first though about a genre ontology/taxonomy

38 views
Skip to first unread message

florent

unread,
Mar 3, 2007, 1:54:08 PM3/3/07
to music-ontology-sp...@googlegroups.com
Hi,

Starting a new thread for the genre/style taxonomy.

I had a look at the wikipedia page and infobox for their music genre. I also looked for what's already existing in this field.

First, wikipedia genre are quite heterogeneous in quality (some infobox missing) but the infobox  http://en.wikipedia.org/wiki/Template:Infobox_Music_genre looks nice
Meta genres are defined in the music genre template : http://en.wikipedia.org/wiki/Template:Music_genres
As I understand, their is two main "genre concept" :
- normal genre
- regional scenes that represent an othogonal classification following geography or culture

A first conceptual view for me look like this
class:genre
- hasSubGenre <-> hasOriginGenre (subgenre and stylistic origins of the infobox)
- birthEvent (would link that to the event ontology, it's the cultural origins of the infobox)
- diffusion (this is this mainstream popularity, i still don't know how to manage it, maybe timeline and geo place)
- main instruments (could be linked to the instrument vocabulary :) )

Now, I got three questions :
- should this FusionGenre be considered as a different subclass of genre or not ?
- same question for the regional scene, should we create a different class with or considere them as generic genre and add a geographical scene concept ?
- what about classical music ? Is there some specific things about it ?

There is also the geographical classification, I guess this could be set as kind of meta genre linked to the regional-scene defined in wikipedia (depend on how we decide to handle this regional scene concept)

I will start to model it (but i can't get the event ontology right now ? is it a problem with me ?)

also, I found one interesting paper ( http://www.csl.sony.fr/downloads/papers/2000/pachet-riao2000.pdf ). They review three classification (MP3.com, amazon and allmusic) and show there is only few crossing between them.
their approach is quite different as one genre as only one fqther and they classify the differences
they have also some nice/strange concepts, like not considering Rap as aspecific genre but as a descriptor for each genre because it's a vocal change... (will we need a vocal taxonomy ????)
They also link only track to genre, not album or artist. I would say that : track, album, and artist can be linked to genre and they can be linked to more than one genre :)

I know it look still a bit messy. I'll try to start something by tomorrow, and i wanted to get review (and questions) first.

++
florent

Yves Raimond

unread,
Mar 3, 2007, 2:02:57 PM3/3/07
to music-ontology-sp...@googlegroups.com
Hi Florent!

Just for information purpose: last year, I did a dump of the amazon
genre taxonomy in OWL. It is available here:

http://moustaki.org/resources/styles.owl

I also have a large subset of the amazon collection annotated with these genres:

http://moustaki.org/resources/collec.owl

It is really a bit messy, but quite fun (and surprisingly useful) though!

Cheers,
Yves

2007/3/3, florent <jah...@gmail.com>:

florent

unread,
Mar 3, 2007, 2:24:40 PM3/3/07
to music-ontology-sp...@googlegroups.com
Yes, I saw that and forgot to ask because i could not get it from your page.

also, i can't access to the event ontology...

Yves Raimond

unread,
Mar 3, 2007, 2:53:08 PM3/3/07
to music-ontology-sp...@googlegroups.com
> I will start to model it (but i can't get the event ontology right now ? is
> it a problem with me ?)


No - it seems the server got hacked yesterday:-( It is switched off
right now, until it is repaired. I hope it will be back online on
monday.

Cheers,
Yves

Reply all
Reply to author
Forward
0 new messages