RE: Cyclic Classes/Properties [was: Re: DAML Correction: Same Is Not A Sub Of Sub]

On Fri, 11 May 2001, Peter Crowther wrote:

> > From: Sigfrid Lundberg, Lub NetLab [mailto:siglun@gungner.lub.lu.se]
> [...]
> > To put this in another way: Could the ability to say (i.e., 
> > in rdf/daml) 
> > that sub-titles "<" titles rather than just that sub-titles 
> > "<=" titles
> > make a more solid ground for inference. I mean, some statements could
> > actually be more precise... I'd reckon that we need both.
> 
> "<" and "<=" here stand for subset relationships.

They do

> To expand, you are stating that the set of all sub-titles should be
> a strict subset of the set of all titles, rather than that the set
> of all sub-titles should be a subset *or equal to* the set of all
> titles.  In both cases, any member of the set of sub-titles is also
> a member of the set of titles.
> 
> Introducing a strict-subset interpretation leads to some interesting issues
> that I don't know how to resolve.

Some messages on this list explained this for me, cf:

http://lists.w3.org/Archives/Public/www-rdf-logic/2001Mar/0033.html

and other messages in the same thread

> In particular, you are defining that there is a non-empty set of
> titles that are not sub-titles without stating criteria for
> inclusion in that set.  How do you name the class corresponding to
> that set?

Sure I can. At a library close to you, you'll find a library catalogue
keeping track of titles with at least the following semantics

ABBREVIATED TITLE, KEY TITLE, UNIFORM TITLE, TRANSLATION OF TITLE BY
CATALOGING AGENCY, COLLECTIVE UNIFORM TITLE, TITLE STATEMENT, VARYING
FORM OF TITLE, FORMER TITLE OR TITLE VARIATIONS

(cf http://lcweb.loc.gov/marc/bibliographic/ecbdtils.html)

Most of which has to be regarded as strict subproperties of dublin
core title. A search service cross searching a library OPAC need to
understand that a search for such a title-blob as the dc:title should
be directed to some of these fields

> Do you ensure that there may never be a singleton subclass of a
> class (generally good modelling practise)?  And if so, what do you
> do while you are developing an ontology and simply haven't fleshed
> that bit out yet, but want to test it?

Don't know. I'm not good enough at ontology modelling. I'm just an
ordinary metadata person, building search engines for my living.

Sigge

Received on Friday, 11 May 2001 08:40:02 UTC