Re: WOWG: Report from WWW 2003 - OWL presentation/issues

At 10:58 AM -0400 6/5/03, Christopher Welty wrote:
Jeremy,

I've argued this with Pat several times.  I'd like to see an 
authoritative definition of what "first-order" means, otherwise we're 
all using our own definitions.  In any dictionary of logic or 
philosophy or mathematics that I've been able to find, "first-order" 
is defined as "not higher order" and "higer order" is defined as 
predication of predicates (or functions of functions).

Until someone produces an authoritative definition of first-order 
that says something else, I don't think it's ever "simply incorrect" 
to call RDFS higher-order.   It is "simply" correct.  It may be 
incorrect according to your (or Pat's) more complicated definition of 
what first-order means, but that is by no means "simple"!

I have claimed from the start that a useful distinction here is to 
say that RDFS is syntactically higher-order and semantically 
first-order.  Pat has not agreed.

More to the point, I believe it to be the case that RDFS is 
undecidable (has this been proven?), and certainly OWL DL is 
decidable (has this been proven?).  Therefore I think it may be more 
useful to make the diagram look like this:

RDFS -> (decidable fragment of RDFS) -> OWL Lite -> OWL DL -> OWL Full
   +--------------------------------------------------------------^

Chris - it's the linear nature of the above that causes a lot of the 
problem.  Given a piece of RDF(S) there is no reason to expect it to 
fall in Lite or DL unless you either have started from a particular 
subset of RDFS or unless some work is done to check/make it 
conformant.  For example, I would assume an OWL DL tool, given an 
arbitrary piece of RDFS as input, would either need to run a 
validator to see if it was appropriate, or would need to have some 
sort of front-end that added annotations, typed some nodes, etc.
  The point is the expectation should be that if you're already using 
RDF and add a little bit of the OWL vocabulary to what you're doing, 
you're likely to be in Full (which is fine).  If you then want the 
reasoning gaurantees, you would do some work to learn about, and put 
your documents into, Lite or DL.
  We need to set that expectation -- the exact labels on the fragment 
of RDFS aren't the important thing.  It's also worth noting that we 
don't seem to have a place where someone can look to see what the 
restrictions are that create the decidable (or FO or whatever) 
fragment of RDFS, and that is a problem that we will eventually need 
to fix (i.e. by adding Sean's document on RDF graphs to our stuff 
somewhere, or the like)
  -JH

-- 
Professor James Hendler				  hendler@cs.umd.edu
Director, Semantic Web and Agent Technologies	  301-405-2696
Maryland Information and Network Dynamics Lab.	  301-405-6707 (Fax)
Univ of Maryland, College Park, MD 20742	  240-731-3822 (Cell)
http://www.cs.umd.edu/users/hendler

Received on Thursday, 5 June 2003 11:33:59 UTC