uos-convene
[Top] [All Lists]

Re: [uos-convene]: Relating ontologies

To: Upper Ontology Summit convention <uos-convene@xxxxxxxxxxxxxxxx>
From: "John A. Bateman" <bateman@xxxxxxxxxxxxx>
Date: Thu, 09 Mar 2006 17:21:26 +0100
Message-id: <44105606.6090804@xxxxxxxxxxxxx>
Brief comments of endorsement ...    (01)

[Nicola]
> 1. First of all, your transcription of the DOLCE axiom is inaccurate,
>  since the "subset" relation you mention in the DOLCE example is 
> indeed the temporal inclusion relation, defined in terms of temporal
>  locations of perdurants. We have no subset relation in the DOLCE 
> vocabulary.    (02)

[Adam]
> I'm not sure how the DOLCE axiom could be inaccurate.  I just copied
>  it from a DOLCE paper.    (03)

This shows it is essential to have a validated computational
form of the ontologies to be compared. Together with Claudio
Masolo and Stefano Borgo of LOA we have a good rendition of
full first-order DOLCE in CASL: structured and typed. We do
not have a KIF import facility, but may look for a short cut
to pick out relevant parts of SUMO.    (04)

Btw: Most of the DOLCE specs given in, for example, WonderWeb
deliverables contain a variety
of small typographical mistakes, parentheses unbalanced,
etc. (sorry Nicola! :-), which are
of course removed in the full formal validated version.
Clearly the later LaTeX version is a lot better than
the previous Word version, but still ... :-) Not only
that but with copy and paste one needs *semantic* copy and
paste where appropriate axioms come along too (that is another
project): cf Michael Gruninger's point below.    (05)

[Nicola]
> 
> My conclusion is that a careful comparison between SUMO and DOLCE 
> concerning the relationship between the mereological structure of 
> processes and their temporal location could actually result in a 
> better understanding of these notions.    (06)

It is precisely in this sense that I would endorse the suggestion
of looking for common subsets in that each such entry will force
us to examine these kinds of close relationships.    (07)

So this is to agree 100% with Pat's:
[PatC]
> I think that the case presented by Adam Pease and the response of 
> Nicola Guarino ...   This case is a good example in part because it
> treats an important and common relation, which should be in any upper
> ontology, and it also elucidates some of the nuances that should be
> made clear in any upper ontology that we would recommend to the
> public for general use.    (08)

Other points:    (09)

> Whether we call them axioms or definitions doesn't matter.  If there 
> are rules or other statements that use the terms, and they aren't 
> shared by both models, they would need to be reconciled.    (010)

For formal and practical differences between these, and
for local definitions that may for theorem proving purposes
later be treated as axiomatic relative to particular modules/theories,
see the CASL documentation.    (011)

[PatC]
> If it were possible to get funding for a project to create a common 
> subset ontology, this kind of analysis will be needed, and will, I 
> think lead at the very minimum to a clear and comprehensive 
> documentation    (012)

We have some funded work in this direction already and I am writing
it in as a fixed component for our next phase of 4 years funded
research. Because we are primarily focusing on space and spatial
entities, our work has mostly focused until now on
spatial definitions; so it will be good if there are as many activities
of this kind funded as possible and again support Pat's hopes.    (013)

[Michael Gruninger]
> This is a nice way to kickoff a case study, but we have the
> opportunity to do this in a more rigorous way. The first step is to
> identify the relevant axioms for each of the terms. The second step
> is to characterize the models of the theories, because we need to
> determine whether the models of one theory are isomorphic to
> substructures of models of the other theory (following the definition
> of definable interpretation). The third step is to actually prove
> that one theory can be definably interpreted in another, or identify
> a new nontrivial theory that generalizes both.    (014)

This is exactly what I was aiming at with my criticism
yesterday of simply comparing terms and how that is
only the start of a much longer (and much more interesting)
road. This is where structured representations of the ontologies
(see my resent email from Till Mossakowski on a related issue
Subject line: "Relating ontologies: structuring and development tools") 
becomes, in our view, essential, together with
sufficiently powerful logical tools for working with them.    (015)

Best,
John B.    (016)






 _________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/uos-convene/
To Post: mailto:uos-convene@xxxxxxxxxxxxxxxx
Community Portal: http://ontolog.cim3.net/
Shared Files: http://ontolog.cim3.net/file/work/UpperOntologySummit/uos-convene/
Community Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?UpperOntologySummit    (017)
<Prev in Thread] Current Thread [Next in Thread>