I'd like to comment on part of an exchange between Ron Stamper and Chris
Mendel: (01)
> > I am all for formal precision once we are sure we have fully grasped
> > what we need to be formal and precise about.
>
> Well, sure, if all that means is that we have to clearly circumscribe
> the intended domain of an ontology. But a major point of introducing
> formal methods, once we are clear about the domain, is to make
> concepts out of ordinary discourse that are initially vague and
> ambiguous clear and precise.
> (02)
There is a way, I think, to achieve an adequate level of precision in
describing the intended meanings of ontology elements using a restricted
subset of ordinary language. There have been a number of versions of a
"controlled English" already developed, but what I think is still missing is
an agreement on a large enough basic vocabulary to accompany the formal
syntax so as to have a sufficiently powerful means of describing meanings in
terms whose meanings themselves have been agreed on. I believe that that
basic vocabulary needs to have at least the content of the Longman defining
vocabulary. With that vocabulary (using only some of the possible meanings
for each word), and a restricted syntax, I believe that there will be enough
expressive power to create natural English-language definitions that can be
automatically converted into the logical axioms that are the necessary
expression of meaning for automatic reasoning purposes. It is not necessary
for people to agree on restricted meanings of words for their own *local*
usage, merely for use in the common logical vocabulary that will allow more
rapid creation of the logical descriptions of the concepts they need to
represent, and thereby greatly facilitate the conversion of local
descriptions into each other via the common logical vocabulary and format.
Any local vocabulary can be mapped to the logical elements created this way. (03)
We have had a discussion on this sort of thing already, and it seems that
there are some who doubt that any restricted vocabulary (even the 3000
concepts represented by the Longman terms) will be adequate to cover all the
terms in all the fields one might want to represent in a computer program.
Nothing wrong with gut skepticism, but the notion of a controlled vocabulary
is highly susceptible to experimental proof or disproof, and if anyone
thinks that this approach could be useful, but has questions about the size
of the minimum vocabulary needed, I would suggest that we jointly conduct a
few test runs. The test could go like this (other suggestions are welcome):
one writes a description of a concept (using any term one wants to label
it), in ordinary English, with the description providing enough information
about that concept to use it in some possible application. I will then try,
using the Longman Defining Vocabulary (LDV), to show how that description
can be grounded in the Longman defining vocabulary. 'Grounding' means that
when there are words used in the original sense description (I use
'description' rather than 'definition' because it may not be in the form of
necessary and sufficient conditions, it may only include necessary
conditions) then those words need to be defined using the LDV, and if there
are any terms in that description that are not in the LDV, those words need
to be described using the LDV, etc., until all words in the original have
been recursively described using the LDV as the basis. The criterion for
whether the English-language descriptions fit the intended meaning of the
author of the description of the original term would be the evaluation of
the author of the original sense description. If any intermediate
descriptions created by me for non-LDV terms used in the original
description do not appear to capture the intended meanings of those terms,
we will need to refine those descriptions until they have a form that the
original author is satisfied does capture the intended meaning. The only
terms that do not need to be described in Longman terms are words already in
the LDV itself. If this procedure can succeed in capturing intended
meanings grounded in the LDV, this presents a potential for creating a
'definition interpretation' program that can automatically convert
English-language descriptions of term meanings into their precise intended
logical form.
There are methodological considerations that some may want clarified, and
we can discuss that if anyone is interested enough to participate in some
test of this kind. (04)
Pat (05)
Patrick Cassidy
MICRA, Inc.
908-561-3416
cell: 908-565-4053
cassidy@xxxxxxxxx (06)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx (07)
|