There is a question from Ken Ewell which I think touches on a topic of
broad interest: (01)
>[PC] It will be easier to get agreement, the simpler the core is, but
>in order to serve the proposed function of enabling interoperability
by
>providing a set of concepts with which definitions of complex concepts
>can be specified, it will have to have a certain irreducible level of
>complexity. Just how complex it can get and still gather a wide range
>of acceptance among user communities is not obvious to me. But over
>the years I have heard many - actually most - ontologists say that
they
>would use another representation if there was a good reason to do so.
>
>
[KE] Would those same ontologist choose to interact using a different
language?
The one that affirms would make for a very interesting, and most
likely,
enlightening case. (02)
[PC] Actually, yes, if there were any benefit to do so. We have ample
precedent of people taking great trouble to learn a natural language
which is not their native tongue in order to communicate with others.
Every international scientific conference I have attended was conducted
in English, and included speakers who obviously were not comfortable
with our language, but took the trouble to learn it because it was the
medium required to communicate with those they wanted to communicate
with. People take a lot of trouble to learn complex programming
languages, because there are examples of programs using those languages
that do very useful things. Conversely, if one already knows a
programming language, it takes a lot more motivation, such as examples
of things a new language can do that the old language can't, that will
induce people to learn a second language. As far as ontologists go, I
will use **any** language that has a large user base and a reasonable
number of public sample applications. I have spoken with others having
a similar attitude. (03)
The whole game is motivation. Up to this point there has been little
if any reason for people to painstakingly learn the details of existing
upper ontology systems because (1) they are complex and difficult to
learn; (2) they are not used enough for third-party developers to
create utilities to make them easier to use and to extend their use;
(3) there are few if any publicly available demonstration programs that
make it clear that the ontologies will do enough useful things to
justify the investment of time in learning them; and (4) since there
are few reasoning systems available that already use that ontology,
there is mostly no communications benefit right now in taking the time
to tune one's own system to use it. (04)
Those who have relatively simple reasoning or representation tasks to
perform may take one look at something as complex as Cyc and conclude -
perhaps justifiably for the immediate future - that the costs of
learning to use it greatly outweigh the benefits. So they make their
own, simpler knowledge classification system. What is lost is the
potential for interoperability with other systems. But at present it
is only a potential. The upper ontologies and the Common Semantic
Model, useful in themselves, are only essential when one wants one's
reasoning system to interoperate semantically with another's (or one
organization's different databases to interact with each other). So a
COSMO is **essential** only if you have a sophisticated reasoning
system and want to interact with others. Getting to the point where
there are enough local practical reasoning systems to begin to gain the
enormous benefits of the networking effect via a COSMO is a slow
process. When there are few local reasoning systems that need to
communicate, the motivation to invest heavily in communication is
absent. That is the current situation. (05)
Nevertheless, it should be quite clear to anyone who has taken any time
to examine the simpler examples of reasoning with ontologies that the
technology will inevitably be extended to provide powerful reasoning
systems with broad and very important capabilities. As with a
programming language, there is a big difference between developing a
small test program and a large and complex operating system. Years of
intensive development involving many people may be required.
Developing an impressive reasoning system will be, I suspect, more
complex than developing an operating system like Windows. But many of
the components are available. Predicting the timetable is risky
because it depends a lot on a number of factors. (06)
Will people take the time to learn and use a complex upper ontology?
Someone else's ontology? Someone else's ontology language? Yes, for
the same reasons that they take the time to learn English and Java.
