Re: SUO: formal SUMO draft - compliance issues
Adam --
Thanks for creating the beginning of a discussion about issues of
conformance or compliance with a standard upper ontology. This
discussion starter is embedded in a document with a filename of
"FormalSUOdraft.rtf" and a title of "IEEE Standard Upper Ontology
(draft proposal)".
The operative section on conformance is copied below, updated with
wording from a note that you posted on 1 February 2001.
| Implementations of SUO are "ontologies" or "information models".
|
| A conforming implementation is an ontology or information model O
| such that all three of the following rules are satisfied:
|
| 1. For every term T occurring in both O and the SUO, the axioms for T
| in O must be exactly those axioms for T in the SUO whose component
| terms all occur in the language of O.
|
| 2. Every term T in O
| a. appears in the SUO, or
| b. has axioms in O and the axioms are well-formed formulas of
| SUO-KIF (or when translated from the language used for O into
| SUO-KIF is a well-formed formula of SUO-KIF) containing only
| the most specific appropriate terms which are axiomatized in
| the SUO or
| c. is axiomatized in using terms which have property 2b.
|
| 3. O is internally consistent and it is consistent with the SUO: a
| contradiction cannot be derived by means of first-order logic from
| the set of statements belonging either to O or to the SUO.
My comments are based on a view of the purpose of a machine-readable
ontology of the kind that I think we're talking about here. I have
often expressed the opinion that ontologies provide a kind of semantic
utility for interoperability among applications. Another way to think
about this is that every application represents an ontology, which may
be implicit in the application (code, database schema, etc.) or rendered
explicit in the form of an externalized ontology.
From this perspective, I question the phrase "implementations of SUO."
I'm less concerned with multiple implementations of an _upper_
ontology than I am with interoperability among applications, facilitated
by domain ontologies that are conformant to some specified degree with
an upper ontology.
The indicated wording leads me to think that this document is aimed at
a particular kind of conformance -- that I might call a minimally variant
subset of the SUO, or a virtual copy of (part of) the SUO.
Taken together, rules 1 and 2 above do not seem to allow for elaboration
of the concepts that are contained in the upper ontology. There is a bit
of wiggle room in Rule 2 that says that there may be terms in O that are
not in SUO. But then the door seems to slam shut when it specifies that
such additional terms must be "defined" by axioms that only contain terms
from SUO.
If we don't allow for robust elaboration, this seems to imply that the
upper level ontology is complete for all applications. Aside from the
unrealistic expectation that this would indicate, it seems to render
completely redundant the concept of "upper" with respect to ontologies.
In a previous posting I tried to start a bit of brainstorming about
issues related to conformance. I've copied this set of starter thoughts
below. In this brainstorm I tried to raise issues of:
* upper vs. domain ontologies,
* applications vs. ontologies,
* copies,
* subsets,
* multi-way conformance,
* vertical vs. horizontal conformance, and
in addition to the issue that has been raised by several members of the WG
about content conformance vs. language conformance.
| I would suggest that some sections of such a document might include:
|
| Definition of upper ontology
| Definition of domain ontology
| Relationship (conformance?) of a domain ontology to an upper ontology
| Relationship or conformance of an application to an ontology,
| for purposes of interoperability
| Requirements for certification
| of copies of standard ontologies
| of subsets of ontologies
| of conformant applications
| to an ontology
| to a subset of an ontology
| to multiple ontologies
| of domain ontologies
| to an upper ontology
| to a subset of an upper ontology
| to another domain ontology (?)
| Interoperability among ontology-expressing languages
| Interoperability at the level of ontological content
| etc.
| etc.
Do you agree that the conformance discussed in the draft document is a
subset of the kinds of conformance that will ultimately be of concern to
the users of an IEEE standard?
Thanks,
Doug
Doug McDavid
Member, IBM Academy of Technology
mcdavid@us.ibm.com
"Imagine all the people ... living life in peace."
Adam Pease <apease@ks.teknowledge.com>@majordomo.ieee.org on 01/31/2002
02:33:56 PM
Please respond to Adam Pease <apease@ks.teknowledge.com>
Sent by: owner-standard-upper-ontology@majordomo.ieee.org
To: standard-upper-ontology@ieee.org
cc:
Subject: SUO: formal SUMO draft
Folks,
A while back we had some discussion about conformance clauses. Frank
Farance and others also had some good input on sections that would need to
be included in a formal draft SUO. Ian and I have put together a document
that attempts to meet this need and incorporates as much of the input we've
been given as possible. I've posted the draft at
<http://ontology.teknowledge.com:8080/rsigma/FormalSUOdraft.rtf>. Note
that to keep the document to a reasonable size for downloading, I've
referenced, but not included the latest SUMO draft that would be part of a
final version of the document.
Constructive feedback would be welcome.
Adam
Adam Pease
Teknowledge
(650) 424-0500 x571