Broadly defined, yes. Pandora, with its Music Genome Project (admittedly a proprietary system that characterizes each song with 200 to 400 different attributes) is an example. See Mike Bergman’s post with comments at http://www.mkbergman.com/322/the-s...

Interesting question. The MGP (Music Genome Project) states it characterizes songs by 200-400 potential attributes each (see this incomplete listing from Wikipedia). What is semantics to you? My ‘glad’ is your ‘happy’? Indeed, there are tremendous difficulties in reconciling semantic heterogeneities, which I earlier outlined on this blog.

The real point, I think, is that Pandora is the example of a real service bringing real benefit using semantic-related data concepts in the background. Actually, I think that is where the vaunted “semantic Web” will really first occur (and is occurring), and not initially as the result of some concepts of theoretical or academic purity.

I don’t mean to be a jerk. It’s just that services such as Pandora should be the community’s current exemplars, not some logically provable inference theorems.

Thanks, Mike

I was just referred to Pandora the other week as well. The service is pretty amazing. I put together a collection of music discovery/recommendation sites that you may be interested in:

http://eric-blue.com/blog/2006/11/how_to_discover_music_you_like.html

To the comment about recommendation != semantic: this is a valid point. However, more and more music services are offering their data via web services, which helps enable applications to find trends, associations and meaning.

Two big examples are Rhapsody (http://webservices.rhapsody.com/) and Last.fm (http://www.audioscrobbler.net/data/webservices/). Data is available in XML, RSS, and MicroFormats.

Recommendation systems are not necessarily related to the Semantic Web