Curt Siffert
2006-04-27 00:38:03 UTC
Yeah, it was fun to try and pull together all the efforts at
once.
Rob has written before that he feels the moderation system needs
to be expanded, and could offer artists the ability to police
their own data a bit more. I don't feel that any of the current
efforts are counter to that direction, so I don't feel too
concerned about the moderation. Just needs a few more volunteers
to work with it.
As far as the rest of the stuff I looked at, the conclusion I've
basically come to is that there doesn't really need to be yet
ANOTHER from-scratch effort to encompass all this; we've got a
lot of resources to pull from already.
I'm really suprised that a file sharer/trader doesn't yet have
musicbrainz support built in and I'm not sure who would be the most
approachable candidate. Napster has started signaturing with
relatable, which I think is really cool, but I really doubt
they are cross-referencing the information against the musicbrainz
database. Scaling considerations aside, it would be very cool
if they did that.
The only thing I think we're truly missing is a good open-source
collaborative filtering system. I know relatable is doing
something there, but I didn't get the feeling it would be
on the same level as musicbrainz in terms of open-ness; our ability
to interact with and extend the functionality. (I know some
relatable folks are on this list; I'd love more clarification.)
I just think it would be ultra cool to have an external open-source
system out there like musicbrainz that would just be a huge recommendation
engine - in a way that would also be beneficial for new artists by
soliciting ratings for them, too.
I'm not sure of the best way to start efforts on such a thing.
The algorithms are cutting-edge A.I./CS stuff that would probably
make my brain hurt. And the datasets would of course be huge.
BUT, one thing I've found out from reviewing the different
research lists is that the folks that play with the algorithms
LOVE datasets so they can test out their own research. So even
if we didn't have the recommendation service set up, an effort
could probably draw attention to themselves in the community simply
by soliciting ratings. It might be as simple as asking Rob to put
in a "Rate this song (1-10)" doohickey in freeamp for while a song
is playing, and setting up a server to receive the ratings (identified
by audio signature).
(writings at http://www.tangrams.com/music/)
Curt
On Tue, May 01, 2001 at 02:53:31PM +1000, Adrian wrote:
> Hey Curt,
>
> I really like what you're doing. It's such a complex environment to analyse
> I know.
> I feel that the whole 'moderation' thing in Musicbrainz is unfortunate.
> There is a dream system out there, one that the open-source community could
> build where artists upload their own content and write their own metadata
> into Musicbrainz or wherever. The question is. . . . is it just a dream . .
> . . or a vision???
>
> ~Adrian
> _______________________________________________
> Musicbrainz mailing list
> ***@musicbrainz.org
> http://www.musicbrainz.org/mailman/listinfo/musicbrainz
once.
Rob has written before that he feels the moderation system needs
to be expanded, and could offer artists the ability to police
their own data a bit more. I don't feel that any of the current
efforts are counter to that direction, so I don't feel too
concerned about the moderation. Just needs a few more volunteers
to work with it.
As far as the rest of the stuff I looked at, the conclusion I've
basically come to is that there doesn't really need to be yet
ANOTHER from-scratch effort to encompass all this; we've got a
lot of resources to pull from already.
I'm really suprised that a file sharer/trader doesn't yet have
musicbrainz support built in and I'm not sure who would be the most
approachable candidate. Napster has started signaturing with
relatable, which I think is really cool, but I really doubt
they are cross-referencing the information against the musicbrainz
database. Scaling considerations aside, it would be very cool
if they did that.
The only thing I think we're truly missing is a good open-source
collaborative filtering system. I know relatable is doing
something there, but I didn't get the feeling it would be
on the same level as musicbrainz in terms of open-ness; our ability
to interact with and extend the functionality. (I know some
relatable folks are on this list; I'd love more clarification.)
I just think it would be ultra cool to have an external open-source
system out there like musicbrainz that would just be a huge recommendation
engine - in a way that would also be beneficial for new artists by
soliciting ratings for them, too.
I'm not sure of the best way to start efforts on such a thing.
The algorithms are cutting-edge A.I./CS stuff that would probably
make my brain hurt. And the datasets would of course be huge.
BUT, one thing I've found out from reviewing the different
research lists is that the folks that play with the algorithms
LOVE datasets so they can test out their own research. So even
if we didn't have the recommendation service set up, an effort
could probably draw attention to themselves in the community simply
by soliciting ratings. It might be as simple as asking Rob to put
in a "Rate this song (1-10)" doohickey in freeamp for while a song
is playing, and setting up a server to receive the ratings (identified
by audio signature).
(writings at http://www.tangrams.com/music/)
Curt
On Tue, May 01, 2001 at 02:53:31PM +1000, Adrian wrote:
> Hey Curt,
>
> I really like what you're doing. It's such a complex environment to analyse
> I know.
> I feel that the whole 'moderation' thing in Musicbrainz is unfortunate.
> There is a dream system out there, one that the open-source community could
> build where artists upload their own content and write their own metadata
> into Musicbrainz or wherever. The question is. . . . is it just a dream . .
> . . or a vision???
>
> ~Adrian
> _______________________________________________
> Musicbrainz mailing list
> ***@musicbrainz.org
> http://www.musicbrainz.org/mailman/listinfo/musicbrainz