RDF Music
If the proposed foaf:tipjar becomes a reality, I want to have artists’ foaf information embedded in ID3 tags, so that iTunes could give me a ‘Donate to this Artist’ button.
In fact, having access to ID3 tag information as RDF would be cool too. Especially if songs had an IFP, since then I could reference a specific song in a music review, and have it all get tangled up in a semantic web. That would also allow media players to leverage all sorts of external annotation. Play a new song, and you get a donate button, lyrics, a blurb explaining that it’s a cover of another song with links to the iTMS where you could download the original, band information including current and past projects by band members, etc.
It’s been a while since I’ve looked at the ID3 specs, but this information could probably be dumped into a comment or text frame.
Comments
great idea (both foaf:tipjar and being able to use RDF tech in music apps). actually, i believe the ID3 standard allows you to define your own tags, at least on an experimental level until they’re adopted into the standard.
as you say, one big issue is defining an IFP for songs (for the non-techies, that’s a unique way to refer to a specific song regardless of whether it’s an MP3, AAC, or any other format). if you haven’t looked at it, relatable’s trm techonology tries to do this. looks like it might be patented, but they come pretty close to identifying songs correctly based on the song’s acoustical properties. musicbrainz uses trm in their database.
Posted by: gary on February 14th, 2004 3:41 PMYeah, I’d be wary of using TRM for this sort of thing. It’s just asking for trouble. Not to mention that TRM would have problems if you wanted to have different versions of the same song by the same band related to the same conceptual song entity. In that case, you’d want Song Instances that could point to their parent Song.
I’d prefer to use a public ontology with preference going to pages on a band’s site if available.
Posted by: kasei on February 14th, 2004 3:50 PMthis is getting a bit off-topic from your original post, but are you trying to identify performances of a song, or written versions of the song (lyrics, notes, etc.)? trm would obviously be used for performances, and i think that’s the right way to go. you can then use rdf to associate the various performances of a song (each with their own unique ID) with the “conceptual song”.
i definitely agree that a public ontology would be better. maybe something like imdb.com for music? discogs.com tries to do this for electronic music. however, i think your idea of pages on a band’s site is somewhat flawed. this goes back to the problem that a person is not a web page, a song is not a web page, an event is not a web page, etc. so while you can say that a certain web page describes a song, you can’t really use it to identify that song. it’s not even an IFP. and have you tried going back to band web pages that you had bookmarked many years ago? they’re almost invariably either completely off-line or no longer maintained (at least for the less well-known bands).
Posted by: gary on February 15th, 2004 4:18 PMI’d want to identify: performances (an instance of a song), a version (by band), and conceptual songs. As far as band web pages, it doesn’t matter that a URL isn’t a song; the URL only has to represent the song. Even if you balk at that idea, though, you could still use an IFP like foaf:homepage to define a song by a band’s song page.
As for your claim that “while you can say that a certain web page describes a song, you can’t really use it to identify that song,” I completely disagree if we take the URI that happens to be a URL for the song page, that URI can absolutely be used to identify the song. That’s exactly what RDF does: it takes URIs (that happen to often resolve as web page URLs), and uses them as identifiers for real world things.
Posted by: kasei on February 15th, 2004 4:31 PM
most of your techno-babble i gloss over but this, you should do. seriously.
Posted by: shuli on February 12th, 2004 11:22 PM