[SUO] Re: Key questions about common upper ontologies
I believe that the answers to Jim's questions are :
(a) Yes, it is desirable and feasible to develop
a common upper ontology within a large organization,
and it is also indispensable and unavoidable in order
to achieve the specified goal of interoperability.
(b) there is no other way to communicate conceptual information
between computers that can begin to approximate the
efficiency of an upper ontology.
------------------------------------
This answer is predicated on the assumption that
we are discussing a standard for *communication**
of conceptual information among **computers** --
to enable them to interpret the conceptual information
and take automatic action that is appropriate to
the purpose of the computer system. Humans will
be involved at some point in the activities, but
an ontology for a computer system is primarily
relevant to determining what the computers can
do without human intervention.
There is another good use for an upper ontology,
alluded to in Danny Ayers' reply, where he appears to
view it as an off-the-shelf repository of good
ideas for conceptual modeling, that one can use
and modify for one's internal purposes. Yes, that is
valuable, but it does not impose the much stricter
requirements for agreement between users that
are necessitated by Jim Schoening's specified purpose:
"a compelling need for data and semantic interoperability"
among diverse users.
For semantic interoperability of computer systems,
the upper ontology can serve as the set of general
concepts used to specify the meanings of more specific
concepts -- in effect a defining vocabulary of the
conceptual language used for computer communication.
If one wants to quibble about what, among the seamless
web of interrelated concepts, constitutes the
"upper" ontology, I would suggest that, for any
two communicating systems, any concept that is
used combinatorially to specify the meanings of
other concepts used in the communication can be
considered as part of the "upper" ontology used by
those two systems. It may not be part of the "upper"
ontology used by other systems.
It must be emphasized that the upper ontology
specifies the essential conceptual referents of
a **language** used for **communication**, so
as to enable reasoning about the referents of the
communication. Communication accurately in
any language requires having the same upper ontology.
Without the same upper ontology, inaccuracies
can occur in communication -- by computers or
people.
If one wants to develop a sophisticated system that
has no need to communicate with other systems, one
is free to use any set of defining concepts and modify
them freely. But if one wants one's computers to
communicate accurately, that goal demands that a
common conceptual standard be used. It is not the
communication standard that restricts what one can do --
it is the **need to communicate** that restricts one's
freedom of what language to use -- i.e., one needs
to use a language understood by the recipient of the
communication. Those who are not (in Danny Ayers'
words) "comfortable with the notion of a single global
Uber-ontology" can be very comfortable doing their
own thing inside their own organizations, but if one,
for example, wants to communicate with the other members
of this discussion group I would recommend strongly that
you don't try to do it in Warlpiri -- even if
you are most comfortable using that language.
And a standard upper ontology is perfectly feasible.
In all of the discussion we have endured on this list about
how different people might represent the same concept
differently, I have seen many examples of how some of us
might **prefer** one representation rather than another,
but not a single example of how one purpose **requires**
one representation and another purpose **requires** a
logically inconsistent representation. The fact that
there are different personal preferences does not
imply that a single standard would not be technically
adequate to serve all users. Within a single organization,
an ontology that is adequate to serve the purposes of
all of the constituent divisions could be developed and
adopted within that organization as a mandated standard.
And assuming that it provided a mechanism for supplementation
and revision, could continue to serve the purposes of the
organization indefinitely.
If anyone can provide examples of a **technical need**
for logically inconsistent representations for different
purposes, please do so. If there was ever a time
and place to explain why logically inconsistent
representations are needed for different practical
purposes, that time is now and the place is here, on
this list. Please, bring on your examples!
Mere personal preference is not the question. We
need not debate whether it is **possible** to represent
the same thing in logically inconsistent ways -- it
does seem so, if one never encounters a situation where
the differences have practical consequences -- but is
there actually a **need** to do so? For what purpose?
It should be a given that any upper ontology will provide
a mechanism for representation of different theories,
possible worlds, hypotheses, belief systems, alternative or
counterfactual scenarios, "what-if" situations, etc., and
this can be done by providing a mechanism for the assertion
of facts relative to specific contexts. The contexts can
be organized as a lattice of theories (some of
which may be logically inconsistent with others), but
true logical inconsistency would be avoided by recognition of
the contexts in which assertions are intended to hold.
This is what I view as a usable upper ontology fit
for industrial consumption.
======================================
So why haven't we agreed on one yet? This is an
interesting issue, but it probably should be
left to a different thread.
As for the question of "deliverables" == I think that
no one will be convinced of the utility of an upper
ontology, or of the superiority of one to another, until
several applications are available that demonstrate its
use in a non-trivial way. We have some suggested upper
ontologies on hand, but to advance the discussion we need
some way to evaluate them in an objective manner.
If we want to develop a demo application soon, the
conceptual information to be interpreted and communicated
by computers must already be encoded in the computer in a
usable form -- which means that unrestricted linguistic
information would not be appropriate for a demo. Perhaps the
best examples of information in potentially usable form
are formal databases with encoded fields, or sensor
information having a restricted set of interpretations
(i.e. not sound recordings or images).
So -- starting a different thread --
is there some formal, freely available public
database that anyone can recommend that could serve as
a set of test concepts to be interpreted and communicated
using an upper ontology (e.g. product catalog data or
corporation financial data or medical information)?
Or (less likely) some set of related sensor data that
could be interpreted for the purpose of automatically
generating a communication -- perhaps weather data?
Any suggestions?
Pat
=====================================
John F. Sowa wrote:
> After a discussion in another forum, Jim Schoening
> raised the following questions, which address the
> requirements for upper ontologies and the likelihood
> that any proposed ontology is going to be adopted
> and used successfully.
>
> > What are the possibilities of a given enterprise
> > adopting a common upper ontology?
> >
> > Or for a longer version of the question: Would it
> > be desireable and feasible for a large and diverse
> > organization (with many domains), with a compelling
> > need for data and semantic interoperability, to adopt
> > or develop a common upper ontology, plus multiple
> > domain ontologies, and to eventually mandate it within
> > the enterprise for those systems requiring interoperability?
> > It would be mandated for drafting consistent requirements,
> > developing data models, etc. Assume, of course, this
> > will be very difficult, will take a long time, and cost
> > mega bucks, but is it feasible and desireable?
> >
> > Second question: If this can't be done, is there any other
> > means of achieving semantic interoperability? Could a system
> > from one domain share data or do inferencing with a system
> > from another domain without them both complying with a
> > common upper ontology, and without a one-to-one mapping?
>
> I believe that these really are key questions that get
> to the heart of what we are supposed to be doing in the
> SUO Working Group. If this is to be anything more than
> a philosophical discussion group, we have to address
> these issues and produce some deliverables that can
> support people who have such problems. But what kind
> of deliverabes should we produce?
>
> I have some thoughts about these topics, but rather than
> bias the discussion by including my own suggestions in the
> same note, I'd like to let people think about the questions
> before we start to tackle them.
>
> John Sowa
>
--
=============================================
Patrick Cassidy
MICRA, Inc. || (908) 561-3416
735 Belvidere Ave. || (908) 668-5252 (if no answer above)
Plainfield, NJ 07062-2054
internet: cassidy@micra.com
=============================================