« Edinburgh | Main | Common Lisp and Web Development »
2006-05-22
RDF++
Yesterday I gave a talk at the W3C Advisory Committee meeting, titled "Identity Crisis and Serendipity" (unprotected version). I proposed to expand the class of applications that can be implemented using RDF (via the use of a reasoner) by "borrowing" two features from OWL. Namely, these features would be:
- something that would allows us to state that two nodes are the same (such as
owl:sameAs), and - inverse functional properties, to do the same automatically based on shared property values.
These features both address the issue of "identity", specifically the observation that many things one would like to describe do not have a URI but can still be uniquely identified. Now we can do, say, FoaF easily by relying on the underlying reasoning engine.
My current "development version" of Wilbur (not yet in CVS) implements this expanded "RDF++" reasoner, and OINK makes use of it. The additions still allow us to keep inference hidden and working behind the scenes, with applications accessing the deductive closure of the contents of the triple store.
I wonder if there are other features that could be seen as equally useful that we should add to RDF++?
Posted by ora at 05:23
Comments
Is it possible to make this PDF publicly available? It's currently password protected.
Posted by: foo at May 22, 2006 05:48 AM
Unprotected version is now at http://www.lassila.org/publications/2006/IdentityCrisisAndSerendipity.pdf
Posted by: Ora Lassila at May 22, 2006 06:05 AM
FYI, the Wilber link is broken (it's pointing at sourceforce, a domain squatter, rather than sourceforge.)
Posted by: Steve Dunham at May 23, 2006 12:34 PM
Broken link fixed...
Posted by: Ora Lassila at May 23, 2006 04:06 PM
About two years ago, Steve Harris and I were looking at a subset of OWL Lite features, tentatively called OWL Tiny, that could be layered on top of RDFS in our 3store triplestore.
In addition to owl:sameAs and owl:IFP, we also included the rest of the property characteristics (transitive, symmetric, functional and inverseOf). These were the OWL features most frequently requested by our users, who wanted something more than RDFS, but didn't want the complexity of full DL reasoning.
Posted by: Nick Gibbins at May 24, 2006 04:49 AM
So is full OWL too complex for real world applications?
Posted by: Steve Rogers at May 28, 2006 10:02 AM
I am not saying that -- categorically -- OWL would be "too much" for real world applications. There are a number of applications where it is appropriate to use OWL. There are, however, many applications that can be implemented using this (slightly enhanced) version of RDF.
Posted by: Ora Lassila at June 14, 2006 12:16 PM
Do you have an idea when the RDF++ reasoner will be checked in to the CVS?
Posted by: Paul Jones at June 15, 2006 09:56 AM
So it's easier to extend RDF slightly than to use OWL for only a small part of it's functionality?
Posted by: Steve Rogers at June 15, 2006 05:48 PM
I am hoping to have RDF++ in CVS in a couple of weeks. I want to reorganize the whole repository first...
As for the question of RDF++ vs. OWL--, personally I don't see much difference. But then, this is a scruffy speaking. The neats may have a different view. :-)
Posted by: Ora Lassila at June 16, 2006 11:58 AM