there are a few other companies on the web trying to do similar stuff but they are trying to do it with the features of the music itself. i have heard good things about pandora and some people even like it better than last.fm for discovering new music. the thing about pandora is it is completely based on human-coded data. those guys sat around and listened to 10,000+ songs and wrote down the features that each song has and use that as their database from which their radio chooses songs. to me this seems crazy. don't get me wrong, i love hand-coded data. it's great to work with and can be far more accurate and detailed than automatically generated stuff.
but musical likes and dislikes are quite general and this is reflected in listening habits. if somebody listens to a bunch of the beatles, led zeppelin, rolling stones and pink floyd do you think they'll like the doors as well? probably. did you need to analyze the common features of the songs of the first four artists and then compare those to the features found in the doors' songs to reach that conclusion? no, you just know that "people that like those bands also like this one." last.fm easily captures this intuitive relationship that exists between artists thanks to the info that around 100,000 people a day are pumping into their database.
what happens when a new artist comes out? the people that like "that kind of music" will buy it, listen to it and it will be in last.fm's database almost instantly. the pandora guys are gonna have to get the music themselves, listen to it and categorize it. hmm, which approach do you think will lead to a more up to date service with better coverage of lesser-know artists?
last.fm is good stuff. get an account. it is fairly painless. even if you don't feel like you would use such a service tomorrow you can still contribute your own personal musical tastes and help other people find good music.
based on your neighbors (people whose tastes are similar to yours) they give you recommendations for artists you have not listened to yet. i've found the recommendations to be quite good so far. when i first starting playing around with last.fm i didn't play my cds on my computer, which meant that the only tracks i was sending it is stuff that i own in digital form, mostly purchased from emusic. and yes, i own it, not like you itunes/napster/yahoo/etc drm chumps that are basically renting your music... but that's a whole other topic.
once i had sent a fair number of my listens to last.fm - the more data the better in statistical magic land - my top ten recommended artists were all ones whose cds i owned. this had me completely sold. i've since started ripping all my cds (slowly, since i have close to 1,000 and a slow as hell cdrom), which has lead to a lot of music rediscovery.
last.fm also provides tons of charts, which is great for an info junkie like myself. it is interesting to see my own listening habits over time. i can spy on other people's listening habits too. they also produce charts for top tracks and artists and artist similarity charts that are another good source for recommendations. it is very community oriented - anyone can start a group with any theme and other people can join and discuss stuff in refreshingly un-moderated forums. there are also the last.fm radio stations. these stations select the tracks that play based on the data they have collected and some initial starting point you give it. you can say "play me stuff like this" and give it an artist and it will play songs from similar artists. just another step in the internet killing off the bullshit mainstream music industry. such a good idea.
what i love about last.fm is that it is based on behavioral data. there are a few other companies on the web trying to do similar stuff but they are trying to do it with the features of the music itself. i have heard good things about pandora and some people even like it better than last.fm for discovering new music. the thing about pandora is it is completely based on human-coded data. those guys sat around and listened to 10,000+ songs and wrote down the features that each song has and use that as their database from which their radio chooses songs. to me this seems crazy. don't get me wrong, i love hand-coded data. it's great to work with and can be far more accurate and detailed than automatically generated stuff.
but musical likes and dislikes are quite general and this is reflected in listening habits. if somebody listens to a bunch of the beatles, led zeppelin, rolling stones and pink floyd do you think they'll like the doors as well? probably. did you need to analyze the common features of the songs of the first four artists and then compare those to the features found in the doors' songs to reach that conclusion? no, you just know that "people that like those bands also like this one." last.fm easily captures this intuitive relationship that exists between artists thanks to the info that around 100,000 people a day are pumping into their database.
what happens when a new artist comes out? the people that like "that kind of music" will buy it, listen to it and it will be in last.fm's database almost instantly. the pandora guys are gonna have to get the music themselves, listen to it and categorize it. hmm, which approach do you think will lead to a more up to date service with better coverage of lesser-know artists?
last.fm is good stuff. get an account. it is fairly painless. even if you don't feel like you would use such a service tomorrow you can still contribute your own personal musical tastes and help other people find good music.
posted by mucker at 12:49 PM
2 Comments:
This comment has been removed by a blog administrator.
Hi this is pervaiz from pakistan i have seen your profile i want make friends can you help me beacuse i am new user of orket & net its to you please help me this is my addres:pervaizsadiq@live.com I w"be thankfull to you
Post a Comment
<< Home