I seem to be eternally fatigued lately. Not really sure what it is - but I'm sure the fact that I'm pretty abusive to my body has something to do with it.

I worked a bit on AudioMatch - it's a website that logs your Winamp songs and also generates a dynamic image so you can link it up at forums or your webpage.



A lot of people misinterpret the aims of this site. It's purpose is not to provide you with a free service; it's to gather data about user listening habits and to extract data that may link up similiar artists.

I've always hated the idea of a musical genre - I find too many musicians cross between borders to effectively label "groups" via genres. Not to mention that most good bands tend to reinvent themselves with every album (e.g. No Doubt). Of course we can't forget bands that produce songs that sound the same over and over (e.g. Linkin Park); to these type of bands genres work fine.

But really. How do you label "No Doubt?"

Rock? Punk? Ska? What?

The idea with AudioMatch is to utilize the frequency of artist transitions within playlist. For instance, there are going to be high instances of users who listen to 2Pac then listen to Dre. Utilizing these frequencies we can determine that 2Pac and Dre are simliar artists and thus we can "recommend" them to users. Let me back up a quick second.

I've essentially determined that there are three primary methods in which people listen to music online:

1.) Some people (e.g. me) select songs on a song-by-song basis so we listen to just what we want.

2.) Via gargantuan playlists with full mp3 collection and shuffled.

3.) Via selective smaller playlists depending on mood

4.) Everyone else

For now, I'm throwing out #4, primarily because those people include ShoutCast users, and our plugin is having trouble with streams. Between the first three methods, most users will fall underneath the idea that you listen to similiar songs; whether they are matched by "genre" or by what you like to listen to when you're in certain moods (do you see where Tabulas fits in with AudioMatch?).

Now as we gather data, there are issues of ... error reporting. Obviously the people who do create gargantuan winamp playlists and listen to Stacie Orrico, then Rage Against the Machine, then Beethoven, then Tupac will obviously mess up the data validity, but I posit that these transitions will be minimal.

The user base will be large enough to marginalize any of these erroneous occurances; I'm essentially positing that as our sample size increases, there exist some clear correlations between certain artists, certain songs, and certain moods.

I do have plans on fully integrating AudioMatch with Tabulas, AND making full use of the mood feature. If I can somehow link up moods and songs, it'd be very interesting; I'm sure people with the mood "desperate" would be more likely to be listening to Radiohead's "Paranoid Android" than Britney Spear's "Oops I Did it Again."

In any case, we're still working out the basic of the site; a flexible database design and optimized PHP scripting to deal with the vast amount of raw information is what I'm really focusing development on.

Neeraj has been busting his ass on the plugin and has started planning the program that will actually do our analyzations. I do want to factor in existing ID3v1 and ID3v2 information ("genre") to help get bearings ...