Re: CR/PR questions

Jeremy, with due respect, I'm afraid I don't actually understand your 
points below.  You say you can encode your problem in OWL, that Ian 
has implemented some tests, and that arithmetic is hard for a pure 
reasoner.  Which of these keep us from moving to PR?
  If your contention is that there are some things for which current 
systems can handle, but optimizations will be needed for larger 
community acceptance, I probably wouldn't disagree, but also wouldn't 
think these hold up the move to PR  - optimizations generally come 
later in the day for these sorts of specs - what's more, there are 
already a couple of implementations where it appears to me that OWL 
reasoners can call out to other processes for doing artihmetic (or 
anything else for which there is a well-known process that is believe 
to be logically consistent) and thus it is a question of these being 
cited and more integrated.
   - jh


At 21:40 +0300 5/4/03, Jeremy Carroll wrote:
>OWL DL seems to differ from OWL Lite in two important respects:
>1) it allows unnamed descriptions and the full remit of set theoretic
>constructions
>
>2) it has the owl:oneOf and owl:hasValue constructions, and hence finite
>classes
>
>3) higher cardinalities, which interact with finite classes in interesting
>ways.
>
>The first seems trivial from an implementation point of view: these things can
>be converted into OWL Lite.
>
>However the second seems to permit a wide range of axioms that are not wdiely
>used in DAML+OIL experience nor over which there seems to be much real
>practical experience of reasoning.
>
>In the TEST editors draft at the moment we have some 3-SAT tests
>
>http://www.w3.org/2002/03owlt/editors-draft/draft/proposed-dl-500-SAT
>
>these are, from a specialist 3-SAT reasoner point of view trivial, (i.e. they
>solve in less than a second) but not ones for which I expect a queue of OWL
>implementations
>
>I am working on expressing some of the Venn diagram problems I have worked on
>in OWL, for instance, the attached picture can be described in OWL DL to a
>surprising degree of accuracy (e.g. its vennness, i.e. there are 5
>overlapping shapes giving 32 distinct regions, the triangularness, in terms
>of 15 intersecting pseudolines etc).
>
>These descriptions depend on giving multiple names to the same geometric
>object (face, edge, line, point), and leaving the reasoner to sort out the
>tangle.
>
>On cardinality, it is possible to embed substantial portions of arithmetic
>within OWL DL, and a practical reasoner intending to process this language
>should be aware of cardinality.
>
>Ian reported having successfully completed the addition tests 901 thru 904
>
>http://www.w3.org/2002/03owlt/editors-draft/draft/proposed-dl-900-arith#proposed-dl-900-arith
>
>however, his reasoner is currently doing this long hand and he had to leave it
>running overnight to work out that 200+300 != 600 because the numbers are
>large.
>
>I am not compelled that Ian will manage when I ask him if 200*300=60000
>
>Jeremy
>
>
>
>Attachment converted: Ontology:venn5.svg (xxxx/«IC») (00136856)
>Attachment converted: Ontology:venn5.png (PNGf/prvw) (00136857)

-- 
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 Monday, 5 May 2003 15:24:14 UTC