re: discussion on part note

To put the discussion back on list
Chris, Debbie,  Pat, Aldo, Evan All

Again, my apologies for missing the meeting.

I shall wait for a clearer summary, but I would make one plea which is to
remember that this is to be a
SIMPLE first note on part-whole relations.  The most obvious thing to me as a
metamessage to me from the comments is that this isn't clear enough in the
introduction.  It needs to cover the things we can agree as basic to those
needingto put together basic part-whole structures.  It needs to counter the statement
that OWL cannot be used for part-whole relations.  Further notes can fill in the
details.

I claim no expertise in cars, and look forward to a corrected example from Evan.
I had deliberately avoided anatomy to try to stick to things with which people
were familiar.  I had foolishly thought that cars would be uncontroversial.

I would strongly object a note for this purpose that took on the full DOLCE (or
BFO, or other) axiomization or indeed that got into the many of the  issues that
Lambrix (and elsewhere Artale et al) discuss, although I would be pleased to see
the references, flavours of part-whole, and perhaps other 'further reading'
extended.  This is not because I don't think these issues are important, but
because a) they are important but can only be understood after people understand
the basics; b) they cover more than most people need; and c) although there are
a lot of ideas, there is less consensus.   (I omitted Lambrix' thesis and papers
from the references, apologies.)

Our recent interaction with users is that "simpler is better".  Most need a very
simple version most of the time.  Points where users have made errors in our
experience.

1)    Mixing part-whole and kiind-of
2)    Not understanding why you need both "all As is-part-of some B" and "all Bs
has_part some A" (apologies for the misprint.  I thought I had corrected that on
the web version.)  OWL, and DL syntax generally, obscures the distinction so it
hasto be made doubly clear.
3)    Not being able to get a part-whole explosion as they would expect
4)    Making transitive relations functional in an attempt to create a tree
5)    Confusing containment, and sometimes other relations, and whole-part
relations.
6)    Having no idea how to get the "fault of the part is a fault of the whole"
inthose situations when they need it.

Given the unfortunate problems of tableaux reasoners with KBs that contain both
is_part_of and has_part, I think this also needs a "health warning".  (We have
seen
20 class ontologies that stop both Racer and FaCT much to users' surprise.)

If we can get these six points across, we will give simple timely advice. If we
wait until we  settle arguments such as that  between Pat and Aldo over
endurants and perdurants our advice will be neither simple nor timely.

Let's keep it simple.  Even simple things are hard enough for new users.

