I agree wholeheartedly with everything that Pat has to say
here.
BS
[PC] A Common Subset Ontology
would be very different from "another
ontology". The structure of a Common Subset Ontology would
mirror
precisely the structure of the corresponding elements in each of the
linked upper ontologies, so that any domain ontology built using the
Common Subset as the basic set of defining elements for the ontology
elements in that domain ontology would be translatable accurately
into
any of the linked upper ontologies. It would not be yet
another
ontology, it is a way of recognizing those parts of the existing
upper
ontologies whose formally described meanings are close enough for
accurate translation among them. This might also apply between 3-D
and
4-D, if the intended meanings of the corresponding elements have
consistent relations.
What would be missing in the Common Subset Ontology that would be
present in each of the linked upper ontologies are: some of the more
abstract higher-level elements of each ontology that have no direct
translation; axioms providing more detailed descriptions of the
properties of certain Types or relations, which are not present in
each
of the linked upper ontologies; some properties or relations present
on
some Types in one ontology and not in the others; and some more
specific elements present in one ontology and not the other.
Those
relations that are present in the common subset should enable
inferences that would not be contradictory to any inferences
generated
by reasoning within any of the linked upper ontologies. There
would,
of course, be additional inferences generated by reasoning within
any
of the more complete linked upper ontologies.
Although the Common Subset would be translatable into only a part of
each of the linked upper ontologies, to be useful in itself it is
likely that some parts of one or more of the linked upper ontologies
would have to be modified. From my discussions with the custodians
of
most of those ontologies, it seems that few of them consider their
existing ontologies to be perfect and immutable. If any of
the
differences are based on alternatives that are not perceived as
required for some purpose but merely chosen for minor reasons, the
changes required may be relatively simple and painless. The
main
question to be decided by each of the panelists with respect to
their
own ontologies is likely to be whether making changes in order to
increase the degree of commonality would be worthwhile. Resolving
that
question, whether positive or negative, will provide the world with
useful information not yet available. It may help potential
users
decide which of the existing upper ontologies is likely to develop a
large enough user base so as to be worth investing time on.
If only two or three of the existing upper ontologies were to form a
common subset, that might be sufficient to gain most of the benefits
that would accrue from having a large user base for some upper
ontology. A larger number of related upper ontologies would be
preferable.
A part of the IKRIS project
(
http://nrrc.mitre.org/NRRC/ikris.htm)
involves finding "bridging axioms" between Cyc and PSL
elements. This
approach does not require any changes in the existing ontologies.
It
will be interesting to see how much commonality can be specified by
that approach. The first results should become public in
April. These
results may provide an existing base from which to develop a common
subset ontology.
The elements one would hope to be present in a common subset are
those
most commonly used in domain ontologies, such as person,
organization,
time, space, artifacts, locations, agents, parts, causality,
grammatical roles in events, and the sorts of things that are
extracted
by "named entity" recognition programs. The more the
better.
---------
*** (2) >> [JB] I am not sure how to fix this. I do not think a
common
subset ontology is necessarily a
sensible approach to establishing
interrelationships. One may emerge as a result of doing
the formalization of the interrelationships but, then
again, it may not. If there is are mappings between
the component ontologies, this is not a problem.
Agreeing on common ontologies always has been a problem.
[PC] Alternatives such as mappings without creating a common subset
should also be discussed. The meeting may end without a decision,
but
even a decision to continue investigating the best approach may be
constructive.
-------
*** (3) [JB] >> Can we get the particular paragraph(s) versions
that
have been proposed side-by-side to hammer them out?
Perhaps with paragraph numbers in the subject-line
or something?
[PC] The function of the Wiki pages is to provide such a more stable
location where points that have been resolved, or alternatives that
have not been resolved, can be readily viewed, modified, or
commented
on.
The Wikis currently in use for the UOS are:
The main Upper Ontology Summit page, which describes the purpose of
the
meeting and has additional discussion of the proposed Common Subset
Ontology:
http://ontolog.cim3.net/cgi-bin/wiki.pl?UpperOntologySummit
The UOS Discussion page, where any discussion of issues related to
the
UOS can be placed. If anyone finds particularly useful comments in
one
of the emails, please feel free to extract that and add it to this
page, with any further discussion desired.
http://ontolog.cim3.net/cgi-bin/wiki.pl?UpperOntologySummit/UOSdiscussi
onPage
The Joint Communique Draft, which can be modified directly by
anyone,
but the list should be notified of any modifications made:
http://ontolog.cim3.net/cgi-bin/wiki.pl?UpperOntologySummit/JointCommun
iqueDraft
The "Extended Remarks" page, which can elaborate on or
clarify
statements made more succinctly in the Joint Communique:
http://ontolog.cim3.net/cgi-bin/wiki.pl?UpperOntologySummit/ExtendedRem
arks
To modify any of those Wiki pages, anyone can click the "create
account" button under the Statue of Liberty icon on the top right
of
the page and create an account (if you are already logged, in the
"logout" button will be visible instead of "create
account"). To edit
the page, the "edit this page" button will be found at the
bottom of
the page.
Pat
Patrick Cassidy
MITRE Corporation
260 Industrial Way
Eatontown, NJ 07724
Mail Stop: MNJE
Phone: 732-578-6340
Cell: 908-565-4053
Fax: 732-578-6012
Email: pcassidy@xxxxxxxxx
-----Original Message-----
From: uos-convene-bounces@xxxxxxxxxxxxxxxx
[
mailto:uos-convene-bounces@xxxxxxxxxxxxxxxx] On Behalf Of John
A.
Bateman
Sent: Tuesday, March 07, 2006 7:03 AM
To: Upper Ontology Summit convention
Subject: Re: [uos-convene] A Common Upper Ontology?
> "(5) As one means of interrelating the existing upper
ontologies, we
> plan to explore the possibility of creating a common subset
ontology
> that will be accurately translatable into each of the linked
upper
> ontologies.
This seems to be pretty much the text that was going around
2 weeks ago. It echoes again the idea of lets build another
ontology. This does not seem productive and echoes the
SUMO-list. There have been a number of other revisions that
came closer to a good aim. But these appear to be going lost
under the mail-volume. The habit of citing huge chunks
of previous emails with inline comments does not help this
IMO.
I am not sure how to fix this. I do not think a common
subset ontology is necessarily a
sensible approach to establishing
interrelationships. One may emerge as a result of doing
the formaliation of the interrelationships but, then
again, it may not. If there is are mappings between
the component ontologies, this is not a problem.
Agreeing on common ontologies always has been a problem.
Can we get the particular paragraph(s) versions that
have been proposed side-by-side to hammer them out?
Perhaps with paragraph numbers in the subject-line
or something?
John B.
_________________________________________________________________
Message Archives:
http://ontolog.cim3.net/forum/uos-convene/
To Post:
mailto:uos-convene@xxxxxxxxxxxxxxxx
Community Portal:
http://ontolog.cim3.net/
Shared Files:
http://ontolog.cim3.net/file/work/UpperOntologySummit/uos-convene/
Community Wiki:
http://ontolog.cim3.net/cgi-bin/wiki.pl?UpperOntologySummit
_________________________________________________________________
Message Archives:
http://ontolog.cim3.net/forum/uos-convene/
To Post:
mailto:uos-convene@xxxxxxxxxxxxxxxx
Community Portal:
http://ontolog.cim3.net/
Shared Files:
http://ontolog.cim3.net/file/work/UpperOntologySummit/uos-convene/
Community Wiki:
http://ontolog.cim3.net/cgi-bin/wiki.pl?UpperOntologySummit
|