Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: SUO: Re: W3C approves RDF and OWL as recommendations




John,
   This issue has been discussed at some length before.  Guarino/Welty 
argue for the same distinction even more strongly than I with their notion 
of 'identity'.  When a class hierarchy is non-monotonic it does make 
reasoning problematic.  Unlike Guarino/Welty I don't think this issue can 
be made completely strict, but it's certainly advantageous to maintain 
identity criteria that are as strict as possible.

Adam

At 12:08 PM 2/16/2004 -0500, John F. Sowa wrote:

>Adam,
>
>When things are related, it is important to show
>the relationships independently of the particular
>syntax as a relation or as a role.
>
>AP>  Could you elaborate on why separating roles
> > and relations is a bad thing?  To be sure,
> > if two notions can be combined, then that does
> > indeed improve sharing and reduce complexity,
> > but only if that is not at the expense of
> > creating an incorrect formalization.
>
>You certainly don't want to say something that
>is incorrect.  But in English you can talk about
>the set of parents who attend the PTA meetings
>without creating any confusion with the notion
>of parent as a relation.  What problem are you
>trying to solve or avoid?
>
>AP> Parent is a transient role and thus has not
> > been formalized as a class in SUMO.  If 'parent'
> > were made a class 'Parent', then an individual's
> > class membership could change over time, which
> > is a problem.
>
>Why?  People don't get confused about the idea
>when stated in English.  Why should they get
>confused when it is stated in logic?  Why should
>you change the syntax depending on how long the
>relationship lasts?
>
>AP> This is often the case for relations.
> > It would make no sense for example, for the
> > 'patient' role also to be a class, for the
> > same reason.
>
>Why not?  A medical information system should
>certainly be able to represent the sets of
>patients, nurses, or doctors at any particular
>point in time.  Those sets are constantly
>changing.  So what?
>
>AP> There are pragmatic boundary cases to this
> > issue, but the general approach to representing
> > transient characteristics in relations is fairly
> > standard and sound.
>
>By what standards?  There's no contradiction in
>the way people talk about these matters in English.
>It would be just as sound and logical in logic.
>
>John