Another day, another moment, another thought, another post

Wednesday, November 23, 2005

Classifying music : Why Tim Berners Lee made so much sense in ISWC.?

I try to abstain from blogging about stuff I do in my research and today I wanted to do something and I realized how semantics can change the way we do things...
The task is simple: Classify all the MP3's I have from time so long ago into artists and genres. I began to use WinAmp to do the job manually for each file and surprisingly the stupid thing doesnt save the genre information. I wanted a genre called Rahmanisque. Come on one cannot box the genius of ARR into a blues r rap r jazz r classical r folk r anything... Same with the songs of the God (read as Ilayaraja). I had to type for each and every song. I downloaded some API to annotate using Java... Small piece of code and still I had to enter, although I found a way around. Now if we all had an ontology of genres, define what each genre means, then all one has to do is to search for the most appropriate genre from this and add it to mp3. Now think about extending this over some P2P sharing like Kaaza or even a point share system like cooltoad or even something thing like blinx or singing fish. Voila we can no longer worry about how Agni Natchatiram is spelt (AgniNatchatram, Agni Nakshatram and so on), but still get what we want. It is not all that difficult. Now each person can build his own ontology and then shares with a community. In the case of music, the community is possibly the fans. Best place to start would be the recording companies, who along with the numerous DRM crap that they want to dish out, can add this information. MP3's have already enough places to capture metadata using ID3. Now extend ID3 to hold more information. Information sharing now needs motivation more than technology. One issue is, once something like this is initiated, we will soon have people writing papers about automatically annotating content and the skeptics will rule over. Scalability, linguistics yada yada yada.... But certainly a manual meta-data addition is not hard. I added meta-data for 10 songs in 2 mins. Now 10 people for 2 mins a day and we have 3000 songs. If we can get 2000 man minutes, that is about 10000 songs a day. Semantic content sharing may not be that hard afterall...
Tim Berners Lee's talk about how ontology creation, perceived more as a social exercise can lead to creation of such a comprehensive shared resource got me thinking..Something for music would be so cool ...


Anonymous Anonymous said...

Hmm... I am not sure I get your idea.. Are you trying to automate the metadata addition to MP3s?? What did you do with your java annotation API?

4:27 PM

Anonymous Anonymous said...

So you do really do research!!! So what if we all start adding more information, meta-data whatever...
I think this approach of ontology creation is very interesting. So can we extend a social network like Orkut (OK You did tell me this long back but still ) to capture all this.. So is semantic web really for real? YES. I hope you dont mean automatic annotation. Manual annotation with some cool gui or ajax based web stuff is really possible.


5:08 PM


Post a Comment

Subscribe to Post Comments [Atom]

Links to this post:

Create a Link

<< Home