Re: implementation report for DL syntax checker

From: "Jeremy Carroll" <jjc@hplb.hpl.hp.com>
Subject: RE: implementation report for DL syntax checker
Date: Mon, 11 Aug 2003 10:53:58 +0200

> Peter:
> > I wrote an OWL DL syntax checker from the CR version of S&AS
> > and ran it over all the OWL tests.
> 
> Congrats!
> 
> >
> > The checker was run on all files in accepted tests that were marked as
> > being in OWL Lite or OWL DL or not being in OWL DL.
> 
> How did you do on the proposed tests - the ones that have not yet been
> successfully checked by any system are those added for the "feature at risk"
> in the CR request. It would be good to know whether I got those right or
> not.

To be done some time this week, if possible.

> > I didn't run the
> > checker on files marked as not being in OWL Lite --- these are probably
> > also not in OWL DL, but this is not inferrable from the test manifest.
> 
> I am not sure what you mean - the intent is that each file mentioned in the
> manifest is labelled as Lite DL or Full. The test itself may have a
> different label, but that's typically more aimed at the syntactic checking.

Several documents are labelled as being invalid for OWL Lite.  As I have an
OWL DL syntax checker, I can't process these tests.  However, the documents
are also probably not in OWL DL as well, and should be so labelled. 

> > The only mistakes that the checker made were on
> > http://www.w3.org/2002/03owlt/miscellaneous/consistent001 and
> > http://www.w3.org/2002/03owlt/miscellaneous/consistent002 where the
> > underlying XML engine that I am using (Galax) choked.  I expect to get a
> > new version of Galax next week and will see if these are then correctly
> > processed.
> 
> There was an XML error in those two files that Sean reported off-list, and
> has now been fixed.

I'll try them again sometime soon.

> > peter
> >
> > PS:  Of course it may be that my syntax checker has errors and they are
> > just not exercised by the approved OWL tests.
> >
> > PPS: It was more difficult than necessary to determine how to process the
> >      manifest file.
> 
> I hope to get round to fixing that at some point, what do you think of the
> plan sketched in:
> 
> http://lists.w3.org/Archives/Public/www-webont-wg/2003Jul/0218.html

This would cut down on the clutter.  The problem I had was determining
which classes and properties to look for.  Perhaps this is written down in
the manifest file itself and I can't see it for the clutter, but a document
making this easier would be helpful.

> donations accepted.

Would you like an air conditioner?  :-)

> Jeremy

peter

Received on Monday, 11 August 2003 06:03:45 UTC