W3C home > Mailing lists > Public > public-owl-dev@w3.org > July to September 2010

RE: How to put "universal" restrictions to members of two classes

From: Michael Schneider <schneid@fzi.de>
Date: Mon, 6 Sep 2010 19:57:17 +0200
Message-ID: <0EF30CAA69519C4CB91D01481AEA06A00211762C@judith.fzi.de>
To: "Yoshio FUKUSHIGE" <fukushige.yoshio@jp.panasonic.com>
Cc: <public-owl-dev@w3.org>
Hi Yoshio!

>Hi Michael,
>Sorry for the delay in responding.
>
>> No, it was correct that sub property chain axioms only take object
>> properties. It is just that in OWL (2) Full all (RDF) properties
>happen to
>> be object properties. Even "rdf:type", as you use it in the property
>chain
>> axiom of your proposed solution, is an object property in OWL Full.
>
>This is because thre is no distinction between individuals and classes
>in OWL Full, right?

It is true that classes and properties are individuals in OWL Full (there is
no strict separation between the different entity types as in OWL DL). But
this is not the relevant point in your concrete usecase. The point is that
OWL Full treats rdf:type as an ordinary property, so you can use it as such
in property chain axioms. In OWL DL, however, rdf:type is /not/ an object
property (at least not in the RDF serialization of OWL DL), so you can't use
it as such.
 
>> Of course, you can still use the term "rdf:type" in property chain
>axioms
>> directly in the OWL 2 syntax, and the OWL 2 Direct Semantics will even
>> interpret it, since the OWL 2 Direct Semantics allows for interpreting
>the
>> whole OWL 2 syntax, not only the particular fraction called "OWL 2 DL"
>(that
>> is, the OWL 2 syntax with the OWL 2 DL constraints being applied; see
>the
>> beginning of Chap. 3 of the OWL 2 Structural Specification). But this
>will
>> still not give you the expected reasoning result, since that
>"rdf:type"
>> property is then only some ordinary object property without any
>connection
>> to class assertion axioms.
>
>Is that because a reasoner doesn't have to be conformant to the RDF
>Based Semantics
>which does define semantics of the term "rdf:type", but can be
>conformant only
>to the Direct Semantics which says nothing about the term "rdf:type"?

The Direct Semantics interprets OWL ontologies according to the OWL 2
Structural Specification, not according to the RDF syntax. The Structural
Specification does not specially treat the name "rdf:type", nor does the
Direct Semantics. For the Structural Specification as for the Direct
Semantics, "rdf:type" is just some name that one can use for all kinds of
entities, including object properties. 

However, a sub-language of the Structural Specification is the OWL 2 DL
syntax, which puts a couple of restrictions to the OWL 2 syntax (without
"DL"!), including disallowing the use of the name "rdf:type" for semantic
entities. It is a bit complicated, and most people will probably not know
about this distinction. Anyway, the bottom line is that the Direct Semantics
does not stop anyone from using "rdf:type" to name an object property but,
unlike the RDF-Based Semantics, it will not give an instanceship-related
meaning to this property.

>> >But does that mean my statements were valid in OWL2 RL??
>>
>> If by "valid in OWL 2 RL" you mean that it was a valid OWL 2 RL
>ontology,
>> then the answer is no. The OWL 2 RL (syntactic!) profile is a sub-
>syntax of
>> the OWL 2 DL syntax. Hence, since your input has not been a valid OWL
>2 DL
>> ontology, it cannot be a valid OWL 2 RL ontology.
>>
>> That you can apply the OWL 2 RL/RDF rules does not mean that you are
>> restricted to the OWL 2 RL syntax. The OWL 2 RL/RDF rules define a
>semantics
>> (defined by the set of first-order logic axioms the rule set consists
>of)
>> and can formally be applied to arbitrary RDF, even to "generalized"
>RDF
>> (e.g. allowing for literals in subject position).
>
>I see. (thogh a kind of puzzled)
>
>> >From looking at it, Pavel's solution seems to work well in OWL 2 DL.
>So go
>> with this solution if you need to use Pellet or some other OWL 2 DL
>> reasoner, or if you want/need to stay in OWL 2 DL. However, I remember
>that
>> you asked whether there can be a specific property between two classes
>which
>> indicates - and allows for inferring - a certain relationship between
>all
>> pairs of instances of these classes. Your own OWL 2 Full solution was
>pretty
>> obvious (from an OWL 2 Full perspective, at least) and really worked.
>So
>> it's now kind of a close-modeling-vs-constraining-technology tradeoff,
>and
>> you have to decide for yourself what has priority for you. In any way,
>you
>> are in the lucky position that there are working solutions for both
>> approaches.
>
>I have to consider the scalability to some extent and I, for my shame,
>don't
>know how to use concrete reasoners of OWL 2 Full/RL with my Joseki.
>So, I'm going to try with the approach suggested by Pavel,
>trying to give some mnemonics to the 'brige' entities/properties,
>to make the developing ontologies easier especially putting N-ary
>(all to all) relationships.

Use the system that works for you! In particular, while your example
happened to work with the OWL 2 RL/RDF rules, other portions of your
ontology might not work as expected, since the RL rules are not very
expressive compared to OWL DL, specifically when it comes to terminological
reasoning. And real OWL Full reasoners (in the sense of "close to complete
for many practical applications") do not yet exist, and are unlikely to
become competitive with DL reasoners concerning speed and scalability, I
guess. But, for your interest, there are several RL rule reasoners listed on
the OWL 2 Implementations page at [1], one of them based on Jena. So you
might, at one time, give it a try. RDF rule reasoners (not only those
conforming to the OWL 2 RL/RDF rules), while not being very expressive
semantically, tend to be pretty scalable. But you have to check yourself for
your concrete usecase.

>Thank you very much for your kind advices.
>
>Best,
>Yoshio

Cheers,
Michael

[1] <http://www.w3.org/2007/OWL/wiki/Implementations>

--
Dipl.-Inform. Michael Schneider
Research Scientist, 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
=======================================================================
Received on Monday, 6 September 2010 17:57:53 GMT

This archive was generated by hypermail 2.3.1 : Wednesday, 27 March 2013 09:32:58 GMT