Re: [Fwd: SIG2 Amsterdam meeting - summary]

At 3:27 PM +0100 10/1/02, GUIDO VETERE wrote:
>Hi Nicola,
>thank you for getting involved in this discussion. You are right: 'part'
>can be considered (and is generally considered indeed) by a pure formal
>perspective, as a relation carrying a specific meaning axiom (weak
>supplementation) and (at least) asymmetry and antireflexivity. From this
>standpoint, 'part' does not depend upon the context, thus is eligible as
>keyword for the ontology definition language.
>Nonetheless, time-space locations cannot be part of objects, concrete
>objects cannot be part of abstract objects, sortal objects cannot be part
>of non-sortal objects, etc. That's what I had in mind when I said that the
>usage of 'part' is affected by the context.

The reason why locations are not parts of objects is due to the very 
nature of objects and locations, not to the specificity of the part 
relation. The part relation is general, various entities behave very 
differently with respect to it, that's why there are various kinds of 
entities, but just one kind of (general) part...

>Consider also that, while
>'part' supplies an identity criterion for 'extensional' classes, this is
>not true for 'intensional' classes. This is, in my opinion, a noticeable

Again, the role parts play in identity criteria of classes is 
different, but this is not an argument in favour of a multeplicity of 
part relations, it is just an argument for a multeplicity of 

>Are we sure that, when we say <'hour' part of 'day'> and
><'wheel' part of 'car'> we are using the same 'part' relation ?


>  My (naive)
>impression is that talking about 'part' is actually talking about a certain
>class of relations carrying weak supplementation, asymmetry, etc. Yet, a
>lot of relations would fall into this class (consider 'cooperate',
>'coexist', and all the verbs tagged with 'co-') that are not (at least
>intuitively) considered as 'part' relations. Here I see a difficulty...

The relations you mentioned don't fall in this class, since they are 
all symmetric (I assume...). If x coexists with y also y coexists 
with x.

>Besides this discussion, introducing 'part' as a keyword in the ontology
>definition language would either imply the adoption of a large number of
>other keywords (including 'abstract', 'extensional', 'sortal') or open to
>language-driven semantic inconsistencies. That's the reason why I  would
>avoid this choice at all.  On the other hand, introducing 'transitive',
>'symmetric', etc.., as modifiers for any binary relation would allow a
>pretty good formalization of the different kinds of part-of relations at
>the model level (e.g. a standard top-level ontology), where domain-range
>constraints can be suitably set.

I agree with the last point: there is no need to introduce a relation 
as primitive if it can be axiomatized. However, you need a suitable 
expressivity for axiomatizing it (for instance, to express the 
supplementation axiom).

>Finally, I insist on the need of introducing 'type' vs. 'role' as modifiers
>for class declarations. This distinction (that I learned from your works)
>woud be crucial for for many practical reasons. Have you any comment on
>that ?

Of course we need these things. Type vs. role is certainly a very 
important distinction for an ontology language.



Nicola Guarino
National Research Council       phone: +39 O49 8295751
LADSEB-CNR                      fax:   +39 O49 8295763
Corso Stati Uniti, 4		email:
I-35127 Padova              
(***updated 19/12/2001 ***)

Received on Wednesday, 16 January 2002 13:08:23 UTC