When examples are available to demonstrate the benefits of using a new
language, they will undertake the effort. (07)
But developing a widely used Common Semantic Model is not quite like
anything else that has been done before, and analogies can hide
significant differences. Developing applications of ontologies is a
more complicated task than developing some simple program in a
programming language, and no community speaks ontology as a native
language. To develop the "installed user base" that will encourage
increasing numbers of people to use, test, and improve a common
ontology may have to proceed in incremental steps of increasing
complexity. Fortunately, we don't have to get universal agreement,
just a large enough base of users to form a self-sustaining community
that can share results with a common conceptual language and help it
evolve and improve. The process could be greatly accelerated if there
were a significant source of funding that could support a large
representative group of users and developers oriented to a single
ontology, to get a variety of impressive applications and utilities to
make the system easy to use. But such funding has not yet
materialized. It is possible that some influential organization like
Microsoft will decide to create their own version for their own
purposes and by monopoly power force the rest of us to use it.
Windology anyone? (08)
There is another benefit of a Common Semantic Model beyond immediate
use in applications, and that is to serve as a common paradigm of
meaning that can help accelerate development of more powerful reasoning
systems. Reasoning with contexts will be necessary to control the
explosion of inference that will occur with first-order logic on even
moderately complex knowledge bases. To meaningfully compare
alternative reasoning methods, and learn what tactics work, it will be
essential to make comparisons using the same realistically large
knowledge base, so that the knowledge variables will be controlled and
the reasoning itself form the subject of experiment. The COSMO can
perform that function, for any community wishing to perform a
comparative evaluation of reasoning methods and to reuse each other's
results. (09)
One thing ONTACWG can do to help is to become a community with a Common
Semantic Model and develop it in incremental steps. At each stage the
investment of time in learning how to use it might be commensurate with
the demonstrated or immediately realizable benefits of the model at
that stage of development. In this way, those who are not convinced
enough of the benefits of complex ontology systems may have less
complex systems available that are easier to learn and evaluate. This
is a model for development that is propelled by a combination of the
complexity of the topic and the absence of direct funding for a broad
community effort. Whether it can succeed will depend on whether our
volunteer participants will focus on the concrete details of
construction and evaluation of the Common Semantic Model in its
increasingly complex and increasingly capable stages. (010)
There have been several proposals in ONTACWG discussion for what might
serve as a the beginnings of a Common Semantic Model. It is possible
that one of the existing upper ontologies could be adopted as a whole.
Thus far there has not been general support for that strategy. Perhaps
the complexity of those systems is not yet balanced by demonstrated
publicly available and impressive applications; the needed motivating
factors may be absent. WordNet has also been suggested as a model, but
is not itself used for logical inference. (011)
All suggested approaches are welcome. Providing specific computational
resources to support an approach will probably increase the chances
that other members will take an interest. There are some commercial
programs that can be useful, and references to those can be helpful,
but if they are expensive it will probably be necessary to provide
powerful motivation by specific examples of utility in order to induce
others to part with their cash. (012)
Right now the only specific small starter version of a COSMO that has
been proposed for ONTACWG is the merged top levels from OpenCyc, SUMO,
and DOLCE (with a few elements from BFO and ISO15926), which I put on
our site at:
http://colab.cim3.net/cgi-bin/wiki.pl?CosmoWG/TopLevel2
. . . and is available in OWL form at: (013)
http://colab.cim3.net/file/work/SICoP/ontac/reference/ProtegeOntologies
/COSMOtopOWL03.owl (014)
Discussions about this and related topics are proceeding now within the
ONTAC-dev email reflector. to subscribe go to:
http://colab.cim3.net/mailman/listinfo/ontac-dev (015)
Pat (016)
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 (017)
_________________________________________________________________
Message Archives: http://colab.cim3.net/forum/ontac-dev/
To Post: mailto:ontac-dev@xxxxxxxxxxxxxx
Subscribe/Unsubscribe/Config:
http://colab.cim3.net/mailman/listinfo/ontac-dev/
Shared Files: http://colab.cim3.net/file/work/SICoP/ontac/
Community Wiki:
http://colab.cim3.net/cgi-bin/wiki.pl?SICoP/OntologyTaxonomyCoordinatin
gWG (018)
_________________________________________________________________
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 (019)
|