Re: A simplistic definition of "ontology"

Mike Uschold <>
From: Mike Uschold <>
Date: Mon, 9 Oct 95 19:25:18 BST
Message-id: <>
In-reply-to: <> (message from Pat Hayes on Fri, 6 Oct 1995 15:30:39 -0600)
Subject: Re: A simplistic definition of "ontology"
Precedence: bulk
% I would be interested if anyone thinks that this distinction is somehow
% mistaken. 

IMHO: Definitely NOT!

% Does anyone else feel this tension between the two perspectives?

I'm not sure what you mean by 'tension', but in my experience, there is
certainly much confusion about these things by anyone trying to
understand what ontologies are and what they are for.  I found your
contribution to be a useful clarification of varying perspectives on

Actually, I see 2 inter-related ISSUES in what Pat discusses, not just
two perspectives:

* the nature of the ontology, in particular how FORMAL is it?
* how will the ontology be USED?

% we have here two rather different kinds of thing
% here, and criteria of success, etc., are different in the two cases; and ...
% we ought to get our ideas clearer... 

Actually there are more than two, but YES, criteria for success certainly
depend on the intended use and we must try to identify what this
dependency is in practical terms.

Elaborating a bit, but mainly just rephrasing what Pat explained, I find
it convenient to think of various [fuzzily defined] points on the
formality scale, including:

A: informal: loose, un-structured nl text and/or diagrams 

B: semi-informal: structured text and/or well defined diagrams
   i.e. specialised technical vocabularies, or what Pat
   calls a glossary.
   e.g. the natural language version of our Enterprise Ontology

C: semi-formal (i.e. ontolingua-style ontologies with axioms here and
   there crisping up things where possible.).
   e.g. the Ontolingua/Kif version of the Enterprise Ontology.
D: [rigorously] formal: i.e. a logical theory as described by Pat
   possibly with proofs of various properties.
   e.g. TOVE (Gruninger and Fox)

Insofar as precisely worded documentation is encouraged and included in 
Ontolingua ontologies, they incorporate information of types B and C. 

There are also many USES for ontologies, important [and very different]
ones include

1. communication of a shared understanding between people 

2. as a unifying framework for translating between different modelling 
methods, paradigms, languages and software tools to achieve inter-operability;
(e.g. the ATOS project.)

3. re-use and sharing: an ontology may be directly inserted and used as
a module in a Knowledge Base (possibly after translation). e.g. KSL's
Library of Ontologies (ideally)

4. to ensure logical consistency is maintained between disparate systems
requiring to be integrated (e.g. for enterprise integration) 
   e.g. IDSE, as described by Menzel and Fillion in recent Ijcai workshop)

5. to assist the process of identifying requirements and defining a
specification for a system (e.g. as done in the Kactus project using
KADS models)

We can note some of the obvious relationships  between uses and
appropriate level of formality as follows:

Use 1 does not require formality, although certain benefits may derive
from more formality.  Where there *is* formality, there MUST be some
special attention paid to good hci to present the formal representation
in an understandable/readable way to humans who may not be technically

Uses 2 and 5 reuqire at least a highly structured text/diagram repn, and
ideally a considerable amount of crisping up via axioms etc.  E.g.  to
enable humans to build semi-automaitic translators.

Uses 3 and 4 require at least a sem-formal formal encoding

Comments and feedback welcome.

Mike Uschold,                 	AI Applications Institute,                    
INTERNET:    The University of Edinburgh,
Tel: (0)131 650 2732            80 South Bridge, Edinburgh EH1 1HN 
Fax:        650-6513		Scotland