- From: Ian Horrocks <horrocks@cs.man.ac.uk>
- Date: Thu, 25 Sep 2003 10:31:17 +0100
- To: Jim Hendler <hendler@cs.umd.edu>
- Cc: webont <www-webont-wg@w3.org>
On September 24, Jim Hendler writes:
> Two things that have come up during CR are causing me "concern" as a chair -
>
> 1 -
>
> At 12:19 PM -0400 9/24/03, Peter F. Patel-Schneider wrote:
> >The incorrect proposed syntactic level test
> >concerns the use of datatypes, where the test suite diverges from S&AS.
>
> This makes me nervous as both S&AS and Test are supposed to be
> normative, and they seem to disagree. Jeremy, Peter - I know this
> has been discussed in the mailing list -- does it need WG action or
> is it something easily resolvable?
>
> 2 -
>
> There are a very small number of tests that have not yet been passed
> - however, two of them are:
>
> [Results] miscellaneous-001 levels:DL, Full [APPROVED: Med L XXL]
> Wine example taken from the guide.
>
> [Results] miscellaneous-002 levels:DL, Full [APPROVED: Med L XXL]
> Food example taken from the guide.
Both these ontologies are in DL, mainly because they use the oneOf
constructor. As they also use cardinality constraints and inverse
properties, they are in that fragment of the language for which we
know that we don't currently have a practical decision procedure
(i.e., a sound, complete and terminating consistency testing
algorithm). For this language, all of the existing systems are (at
best) sound but incomplete w.r.t. *inconsistency*, i.e., they can
sometimes prove inconsistency, but if they fail to do so then we know
nothing. Given that the tests in question are consistency tests, none
of these systems can be expected to report pass.
Note that it would in principal be possible to design an algorithm
that was sound but incomplete for *consistency* in this language, but
I don't know of anyone who is working on this aproach.
If the use of individuals in oneOf constructs is eliminated in favour
of unions of classes, then the ontologies could (in principal) be
proved consistent by several of the existing systems. (In fact with
such a modification we are able to completely classify the wine
ontology in about 10s and the food ontology in less than 1s using the
latest version of the FaCT++ system.) I believe that we should make
such a change: the extensive use of oneOf in the wine and food
ontologies is largely gratuitous (it seems to be mainly the result of
their origin in a language that supported this constructor but did not
support unions of classes), and is setting a bad example to
prospective users - it encourages the use of statements that are, in
most cases, stronger than is needed/intended, and that are known to be
difficult to reason with.
One further point. Given the elimination of oneOf, then the wine and
food ontologies could even be transformed into OWL Lite, although this
would result in some mangling of the syntax (in order to capture
negation and disjunction).
Regards, Ian
>
> both of which are DL tests taken from examples from the Guide. In
> the worst case we could always make the tests extra credit or simply
> change the Guide a bit -- but as Chair it concerns me that a document
> created by our own WG to be an example of a typical use of OWL is
> generating a test that none of our reasoners seem to be able to pass.
> It might be worth the WG taking a look at these tests, seeing why
> they are hard to pass, and deciding if anything in either our design
> or our documents needs reworking (for those not following the Tests -
> these tests are more or less equivalent to proving that wine.owl and
> food.owl are consistent documents).
> In particular, I could sleep better at night if I understood why
> some other large ontologies (like galen.owl) are not causing problems
> but this one is -- and whether anyone has reasoners that are likely
> to be able to handle our own best example!
>
> -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 *** 240-277-3388 (Cell)
> http://www.cs.umd.edu/users/hendler *** NOTE CHANGED CELL NUMBER ***<!doctype html public "-//W3C//DTD W3 HTML//EN">
> <html><head><style type="text/css"><!--
> blockquote, dl, ul, ol, li { padding-top: 0 ; padding-bottom: 0 }
> --></style><title>Chair's Concerns re: Test Cases and
> CR</title></head><body>
> <div>Two things that have come up during CR are causing me
> "concern" as a chair -</div>
> <div><br></div>
> <div>1 -</div>
> <div><br></div>
> <div>At 12:19 PM -0400 9/24/03, Peter F. Patel-Schneider wrote:</div>
> <div>>The incorrect proposed syntactic level test</div>
> <div>>concerns the use of datatypes, where the test suite diverges
> from S&AS.</div>
> <div><br></div>
> <div>This makes me nervous as both S&AS and Test are supposed to
> be normative, and they seem to disagree. Jeremy, Peter - I know
> this has been discussed in the mailing list -- does it need WG action
> or is it something easily resolvable?</div>
> <div><br></div>
> <div>2 -</div>
> <div><br></div>
> <div>There are a very small number of tests that have not yet been
> passed - however, two of them are:</div>
> <div><br></div>
> <div>[Results] miscellaneous-001 levels:DL, Full [APPROVED: Med L
> XXL] Wine example taken from the guide.<br>
> <br>
> [Results] miscellaneous-002 levels:DL, Full [APPROVED: Med L XXL] Food
> example taken from the guide.</div>
> <div><br></div>
> <div>both of which are DL tests taken from examples from the
> Guide. In the worst case we could always make the tests
> extra credit or simply change the Guide a bit -- but as Chair it
> concerns me that a document created by our own WG to be an example of
> a typical use of OWL is generating a test that none of our reasoners
> seem to be able to pass. </div>
> <div> It might be worth the WG taking a look at these tests,
> seeing why they are hard to pass, and deciding if anything in either
> our design or our documents needs reworking (for those not following
> the Tests - these tests are more or less equivalent to proving that
> wine.owl and food.owl are consistent documents). </div>
> <div> In particular, I could sleep better at night if I
> understood why some other large ontologies (like galen.owl) are not
> causing problems but this one is -- and whether anyone has reasoners
> that are likely to be able to handle our own best example!</div>
> <div><br></div>
> <div> -JH</div>
> <div><br></div>
> <div><br></div>
> <div><br></div>
> <x-sigsep><pre>--
> </pre></x-sigsep>
> <div>Professor James Hendler<x-tab>
> </x-tab><x-tab>
> </x-tab><x-tab>
> </x-tab><x-tab> </x-tab>
> hendler@cs.umd.edu<br>
> Director, Semantic Web and Agent
> Technologies<x-tab> </x-tab>
> 301-405-2696<br>
> Maryland Information and Network Dynamics Lab.<x-tab>
> </x-tab> 301-405-6707 (Fax)<br>
> Univ of Maryland, College Park, MD 20742<x-tab>
> </x-tab> *** 240-277-3388 (Cell)<br>
> http://www.cs.umd.edu/users/hendler ***
> NOTE CHANGED CELL NUMBER ***</div>
> </body>
> </html>
Received on Thursday, 25 September 2003 05:33:33 UTC