If such a project could be funded, it might well serve as a good start
for the development of a foundation ontology that is widely accepted - one
of the two main criteria for utility as a basis for interoperability (the
other being technical adequacy).
Among your suggestions I would only have some questions about one, and a
comment about another:
(1) How 'lightweight' do you think it has to be? The size, format, and
degree of axiomatization will all be important. I am inclined to think that
the optimal format for a common FO would be to have a version in OWL that is
automatically converted into a Common-logic format such as KIF. For this to
work, the OWL version would need to have representations of rules,
functions, and higher-arity relations even though these have no effect on
the DL reasoning that is used with OWL.
Provision for rules will be important in order to allow alternative
methods of representing concepts (such as representing 'mother' by a
relation and/or by a type). The translation rules will only be executed in
a FOL reasoner, but need to be present in every version so that all views
will be properly represented. This raises the issue of how to extract out
only the parts one needs for a particular application - another comment on
that is below. (01)
(2) Mapping to WordNet synsets is probably useful, to some extent. In
building the COSMO I have begun to do that, with the result that the
development rate slows significantly. In the process, it seems clear that
the top levels of WordNet do not conform to an inheritance hierarchy that
would work for logical inference, even though most of the WordNet structure
has valuable linguistic and semantic insight. I find referring to the
WordNet as helpful in building the COSMO structure, sometimes very helpful.
The parts of WordNet that could use some reorganization probably include
most of the parts that would be in a Foundation Ontology. Nevertheless, I
would want to continue mapping and then it would be possible to construct a
'modified WordNet' in which the WordNet synsets that are not mapped to the
FO elements are included by subsumption under one or more of the synsets
that are mapped.
I suspect that current applications that use the WordNet will not "break"
if the hierarchy is modified somewhat, and existing applications may not
need to interoperate with new ones. But I think it is more important to
focus on usefulness in applications yet to be developed, of which there will
be many many more than existing ones. (02)
At some point (not necessarily in the first phase) thought needs to be
given to what kind of utility can be developed to allow potential users to
extract out of the full ontology only those parts that are needed for a
particular application. There are some precedents, such as the utility used
with the NIEM to create the "Information Exchange Package Documentation"
(IEPD) that is used to transfer information. The Protégé PromptTab plug-in
also has a tree-extraction utility that provide more than minimal
functionality, pulling out some links as well; that kind of function would
probably be needed, as well as others. (03)
The issue remains of where the funding for such a project could come
from. As we know from experience, only a very limited amount of serious
time-consuming work will be available without funding, and that only from a
few people. (04)
> -----Original Message-----
> From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx [mailto:ontolog-forum-
> bounces@xxxxxxxxxxxxxxxx] On Behalf Of John F. Sowa
> Sent: Wednesday, August 20, 2008 9:13 PM
> To: [ontolog-forum]; Patrick Cassidy
> Cc: standard-upper-ontology@xxxxxxxxxxxxxxxxx
> Subject: Re: [ontolog-forum] Foundation Ontology [was Semantic Web
> If you like the term 'foundation ontology', I won't complain. But
> please note that the method of the Longman dictionary is totally
> different from anything that we have been calling an ontology. Their
> defining vocabulary, as they say in the introduction, is intended for
> people who are learning English as a second language. It is not an
> ontology like Cyc, SUMO, BFO, Dolce, etc.
> The people who use the Longman Learners' Dictionary have an enormous
> amount of background knowledge beyond any computer system today.
> They are able to use that knowledge to interpret the very incomplete
> so-called definitions. That may be useful for people, but it's not
> an ontology that could be used for deductive reasoning by computers.
> On the other hand, those very incomplete definitions are not bad for
> specifying an incomplete type hierarchy, somewhat along the lines of
> WordNet. Something at that level can be extremely valuable, as the
> many applications of WordNet have demonstrated. In fact, it could
> also be useful as a basis for a *lightweight* low-cost foundation.
> Following is my recommendation for a foundation ontology (FO).
> I'm sending this note to both ontolog-forum and the SUO mailing
> list, because something along these lines could be appropriate
> for an IEEE standard:
> 1. A lightweight, low-cost foundation, an initial version of which
> could be developed relatively quickly without a large investment
> of cash, but which could evolve into something much more complete.
> 2. The initial components of the FO would take advantage of resources
> that have proved to be successful in practical applications. But
> the principles should also have a sound logical basis to enable
> a smooth evolution and transition toward a more complete system.
> 3. A simple, but widely used resource is WordNet. Its advantage is
> wide coverage, and its lack of detailed axioms enables it to be
> adapted to multiple purposes without creating contradictions.
> However, many aspects of WordNet, such as its top-level
> would require revisions or replacement before being adopted and
> adapted into the FO type hierarchy. Many other resources could
> also be added, but with considerable revisions to avoid conflicts.
> The FO hierarchy would initially have very few axioms, of which
> the primary ones would be the subtype/supertype relations.
> 4. Other important resources are the standards for dates, times,
> geographical locations, units of measurement, monetary units,
> chemical elements and compounds, etc. The terminology and the
> mathematical relations among terms should be related to the FO
> hierarchy and made available for all applications.
> 5. Organizations for the sciences, engineering, law, medicine,
> businesses, governments, agriculture, etc., have established
> standardized terminology with standard definitions and detailed
> specifications. These terms should be related to the basic
> FO hierarchy, but a suitable naming scheme is necessary to
> distinguish homonyms used in different standards and revisions.
> 6. The development of the FO should be coordinated with existing
> bodies such as ISO, W3C, and various governmental and non-
> governmental organizations. The naming scheme should enable
> different bodies to control their own terminology while relating
> them to the basic FO type hierarchy.
> There is a lot more to be said, but I believe that something along
> these lines would be (a) relatively inexpensive to get started,
> (b) upward compatible with existing practices, (c) immediately useful
> for practical applications, and (d) compatible with both formal
> deductive systems and much more informal tools used for information
> classification and retrieval.
> If we do a good job, it could become an IEEE standard.
> John Sowa
> Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
> Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontolog-
> Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
> Shared Files: http://ontolog.cim3.net/file/
> Community Wiki: http://ontolog.cim3.net/wiki/
> To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx (08)