Re: On the definition of "ontology"

"Nicola Guarino" <guarino@ladseb.pd.cnr.it>
Message-id: <guarino.1163210663C@150.178.2.3>
Date: Wed, 4 Oct 95 17:10:23 +0100
From: "Nicola Guarino" <guarino@ladseb.pd.cnr.it>
Subject: Re: On the definition of "ontology"
To: "Paul van der Vet" <vet@cs.utwente.nl>, "Pat Hayes" <phayes@cs.uiuc.edu>
Cc: srkb@cs.umbc.edu
X-Mailer: VersaTerm Link v1.1
Sender: owner-srkb@cs.umbc.edu
Precedence: bulk

I wrote:

>> I would now propose YET ANOTHER definition of an ontology, slightly
>> different (in the form, but not in the content) from that discussed in the
>> paper "Ontologies and Knowledge Bases: Towards a Terminological
>> Clarification", accessible on the web site reported below:
>> 
>> "An ontology is a specification of the intended models of a logical
language".

Paul van der Viet writes:

>
>I'm somewhat surprised by the plural - when you develop a
>knowledge-based system you'll have one intended model in mind.
>

Pat Hayes writes:

>Fine, as long as we don't forget that the models usually will include
>'nonstandard' models which we *didnt* intend, so we shouldnt assume that
>constructing an ontology somehow fully captures an idea.

Suppose you have a logical language (signature) where the binary predicate
"on" is defined. You want to give it the meaning of a particular physical
relation between two objects, in order to be able to say, *for instance*,
that block A is on block B. In this case you may want to exclude the
possibility for block A being on itself, by means of a suitable axiom
(irreflexivity of "on"), which becomes part of your ontology. 
You can go on excluding other unwanted models (for instance, imposing
restrictions on the arguments of "on"), but the models of your ontology will
always be a huge number. The reasons of this molteplicity are:

1. you don't want to commit to a *single situation* (say, a specific
configuration of blocks), but rather allow for *many* possible intended
situations.

2. as stated by Pat, there will always be further unintended models (except
for trivial predicates, maybe)

In order to stress this molteplicity of models, we may refine the definition
as follow:

"An ontology is a *partial* specification of the intended *possible* models
of a logical language"

where "partial" accounts for point 2, and "possible" accounts for point 1.

-- Nicola


                         ---------------------

Nicola Guarino
National Research Council     phone: +39 49 8295751
LADSEB-CNR                    fax:   +39 49 8295778
Corso Stati Uniti, 4          email: guarino@ladseb.pd.cnr.it
I-35127 Padova                WWW: http://www.ladseb.pd.cnr.it/infor/infor.html
Italy