I find myself on both sides
of this issue. By way of yet another anecdote:
About 20 years ago, I participated in a standards development activity
whose intent was to standardize an exchange form for 'conceptual
schemata', which led to the need to construct a 'conceptual schema of
conceptual schema concepts', what we have learned from the OMG to call
a 'metamodel'. (My contribution was a NIAM diagram of NIAM concepts.)
The participants were, in fact, very learned information modelers,
several of whom were the authors of, or students of the authors of,
some of the information modeling methods of the 1980s. There began a
discussion of fundamental concepts using favorite terms, which rapidly
descended into disagreement. After a while, I wrote five definitions
on the blackboard, creating a nonsense term for each, and using it in
subsequent definitions. Everyone agreed to substitute the same
term-of-art for the first nonsense term. Two suggestions for the
second one begot the complaint that that term should have been the
third definition or the fourth definition. It thus became apparent
that the experts thought they knew what the terms meant, and most of
the disagreement was about mismatches between terms and concepts.
Within 15 minutes, we agreed on changing one of the definitions, and
adopting those five concepts as the base concept set. We then spent
another hour arguing about which term to use for what. One of the
experts observed that he was steadily becoming more comfortable with
"zarg" as the designation for one of the concepts, since all of the
other choices were contested and therefore certain to be misunderstood
by some part of the audience. The final solution was to explicitly use
a namespace indication in front of every usage of each of the
(nominally meaningful) terms we adopted.
As Doug Foxvog observed, in theory you can get people to look at the
definitions and use the terms as defined, but in practice, it is very
hard to get them to do that. You need a device to break the
instinctive associations.
One more anecdote: In some NIST control systems architecture meeting,
we had a battle about terminology and finally agreed to a particular
set of terms for a set of concepts that we all understood. 30 minutes
later, one of the contributing academic experts argued about some
proposed architectural requirement in a way that made no sense to some
of us. And someone asked him point-blank whether his use of the term
XXX matched the definition we had agreed on. His response was "No, I
use the term with the definition common in the ZZZ literature. The
definition you agreed to is wrong." The Team Manager, the late Dr.
James Albus, then said to him: "In that case, Dr. X, you will have to
leave the room, since it will be impossible for us to communicate with
you."
As John Sowa observed, human nature trumps formal methods. Maintaining
a discipline is hard.
-Ed
Hassan Aït-Kaci wrote:
Hello,
All this discussion on meaningful vs. meaningless labels in an
ontologies brought back to me the memory of a paper I wrote back in 1982
[!] while a grad student at Penn describing an experiment on what sense
a human user is able to make out of an ontology using English labels as
opposed to to dummy labels.
What I used then to represent the ontology was Ron Brachman's KL-One and
the domain was the description of a Production-Distribution-Inventory
optimization system. Subjects (all unfamiliar with the domain) were
asked to answer questions regarding the ontology (half of the experiment
subjects were presented with meaningfully named nodes, and half with
dummy-labeled nodes).
The conclusion of the experiment was, to quote part of my paper's
conclusion, that "when entities bear English names, natural language
productions are more compact, more "natural", and syntactically more
elegant." Descriptions given by the subjects seeing only dummy-labels
were verbose, and tended to carry no clear meaning, even with a clearly
detailed ontological structure.
Here's the paper:
http://www.hassan-ait-kaci.net/pdf/Describing-Knowledge-from-Semantic-Networks.pdf
Cheers,
-hak
--
Edward J. Barkmeyer Email: edbark@xxxxxxxx
National Institute of Standards & Technology
Systems Integration Division, Engineering Laboratory
100 Bureau Drive, Stop 8263 Tel: +1 301-975-3528
Gaithersburg, MD 20899-8263 Cel: +1 240-672-5800
"The opinions expressed above do not reflect consensus of NIST,
and have not been reviewed by any Government authority."
|
_________________________________________________________________
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 (01)
|