Friday, December 16, 2005

Semantic Web Übernodes

I hope that since I just discovered Stephan Decker’s blog that I might be forgiven for commenting on one of his older posts. In it he takes about the relation between Web 2.0, the Semantic Web and how he believes that both may come together in the “Semantic Web 2.0”.
I started out to write a rebuttal of his thesis, but in the process ended up actually understanding (and sharing) his arguments (I think, or I just bended them until I could agree).

I always tend to think of the Semantic Web as a web of personal homepages where everyone annotates his stuff. This vision seems to be at odds with the current state of the Web 2.0 which is dominated by large, centralised applications like or flickr. They offer lots of API’s – but are centralised nonetheless. It would be possible to build a really decentralised flickr - just embed metadata in each and every image you post on the web, give each image a unique uri, use a couple of trust services and define a suitable machine understandable comment format. This decentralised flickr, however, would be inconvenient* and quite difficult to build; its unrealistic that it will appear anytime soon.

But actually I wasn’t seeing the forest for the trees (not seeing the Semantic Web for the data?): Why do the semantic web notes need to be so small, if this causes inconvenience? Why not have one large node for pictures (or five for that matter), two for events, ten for reviews … ? This vision is less distributed than the current web, but has less technical challenges and is easier to understand & use (than my above sketched vision of a fully decentralised Semantic Web). Large database driven website may eventually become their own node in the Semantic Web. Everyone else (who’s currently using some simple shared web hosting or does not yet have a web presence) will not host their own Semantic Web data, but put their data on a few Semantic Web Übernodes.

* for example everyone posting images would need to worry about metadata standards and uris.

Manual Trackback: Richard Cyganiak writes:

He’s right. Here’s something I realized when I built the FOAF crawler for our Google Base upload experiment: Most RDF geeks hand-craft their FOAF files, or use RDF tools like the FOAF-a-matic or FoafMe. There seem to be around 500 of these profiles out there in the vast wilderness of the WWW. Then, there are two million FOAF profiles created automatically by LiveJournal for their users. Five hundred! Two million! [...].