W3C home > Mailing lists > Public > public-owl-wg@w3.org > November 2007

Re: Rich Annotations Use Cases

From: Vojtech Svatek <Svatek@vse.cz>
Date: Tue, 13 Nov 2007 01:44:38 +0100
To: "Peter F. Patel-Schneider" <pfps@research.bell-labs.com>
Cc: public-owl-wg@w3.org
Message-ID: <OF9BDFB110.F703291A-ONC1257392.00041615-C1257392.0004163A@kotelna.vse.cz>

Hi Peter,

(coming back after a delay caused by my own illness plus mailbox crash...)

You are right that things like 'reified n-ary properties' *can* be modelled
in different ways (though you only showed one alternative). The question
however is what *good* modelling practice looks like.

The notion of 'reified n-ary property' is closely linked to the ontology
language expressivity, and, consequently, to the core of the
conceptualisation-coding process. It is not just a very specific feature to
be included 'in advance' so as to be possibly used by some arcane tools at
later phase (and thus only deserving to be imported through an ontology
specifically tailored to the needs of these tools). Maybe there are indeed
other alternatives to using annotations - I don't stick to this one - but
that you mentioned is not the right thing IMHO.

To me, the lack of explicit n-ary properties is, in a way, a deficiency of
OWL wrt. other languages often used for 'ontological' modelling purposes. I
understand that complying with DL semantics is critical. On the other hand,
I believe that simple steps that ease compatibility with other, less
AI-ambitious modelling languages (such as UML or Topic Maps) in this and
other respects might improve the perception of OWL by kind-of external

All such 'pattern-oriented' annotation types I have in mind are, in Bijan's
terms, 'canIgnore' ones (although I am also interested in e.g. uncertainty
extensions - but that's another story).

My plan now is to have a bit of discussion with some 'patterns' guys first,
offline, and then return with some consolidated proposal. Bijan, would it
be OK if I then add a few bullets to [1] at some point?



[1] http://www.w3.org/2007/OWL/wiki/Annotation_System#Examples

  Vojtech Svatek, University of Economics, Prague
  Nam.W.Churchilla 4, 13067 Praha 3, CZECH REPUBLIC
  (CHANGED!!!) phone: +420 224095495, e-mail: svatek@vse.cz
  web: http://nb.vse.cz/~svatek

-----"Peter F. Patel-Schneider" <pfps@research.bell-labs.com> napsal: -----

>Komu: Svatek@vse.cz
>Od: "Peter F. Patel-Schneider" <pfps@research.bell-labs.com>
>Datum: 06.11.2007 14:26
>Kopie: public-owl-wg@w3.org
>Předmět: Re: Rich Annotations Use Cases
>From: Vojtech Svatek <Svatek@vse.cz>
>Subject: Re: Rich Annotations Use Cases
>Date: Wed, 31 Oct 2007 22:08:30 +0100
>> Dear all,
>> sorry if I am wrong - I am entirely new to the WG and may not have
>> understanding of many issues - but it seems to me that annotations
>could be
>> used for a lot of interesting and 'semantic' purposes connected to
>> patterns, visualisation etc.
>> For example:
>> - indicating that a certain concept is a reified n-ary
>relationship, cf.
>> http://www.w3.org/TR/swbp-n-aryRelations/
>> - naming the feature that is responsible for a class partition (or
>> generally subclassing), e.g. if the class Person is partitioned to
>> Employee, Student and Retired, the partition could be labelled as
>> 'professional status' or the like.
>> It is important that such distinctions are not only consumable by
>> When (semi-)automatically mapping ontologies, it often occurs that
>> parts of a domain are modelled differently (say, using different
>> patterns) in different ontologies, and such additional information
>> help a lot I believe.
>> Regards
>> Vojtech
>Sure, but how would this work, and what support from the WG is
>I can imagine doing this already in a number of ways, some of which
>utilize annotations and some of which don't.
>For an example of a way that needs neither annotations nor WG
>you could just do something like:
>    Declaration(OWLClass(ex:n-ary-relation-superclass))
>and make concepts representing n-ary relations subclasses of this
>structural class, even use classes-as-instances and make the class
>instances of a particular individual, perhaps ex:n-ary-relation.
>The extra-OWL import of these special names would then be used by UI
>ontology integration tools.
>NB: I'm not advocating these as reasonable solutions, just pointing
>that there can be other ways of providing "structural" information
Received on Tuesday, 13 November 2007 00:45:34 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:42:00 UTC