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

Re: Reified relations in OWL

From: Alan Rector <rector@cs.man.ac.uk>
Date: Mon, 1 Mar 2010 12:34:12 +0000
Cc: "Uli Sattler" <sattler@cs.man.ac.uk>, <public-owl-dev@w3.org>
Message-Id: <10763DA3-C303-49AB-BE13-FBCC3876BFC4@cs.man.ac.uk>
To: Paul Oude Luttighuis <Paul.OudeLuttighuis@novay.nl>
Paul


On 26 Feb 2010, at 16:16, Paul Oude Luttighuis wrote:

> Dear Uli,
>
> Thanks for your swift reply.
>
> I'm hesitant about extensively elaborating on our modelling approach  
> in this forum. It's not the place. If you like, we can do that via  
> direct e-mail.
>
> But, to answer your last question: basically, I would like relations  
> to be able to participate in relations themselves. In fact, in the  
> modelling approach we study, this is the rule, not the exception.  
> Relations are hence first-class citizens, or even, the only  
> citizens ... Each concept is represented by a relation that connects  
> the concepts it is existence-dependent upon. So, it requires  
> relations to have their own instances. This is what I know as  
> "reified relations".
>
> We are looking for modelling languages that might nicely express  
> such reified-relations-only models. OWL is only one of them. Our  
> first concern is not reasoning, but plain representation, preferably  
> with a nice graphical syntax.
>
> Does this help?

There are at least two ways to model reified relations:

a) Focusing on the endpoints, in which case you get expressions in OWL  
of the form

	C1 has-relation some (Relation that has_target some C2)

This says that every C1 is related to a Relation that is related to  
some C2.

b) Focusing on the relations themselves.

	R1 has_topic some C1 AND has_target some C2.

This says that every relation R1 is related to a C1 and a C2, and  
implies
that some C1s are related to some C2s via R1 (provided any R1s exist).

For many purposes it is more convenient and semantically correct than  
the version in the
n-ary relations note, since it makes it easy to define the domain and  
range of R1, if that
is the task, and to create whatever relation and class hierarchies are  
required.

As for nice graphical syntaxes, I am afraid that is still a serious  
lack in OWL, perhaps
appropriately, since OWL is sufficiently low level that it can be used  
in many different
ways.  It is unlikely that any one syntax would be suitable for all  
purposes.  Nonetheless,
I agree that it would be good to have a generic OWL graphical syntax  
but can't provide one.

Regards

Alan


>
> Regards,
>
> Paul
>
> From: Uli Sattler [mailto:sattler@cs.man.ac.uk]
> Sent: vrijdag 26 februari 2010 15:24
> To: Paul Oude Luttighuis
> Cc: public-owl-dev@w3.org
> Subject: Re: Reified relations in OWL
>
>
> On 25 Feb 2010, at 21:53, Paul Oude Luttighuis wrote:
>
>> Dear OWL mailing list,
>>
>> In a current project, we take an approach to semantic modelling  
>> based on reified relations only. In that context, I am looking for  
>> ways of using reified relations in OWL. I am aware of the ways of  
>> mimicking reified relations proposed in http://www.w3.org/TR/swbp-n-aryRelations/ 
>> , but these yield awkward models. Also, such mimicking takes away  
>> the graphical visualisations we need, as these tend (as in Protégé)  
>> to visualise the subtyping hierarchy only, not the relations  
>> (properties). In our case, subtyping is refrained from/not  
>> interesting.
>
> sorry for skipping over your initial question: why would you refrain  
> from 'subtyping'?! The fact that you use the term 'subtyping' seems  
> to indicate a misunderstanding of the OWL semantics...so perhaps  
> this should be clarified first: e.g., in OWL, if you define the  
> class "Waterbirds" as those Birds who live close to water, and Ducks  
> as those Birds who say quack and live close to water, then the  
> subclass relationship between Ducks and Waterbirds will be entailed,  
> and thus inferred by the reasoner...as a consequence, every instance  
> of Duck will also be an instance of Bird, automatically, simply due  
> to the OWL semantics...
>
> Now, perhaps you explain your 'reified relation' scenario a bit  
> more: I assume you want to model n-ary relations/tuples, for n>2? Do  
> you have an example?
>
> Cheers, Uli
>
>>
>> Am I overlooking something. Can anybody point me to more natural  
>> ways of modelling and graphically representing reified relations  
>> with OWL?
>>
>> I'd be grateful for your suggestions.
>>
>> Best regards!
>> Paul
>>
>> POSTADRES
>>
>> BEZOEKADRES
>> MOBIEL
>> TELEFOON
>> FAX
>> INTERNET
>>
>> Postbus 589
>> 7500 AN Enschede
>> Brouwerijstraat 1
>> +31 65 088 17 54
>> +31 53 485 04 18
>> +31 53 485 04 00
>> www.novay.nl	<Novay Signature NL + Mobiel.1>
>>
>>
>> Paul Oude Luttighuis
>> principal researcher & consultant
>>
>> E-MAIL
>>
>> Paul.OudeLuttighuis@novay.nl	
>>
>>
>>
>>
>>
>>
>> Verbinden, Ontmoeten, Inspireren, Ontdekken, http://co-inspire.novay.nl
>>
>>
>

-----------------------
Alan Rector
Professor of Medical Informatics
School of Computer Science
University of Manchester
Manchester M13 9PL, UK
TEL +44 (0) 161 275 6149/6188
FAX +44 (0) 161 275 6204
www.cs.man.ac.uk/~rector
www.co-ode.org
http://clahrc-gm.nihr.ac.uk/
Received on Monday, 1 March 2010 12:34:44 GMT

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