RE: [OWLWG-COMMENT] Example why current RDF mapping for QCRs might hurt OWL-1.1-Full

Hi Jim!

Jim Hendler wrote:

><flame on - but not at Matthew>
>-but  I note also that there are many widely used KR features 
>lefts out of OWL because the WG either does not have consensus 
>on how to define, or apparently, now things are left out if 
>they aren't decidable in OWL DL (I note the most used OWL 
>feature of any ontology to date, the inversefunctional email 
>sha-sum of FOAF, would not have been accepted under this rule) 

So this would be an InverseFunctionalDataProperty (IFDP), which indeed is
not in the current OWL-1.1-DL draft. But I want to note that, AFAICS, this
topic is really considered to be a serious topic by several people:

  * There is an OWLED taskforce dedicated to this topic:
    <http://code.google.com/p/owl1-1/wiki/DatabasEsque>

  * Bijan had a report on the current state of the discussion, starting with
 
<http://lists.w3.org/Archives/Public/public-owl-dev/2007JulSep/0211.html>
    (but I'm afraid this discussion got a bit lost at this time...)

  * There have been slides by Uli at the WG's F2F which cover this topic:
    <http://lists.w3.org/Archives/Public/public-owl-wg/2007Dec/0088.html>

  * Uli has made a pragmatical proposal for IFDPs in
    <http://lists.w3.org/Archives/Public/public-owl-wg/2007Dec/0035.html>

I cannot do more than just pointing you to these resources, because I am
pretty much behind in this topic.

>-- there are lots of such examples Part/whole being the 
>obvious it is used in the gene and OBO ontologies, and in 
>most of the scientific ontologies I've seen for physics, 
>astronomy, and geology - but the OWL 1.0 WG punted on it and 
>the OWL 1.1 hasn't even considered it.

I think "part/whole" would have to be either a property characteristic, or
there would have to be a dedicated property with the "part/whole" semantics,
from which more specific properties would have to be sub properties. 

I wouldn't have a good feeling for both of these options, because I do not
see that "part/whole" is always understood in the same way in all possible
situations. For example, sometimes I will regard something as a part of
itself, and sometimes not. And sometimes I will regard a "part/whole"
relation to be transitive, and sometimes this sounds absurd to me. And I may
or may not find it natural to regard a specific "part/whole" relation in
connection with another semantic relation like "owns".

After all, OWL-1.1 gives me at least some formal tools to model my own
"part/whole" relation. For example:

  * isPartOf a owl:TransitiveObjectProperty
    ("if a is part of b, and b is part of c, then a is part of c")

  * isPartOf a owl:AsymmetricObjectProperty
    ("if a is part of b, then b is *not* part of a")

  * isPartOf o hasOwner <= hasOwner
    (a sub property chain:
     "parts of things owned by X are themself owned by X")

I'm afraid that not everything, which might make sense, is actually possible
to model. For instance, I cannot model antisymmetrical relations at the
moment. But the situation has much improved compared to OWL-1.0.

> I'd like to see the WG 
>being a bit more consistent with respect to deciding what will 
>and will not be used.  I see the current 1.1 design as 
>something of a hodgepodge, with KR considerations being far 
>more weighted than other issues - decidability has become a 
>requirement (without any rewrite of the requirements document 
>that I ca find) 

This is the heretic "Consider slaughtering the holy decidability cow!"
discussion I have tried to start two times in the past:

  *
<http://lists.w3.org/Archives/Public/public-owl-dev/2007OctDec/0142.html>

  *
<http://lists.w3.org/Archives/Public/public-owl-dev/2007OctDec/0170.html>

But I want to make it clear that this was never meant by me to be a topic
for the current OWL-1.1 WG. As I understand it, it was the idea for this WG
to build a robust incremental update to OWL-1.0. And I believe that
carelessly making OWL-1.1 undecidable may easily make it unusable, or at
least it may result in unpredictable effects later in real-life use. I
believe that a lot of research has to be done before it will be possible to
come up with an undecidable OWL-DL successor which works well in practice.
That's simply out of scope for the current WG, but it might perhaps be an
interesting topic for future OWLED workshops!?

>and people are doing what I see as very 
>arrogant, such as people telling the Oracle rep what it is 
>that his customers should want - rather than listening to 
>Oracle explain what Oracle needs and wants
><flame off>
>
>sorry had to get that off my chest 
> -JH

Regards,
Michael

--
Dipl.-Inform. Michael Schneider
FZI Forschungszentrum Informatik Karlsruhe
Abtl. Information Process Engineering (IPE)
Tel  : +49-721-9654-726
Fax  : +49-721-9654-727
Email: Michael.Schneider@fzi.de
Web  : http://www.fzi.de/ipe/eng/mitarbeiter.php?id=555

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 Regierungspräsidium Karlsruhe
Vorstand: Rüdiger Dillmann, Michael Flor, Jivka Ovtcharova, Rudi Studer
Vorsitzender des Kuratoriums: Ministerialdirigent Günther Leßnerkraus

Received on Monday, 17 December 2007 15:11:23 UTC