Regards

Alan



> Uhuh, you are inviting me to a mice party in a giant cheese cake!
>
> At 16:50 -0600 17-02-2005, Pat Hayes wrote:
> >>Also for me Chris: do we move discussion to the list? maybe you can
> >>make a compilation and move things there ...
> >>
> >>Concerning part-whole, consider the existing axiomatization and
> >>typology (parts, proper parts, components, features, membership,
> >>temporal indexing of parthood, transitivity issues, universe
> >>restrictions, etc.) in DOLCE and its extended library
> >>(DOLCE-Lite-Plus), existing in both FOL and OWL, with a rich
> >>documentation at: http://dolce.semanticweb.org.
> >
> >Well, *consider* it, but then I would strongly recommend rejecting
> >it, on the grounds that the central distinction it bases itself on
> >(the perdurant/endurant distinction) has no useful place in a
> >working ontology, and in fact is actively harmful to most practical
> >part/whole reasoning. Which is a pity, as much of the DOLCE
> >structure seems useful and well-thought-out; but this useless and
> >harmful distinction runs through it like a fault line through a
> >landscape.
>
> I could just tell you that a honourable distinction, existing in many
> natural languages and much common sense reasoning, cannot be said to
> be useless. But I take you earnestly.
>
> I think no distinction is harmful *per se*, provided it is explicit
> enough to be evaluated for applicability.
> Premise: DOLCE is not the only way to draw distinctions: we have
> built it as an example of an axiomatic ontology with an attempt to
> base it on solid grounds, but in our Lab four-dimensionalist (or
> n-dimensionalist) ontologies are being investigated and developed as
> well.
>
> If your main criticism is against 3D ontologies, i.e. those that
> assume that no temporal parts of objects can be predicated, e.g.
> PatHayesYesterday -PART-OF-> PatHayesAsTemporalWorm), I should say
> that it is just a matter of taste rather than harmfulness (not that
> aesthetic appreciation cannot guard ourselves from harmfulness, but
> some see 4D as cognitively noxious as well!).
> There are good arguments for the distinction, and other against it.
> Most you can express in 4D can be expressed in 4D, and vice-versa:
> some cases will be easier to model in one paradigm, others the
> opposite.
>
> And distinctions go far beyond 3D vs 4D ... for example, do you
> really think the distinction between objects and events has no room
> in 4D?
> More practically: do you think it's the same part-of relation applied to:
>
>   i) PatHayesYesterday -PART-OF-> PatHayesAsTemporalWorm
>
> as to:
>
>   ii) PatHayesLiverYesterday -PART-OF-> PatHayesYesterday?
>
> or even as to:
>
>   iii) PatHayesLiverAsTemporalWorm -PART-OF-> PatHayesAsTemporalWorm?
>
> If you do not think they are the same relation (in the sense that its
> universe is partitioned by appropriate axioms on different types of
> entities), what's your criticism about? In DOLCE, you can talk of
> PatHayesLife, parts of that life, of being part of Pat Hayes at time
> t or forever, etc.
>
> If you think they are the same, then you accept that I sensibly say
> something like:
>
>   iv) PatHayesLiverYesterday -PART-OF-> PatHayesAsTemporalWorm
>
> ... uhm ... not that one cannot tell that, but this last use of
> part-of implies (in 4D) that:
>
>   v) PatHayesLiverYesterday -PART-OF-> PatHayesYesterday -PART-OF->
> PatHayesAsTemporalWorm
>
> And this composition of relationships is logically different from its
> component relationships, even if you state transitivity on all uses
> of part-of, which is not necessarily a good practice.
>
> I know this is just the beginning :)
> A
>
> >
> >Pat
> >
> >>More precise comments after I read your note carefully.
> >>Aldo
> >>
> >>At 16:28 -0500 17-02-2005, Christopher Welty wrote:
> >>>We had an OEP telecon today and had some lively discussion on the part note.
> >>>
> >>>First of all we all agreed it is a good start. Despite numerous
> >>>philosophical/ontological issues creeping into the discussion, we
> >>>reached a consensus that for the simple note we shouldn't change
> >>>it too much, and consider deeper issues for the longer note.
> >>>
> >>>We discussed for a while specific criticisms to the example, its
> >>>general usefulness and correctness (wrt reality).  I suggested a
> >>>change to a medical example, for which these criticisms had ready
> >>>answers and in particular lay in Alan's expertise - even more, we
> >>>could take the examples from actual usage.  In the end we
> >>>convinced ourselves that this example was a good place to start
> >>>because of its familiarity and general reusability.  Evan took the
> >>>action to work on a corrected version of the example that is
> >>>accurate wrt the anatomy of cars.  (Evan, be sure to include the
> >>>critical issue of unsprung weight).
> >>>
> >>>Some specific comments:
> >>>
> >>>- It woudl be very useful to mention in this note the limitations
> >>>on transivity in OWL DL (no cardinality restrictions) and perhaps
> >>>exemplify it.
> >>>- Brush up the introduction section.  Rephrase "the key thing to
> >>>represent about PW relations is that they are transitive", which
> >>>seems to strong .  Add a brief discussion to the point that there
> >>>are many "kinds" of PW relations and try to describe which one
> >>>this note deals with.
> >>>
> >>>Also, Deb and perhaps others will send suggested references to add.
> >>>
> >>>I am willing to take a pass on it to address these issues.  Is the
> >>>editor's draft the latest version?
> >>>
> >>>-Chris
> >>>
> >>>
> >>>Dr. Christopher A. Welty, Knowledge Structures Group
> >>>IBM Watson Research Center, 19 Skyline Dr., Hawthorne, NY  10532
> >>>USA
> >>>Voice: +1 914.784.7055,  IBM T/L: 863.7055, Fax: +1 914.784.7455
> >>>Email: welty@watson.ibm.com, Web:
> >>>http://www.research.ibm.com/people/w/welty/
> >>
> >>
> >>--
> >>
> >>
> >>
> >>Aldo Gangemi
> >>Research Scientist
> >>Laboratory for Applied Ontology
> >>Institute for Cognitive Sciences and Technology
> >>National Research Council (ISTC-CNR)
> >>Via Nomentana 56, 00161, Roma, Italy
> >>Tel: +390644161535
> >>Fax: +390644161513
> >>a.gangemi@istc.cnr.it
> >>
> >>*******************
> >>!!! please don't use the old gangemi@ip.rm.cnr.it
> >>address, because it is under spam attack
> >
> >
> >--
> >---------------------------------------------------------------------
> >IHMC           (850)434 8903 or (650)494 3973   home
> >40 South Alcaniz St.   (850)202 4416   office
> >Pensacola                      (850)202 4440   fax
> >FL 32502                       (850)291 0667    cell
> >phayes@ihmc.us       http://www.ihmc.us/users/phayes
>
> --
>
> Aldo Gangemi
> Research Scientist
> Laboratory for Applied Ontology
> Institute for Cognitive Sciences and Technology
> National Research Council (ISTC-CNR)
> Via Nomentana 56, 00161, Roma, Italy
> Tel: +390644161535
> Fax: +390644161513
> a.gangemi@istc.cnr.it
>
> *******************
> !!! please don't use the old gangemi@ip.rm.cnr.it
> address, because it is under spam attack

--
Alan L Rector
Professor of Medical Informatics
Department of Computer Science
University of Manchester
Manchester M13 9PL, UK
TEL: +44-161-275-6188/6149/7183
FAX: +44-161-275-6236/6204
Room: 2.88a, Kilburn Building
email: rector@cs.man.ac.uk
web: www.cs.man.ac.uk/mig
        www.opengalen.org
        www.clinical-escience.org
        www.co-ode.org





--
Alan L Rector
Professor of Medical Informatics
Department of Computer Science
University of Manchester
Manchester M13 9PL, UK
TEL: +44-161-275-6188/6149/7183
FAX: +44-161-275-6236/6204
Room: 2.88a, Kilburn Building
email: rector@cs.man.ac.uk
web: www.cs.man.ac.uk/mig
        www.opengalen.org
        www.clinical-escience.org
        www.co-ode.org

Received on Friday, 18 February 2005 13:12:24 UTC