W3C home > Mailing lists > Public > semantic-web@w3.org > June 2013

Re: Property chain axioms and cyclic dependencies

From: Michael F Uschold <uschold@gmail.com>
Date: Tue, 4 Jun 2013 11:53:38 -0700
Message-ID: <CADfiEMN27a35=9Q1EGkzNS0xMHCv5e1WGe6=AoH80_zsZg+aOw@mail.gmail.com>
To: Michel Dumontier <michel.dumontier@gmail.com>
Cc: Matías Parodi <mparodilabs@gmail.com>, "semantic-web@w3.org" <semantic-web@w3.org>
This sounds about right from my experience.  I found that when defining
property chains, I ran into problems and discovered that I could do any two
of the following three things:
* keep the property (PropA) defined as a property chain
* have PropA defined to be a subproperty of some other property
* use a min or max cardinality with 2 or more, or an exactly.

To repeat for emphosis, what this means is that if I did any of the
following, the inferencing error went away:
* don't have PropA as a property chain
* don't have PropA be a subproperty of anything
* only use min 1 or max 1, vs. 2 or more, or an exactly in restrictions

It was pretty mysterious behavior that is not described in an easy to
understand way in the spec (if you are not a logician) and is much better
explained above by Michael Dumontier.


On Sun, Jun 2, 2013 at 4:14 PM, Michel Dumontier <michel.dumontier@gmail.com
> wrote:

> The relevant section is here:
>
> http://www.w3.org/TR/owl2-syntax/#Global_Restrictions_on_Axioms_in_OWL_2_DL
>
> basically, you can't use transitive roles and composed roles (e.g. from
> property chain) in min/max/exactly/self axioms or in combination with
> functional, inverse functional, irreflexive, asymmetric and disjoint
> properties.  there's also restrictions on the property hierarchy, in that
> the composed roles can't lead to a cycle (must maintain partial order; see
> text for details), along with properties that have non-simple subproperties.
>
> m.
>
>
>
>
> On Sun, Jun 2, 2013 at 9:31 PM, Michael F Uschold <uschold@gmail.com>wrote:
>
>> This is tricky, read very carefully the OWL spec.  There are obscure and
>> bizare (from a common sense user perspective) restrictions that result from
>> the inference needs.  Good luck.
>>
>>
>> On Wednesday, May 29, 2013, Matías Parodi wrote:
>>
>>> Hello,
>>>
>>> I'm trying to understand in which cases it's allowed to use
>>> owl:propertyChainAxiom and when it's not.
>>>
>>> I've written a small ontology to understand it (I pasted it at the end),
>>> but I still couldn't figure out why HermiT (using Protege) says "The given
>>> property hierarchy is not regular" and "There is a cyclic dependency
>>> involving property :isRelative".
>>>
>>> I know there's a loop there but what is the real problem? After all, how
>>> is that different from a owl:TransitiveProperty? Any way to intuitively
>>> understand what is wrong with it?
>>>
>>> Any idea?
>>>
>>> Thank you,
>>> Matt
>>>
>>> ---
>>> @prefix : <http://foo/bar#>.
>>> @prefix owl: <http://www.w3.org/2002/07/owl#>.
>>> @prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#>.
>>> @prefix rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>.
>>> @prefix xsd: <http://www.w3.org/2001/XMLSchema#>.
>>>
>>> :Person a owl:Class.
>>>
>>> :isRelativeOf
>>>   a owl:TransitiveProperty;
>>>   rdfs:domain :Person;
>>>   rdfs:range :Person;
>>> ..
>>>
>>>   :isAncestorOf
>>>     rdfs:subPropertyOf :isRelativeOf;
>>>     owl:propertyChainAxiom (:isRelativeOf :isAncestorOf);
>>>   .
>>> ---
>>>
>>
>>
>> --
>> Sent from an input-challenged device.
>>
>
>
>
> --
> Michel Dumontier
> Associate Professor of Bioinformatics, Carleton University
> Chair, W3C Semantic Web for Health Care and the Life Sciences Interest
> Group
> http://dumontierlab.com
>



-- 

Michael Uschold
   Senior Ontology Consultant, Semantic Arts
   http://www.semanticarts.com
   LinkedIn: http://tr.im/limfu
   Skype, Twitter: UscholdM
Received on Tuesday, 4 June 2013 18:54:05 UTC

This archive was generated by hypermail 2.4.0 : Tuesday, 5 July 2022 08:45:33 UTC