[sc34wg3] From SC32 and SUO: Working on a definition of "ontology"

[Mason, James David (MXM)]

 Forwarded from other committees.

Jim Mason

-----Original Message-----
From: standard-upper-ontology at ieee.org
[mailto:standard-upper-ontology at ieee.org] On Behalf Of Frank Farance
Sent: Monday, April 10, 2006 6:40 PM
To: standard-upper-ontology at IEEE.ORG
Subject: Working on a definition of "ontology"

Hello-

We haven't seen much traffic on this list in a while.  We know this topic has
probably been debated to death.  However, we have given a lot of though to
this problem, especially when trying to apply the terminological methods and
methodology of ISO TC37 (Terminology).

We've heard the term "ontology" used in many ways and we're thinking about
how this is used in the information technology sense, i.e., the kinds of
things we talk about here when we speak of "ontology".

We've looked at a couple well-known definitions, such as:

        - specification of a conceptualisation of a knowledge domain. An
ontology is a controlled vocabulary that describes objects and the relations
between them in a formal way, and has a grammar for using the vocabulary
terms to express something meaningful within a specified domain of interest.
The vocabulary is used to make queries and assertions. Ontological
commitments are agreements to use the vocabulary in a consistent way for
knowledge sharing. ...

The "specification" part seems problematic.  A specification of X,
inherently, is about identifying a class of objects that satisfy certain
criteria.  So a "specification of a conceptualization" means that someone is
identifying one or more conceptualizations that meet certain criteria.

In the end, this doesn't address anything computable, which is something we
see as essential to "ontology" -- at least the use of "ontology" in the IT
sense.

        - Ontologies resemble faceted taxonomies but use richer semantic
relationships among terms and attributes, as well as strict rules about how
to specify terms and relationships. Because ontologies do more than just
control a vocabulary, they are thought of as knowledge representation. The
oft-quoted definition of ontology is "the specification of one's
conceptualization of a knowledge domain."

Then "ontologies" according to this definition are no different than "concept
systems", as defined by ISO TC37.  TC37's concept systems have relationships
among concepts (with no particular requirements on their simplicity or
complexity), so the description above is equivalent to TC37's "concept
system" ... and if "ontologies" are only "concept systems" then we should
call them as such, but we believe there is more.

        - Ontology is the newest label attached to some KOSs. Ontologies are
being developed as specific concept models by the Knowledge Management
community. They can represent complex relationships between objects, and
include the rules and axioms missing from semantic networks. Ontologies that
describe knowledge in a specific area are often connected with systems for
data mining and knowledge management.

We agree somewhat with this definition because there is clearly the
computational component to it.  All "knowledge organization systems" are TC37
"concept systems" (these KOS might have more than TC37 requires in its
definition).  And the "rules and axioms" imply some thought towards
computation, as evidenced by the reference to "often connected with systems
for data mining and knowledge management".

So in conclusion, we have found the following definition to be a reasonable
definition that is reasonably precise:

        [context: information technology]
        ontology: concept system and its computational model

With the above definition, taxonomies and knowledge organization systems all
could be ontologies, as long as they include their computational model.  For
example, a taxonomy of objects in the field of X is just a concept system,
but a taxonomy of objects in the field of X that includes a computational
model (e.g., producing common/differing characteristics, finding lowest nodes
with common features, etc.) could then be an ontology.  Whether the
computational model is expressed in procedural code (e.g., programs) or
non-procedural code (e.g., axioms and statements) is irrelevant.

We'd like feedback on this.  We're working on this in INCITS/L8 (the US
metadata standards committee) and its corresponding ISO committee ISO/IEC
JTC1 SC32 WG2 Metadata.

If you have some improvements to suggest, please feel free to contact us (put
"ontology" in the Subject line).  Naturally, this can all start a flame war
-- no one has the time for this.

Thank you in advance for your comments.

Frank Farance (Farance Inc.)
Dan Gillman (Bureau of Labor Statistics)

______________________________________________________________________
Frank Farance, Farance Inc.    T: +1 212 486 4700   F: +1 212 759 1605
mailto:frank at farance.com       http://farance.com
Standards/Products/Services for Information/Communication Technologies