Regarding John Sowa's recent posting, I would like to focus on one
problematic statement; (01)
[JS] The SUO work over the past five years has been interesting,
and we all learned a lot. But the most important thing we
learned is that assuming a fixed and frozen set of upper-level
axioms does not promote interoperability. Instead, the axioms
introduce irrelevant contradictions that are a major barrier
to communication and sharing. The solution is to minimize the
axioms at the top levels and to introduce them as needed at
the problem-oriented lower levels. (02)
------------- (03)
I have a very different interpretation of what we learned from the
IEEE-SUO experience: (04)
(1) What certainly does not promote interoperability is having several
proposed upper ontologies with alternative axiomatizations and no
mappings between them. That is the current state of the IEEE-SUO
project. (05)
**BUT** to conclude that having *0ne* upper ontology does not promote
interoperability is a completely unwarranted conclusion, since neither
IEEE-SUO nor the broader ontology community has never had one. I know
of no data to support such a conclusion. To draw that conclusion, one
would have to be able to point to some failed attempt to achieve
interoperability using some specific common upper ontology. I am
unaware of any such case, and if anyone does know of such a negative
experiment, please let us know the details, so we can learn from it.
On the other hand, every company (including Cyc) that asserts that they
will integrate or harmonize databases does so with a common reference
ontology or terminology - specific for that company. The axioms of an
upper ontology do not introduce contradictions within that ontology (if
it is sound), but may cause contradictions between different upper
ontologies - and may not. (06)
(2) As for "fixed and frozen", I really wish our discussants would not
use such terms since every ontologist I know assumes that an upper
ontology will continue to evolve as new knowledge is discovered and new
domains are integrated with it. If there is anyone who actually wants
a "fixed and frozen" upper ontology please let us know who you are, so
we can actually determine what virtues someone may see in such an
artifact. (07)
(3) On the other hand, I do agree that having a relatively simple and
not too extensively axiomatized common ontology would be a useful
**first step** in promoting interoperability. It will ultimately not
be sufficient, because there will be circumstances where the sparse
axiomatization leaves ambiguities that need to be resolved, so that
those who use the upper ontology (and the programs they use) know
precisely what is intended, among several alternatives. As a start,
potential ambiguities and the intended interpretation can be described
in detail in the documentation for each class or relation, even if the
axioms that formalize those descriptions are not included in the
ontology itself. Also, as John suggests, unambiguous classes or
instances can be assigned to particular categories so that their
meaning is less ambiguous, at least to the humans who know what those
instances are. More detailed axiomatizations may be included in more
richly axiomatized ontologies that are consistent with the simpler
ontology. It may be a good idea to have a relatively simple upper
ontology as a "starter" upper ontology that may suffice for many cases;
the representations in such an ontology should be reusable in any more
highly axiomatized ontology that is logically consistent with the
simpler ontology; there may be several such compatible but more richly
axiomatized ontologies. Representations in the simpler ontology could
in effect be abstracted "views" of one or more richly axiomatized
ontologies. As the community develops experience with applications
using that simple common ontology, additional agreements may well
develop on more detailed meanings and axiomatizations. I don't see how
we can ever discover the actual extent to which agreement may be
possible until we develop publicly available experience with
applications using some common ontology. To gain that experience is
one of the goals of the ONTACWG. Finding agreement on a simple common
upper ontology is an issue that has been raised in the Ontolog forum
and I think is worth examining further. (08)
As a concrete step in advancing this discussion, I would request that
any ONTACWG member interested in such a simplified ontology review the
proposed merger of the 100 topmost classes in the OpenCyc and SUMO
ontologies which I posted on our Wiki:
http://colab.cim3.net/cgi-bin/wiki.pl?CosmoWG/TopLevel (09)
Discussion of whether that bare hierarchy appears sound, and
suggestions for changes or additions would be very welcome. If more
information is needed to provide comments, let me know. (010)
(4) Looking for a common terminology, with words having a fixed
meaning, appears from the discussions I have had to be vastly more
difficult than agreeing on the logical structures of an upper ontology,
which may be labeled by different terms. Agreeing on a common
terminology could indeed foster interoperability, but the desire of
different communities to use different terminologies is in my view
precisely the problem that we are trying to solve with a common
ontology (or lattice thereof); getting those communities to agree on a
single terminology is not a likely solution. Perhaps there can be
agreement on a terminology for some high-level taxonomy, but this does
not appear to me to be prima facie any easier than agreement on a
common ontology. (011)
Pat (012)
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 (013)
_________________________________________________________________
Message Archives: http://colab.cim3.net/forum/ontac-forum/
To Post: mailto:ontac-forum@xxxxxxxxxxxxxx
Subscribe/Unsubscribe/Config:
http://colab.cim3.net/mailman/listinfo/ontac-forum/
Shared Files: http://colab.cim3.net/file/work/SICoP/ontac/
Community Wiki:
http://colab.cim3.net/cgi-bin/wiki.pl?SICoP/OntologyTaxonomyCoordinatin
gWG (014)
_________________________________________________________________
Message Archives: http://colab.cim3.net/forum/ontac-forum/
To Post: mailto:ontac-forum@xxxxxxxxxxxxxx
Subscribe/Unsubscribe/Config:
http://colab.cim3.net/mailman/listinfo/ontac-forum/
Shared Files: http://colab.cim3.net/file/work/SICoP/ontac/
Community Wiki:
http://colab.cim3.net/cgi-bin/wiki.pl?SICoP/OntologyTaxonomyCoordinatingWG (015)
|