Re: release candidate of the "rdfbased" testcollection

Ian Horrocks wrote:
> Excellent! I look forward to hearing about the results with Ivan's
> implementation and to seeing these posted on the results page [1].

It has already revealed some small problems in the last few days, but
the version on the web is now o.k. as far as I know...

Just my own impressions as implementer; nothing really new for this
group on the technical level, but good to know if others come around to
possibly implement the RL rule set with RDF in mind...

Implementing the core rule set is, in general, close to trivial once
some sort of a general framework is around (either a rule system or some
sort of a pedestrian framework to handle the rules procedurally). Some
rules (eg, the property chains) are a little bit trickier, so it is a
good exercise for an undergraduate:-)

There are some hiccups, however:

- if one relies on an existing RDF environment (as I did) then the fact
of using 'generalized' triples can be tricky. One should be able to
handle triples with blank node predicates and literals in a subject
position. (I was lucky for the first one, because RDFLib does not yell
at me if I have a BNode as a predicate; for literals I used the trick
described in the RDFS document to exchange all literals with a BNode at
the start of processing and switch them back at the end. So these worked
after all, but one has to be prepared for something like that.)

- handling the datatypes may be tricky (we did have a bunch of mails
between me and Michael the last few days). If I want to check, eg, the
disjointness of the various datatypes (more exactly, whether some
literals break this), whether the lexical form is kosher or not, etc, it
is very much dependent on what the underlying environment offers. I am
busy trying to add a minimal number of extra triples for, eg, the
disjointness issue so that the RL rule processing would automatically
reveal problems with the data, but this is not efficient. ie, if one
relies on existing RDF environment, the more precise nature of datatype
handling in OWL 2 is an extra implementation burden for RL (though
probably well worth it for users!). I am not saying we should change
that, but the value of the CR phase is to gather implementation
experience so, well, that is it...

Ivan


> 
> Thanks,
> Ian
> 
> [1] http://www.w3.org/2007/OWL/wiki/Test_Suite_Status
> 
> 
> On 25 Jun 2009, at 00:47, Michael Schneider wrote:
> 
>> Hi!
>>
>> The testcollection for the RDF-Based Semantics, which I have announced
>> several times in the past, is now essentially finished after two
>> months of work. I have (again) attached only the source version
>> (easier to inspect), and will have to translate all the testcases into
>> the final Wiki-Page format and upload them to the Wiki in the
>> following days.
>>
>> One interesting aspect of this final version of the testcollection is
>> that it additionally covers the OWL 2 RL ruleset. So there is now a
>> dedicated subset of testcases (ca. 40%) being explicitly marked as
>> "owl2rl", which means that these testcases have been selected based on
>> the different rules in Section 4.3 of the Profiles spec.
>>
>> However, since conformant OWL RL entailment checkers may provide
>> entailments beyond the ruleset up to what the whole OWL 2 RDF-Based
>> Semantics provides, one can even regard the /complete/ testcollection
>> as an OWL RL test suite. Implementers of RL systems, who want to
>> provide a more expressive but still conformant RL reasoner can
>> therefore apply the whole testcollection to get a coarse notion of
>> where their implementations stands "on the path" between the RL
>> ruleset and the whole of OWL 2 Full.
>>
>> The testcases have been carefully audited, but I am going to apply
>> them to Ivan's RL implementation, which might then still reveal some
>> errors (on whichever side :)).
>>
>> Cheers,
>> Michael
>>
>> -- 
>> Dipl.-Inform. Michael Schneider
>> Research Scientist, Dept. Information Process Engineering (IPE)
>> Tel  : +49-721-9654-726
>> Fax  : +49-721-9654-727
>> Email: michael.schneider@fzi.de
>> WWW  : http://www.fzi.de/michael.schneider
>> =======================================================================
>> FZI Forschungszentrum Informatik an der Universität Karlsruhe
>> Haid-und-Neu-Str. 10-14, D-76131 Karlsruhe
>> Tel.: +49-721-9654-0, Fax: +49-721-9654-959
>> Stiftung des bürgerlichen Rechts, Az 14-0563.1, RP Karlsruhe
>> Vorstand: Prof. Dr.-Ing. Rüdiger Dillmann, Dipl. Wi.-Ing. Michael Flor,
>> Prof. Dr. Dr. h.c. Wolffried Stucky, Prof. Dr. Rudi Studer
>> Vorsitzender des Kuratoriums: Ministerialdirigent Günther Leßnerkraus
>> =======================================================================
>>
>>
>> <tests-rdfbased-090624.zip>
> 
> 

-- 

Ivan Herman, W3C Semantic Web Activity Lead
Home: http://www.w3.org/People/Ivan/
mobile: +31-641044153
PGP Key: http://www.ivan-herman.net/pgpkey.html
FOAF: http://www.ivan-herman.net/foaf.rdf