Re: Question regarding open vs closed formulas (n3)

Seth,

It sounds like your problem is that the Semantic Web is an open world.
Think of it this way, if you don't already have a closed list of what
Bob owns, then how do you know when you've come across all of the
triples of form :Bob :owns :x? There could always be another one on some
website you haven't see yet.

In the database community, there has been work on something called Local
Closed World (LCW) information to deal with these kinds of problems. In
effect, an LCW statement says that such and such a source has complete
information with respect to some topic. Unfortunately, there is nothing
like this in RDF, DAML+OIL, or OWL. Hector Munoz-Avila and I wrote a
paper last year arguing for LCW on the Semantic Web, and presented
extenions to two semantic web languages (SHOE and DAML+OIL) that
supported it. If you're interested, you can find the paper at:

http://www.cse.lehigh.edu/~heflin/pubs/#aaai-02

Jeff

Seth Ladd wrote:
> 
> Hello,
> 
> If the outer most formula in cwm is open (which I'm assuming because I'm
> able to add triples to it), how can a inner formula, such as {:seth :has
> ?X} => {:seth :likes :?x}, ever return (that is, find the set of paths
> that match :seth :has ?X)?
> 
> How can it, theoretically, ever know it's gotten all the results for the
> query, and therefore, ever do the log:implies? In an open world, the
> query :seth :has ?x can't be proven to return all the matches, can it?
> 
> If so, does that mean the world is closed for just a moment?
> 
> I'm having a hard time writing a solution to this use case: I want to
> buy a Jimmy Buffet CD for Bob, one he doesn't have, and it's under $10
> dollars. Steve and Justin know what Bob has, so ask them. Ask Amazon
> then for prices.  It's the "CDs that Bob /doesn't/ own" that I'm having
> difficulty with. I'm worked out a owl:complementOf rules, but it's the
> act of computing the rdf:List of CDs Bob owns that gets me. I'm not sure
> if I can compute that List (or merge two Lists).
> 
> If I found a pre-existing List of CDs Bob owned, I could easily find
> those that he does NOT own.  But if I only find the facts that :Bob
> :owns :CD1. :Bob :owns :CD2, then I haven't been able to turn those two
> facts into a rdf:List (using n3's rules).
> 
> Maybe the question I should be asking is: Is this possible?  It would
> seem like the above use case is valid, and given a few facts on the
> network about what CDs Bob owns, I should be able to infer what he
> doesn't own.
> 
> Any ideas or tips?
> 
> Thanks very much,
> Seth

Received on Thursday, 26 June 2003 09:52:42 UTC