Dear Matthew, Barry, and Arun, (01)
Thanks for the comments. (02)
I'll start with a comment on Arun's first point: (03)
> AM: My main concern is the buy-in from industry. Mostly,
> they will determine what gets used. There are many ISO
> standards that are rotting on standards shelves and never
> get used. We want to ensure as wide a buy-in as possible
> by industry else we will fall into the same dark, bottomless
> abyss. (04)
That is certainly true, and it implies that one of the most
important criteria for evaluating any ontology is the ease
of migrating legacy systems to the new approach. Anything
that requires a total overhaul of the old systems will not
be accepted, no matter how good, pure, and elegant it may be. (05)
> MW: I haven't been able to find something I can download,
> either it asks for a password or there is no file. (06)
On the first page, click "THE WORLD DIRECTORY OF THINGS"
and you get a PDF file of 1717 pages. By the way, there
are many other systems that are worth looking at. Arun
suggested the BeingMeta project at the MIT Media Lab: (07)
http://www.beingmeta.com/technology.html (08)
This has pointers to four open-source knowledge bases,
FramerD, BRICO, Gnowsys, and BabelVision. (09)
> MW: ... Ontology is extremely immature as a subject and
> to coalesce now around a single ontology would stifle
> progress. I would see the ideal situation as there being
> a handful of major ontologies with competition between
> them. I think this is most likely to drive the improvement
> that I believe is necessary. (010)
I mostly agree. But that raises the question of how a variety
of systems based on incompatible ontologies can interoperate.
I believe it is possible, because today we have many systems
that work together successfully even though *none* of them are
based on an explicit ontology. Ideally, an ontology should
not make it more difficult for systems to cooperate. (011)
> JS: Is interoperability fundamentally task dependent?
>
> MW: No. Certainly not if the immense effort we put into
> making our interfaces consistent is anything to go by
> and the very network nature of the interfaces. (012)
I should have stated the question more precisely. One could say
that every interface is designed for a particular type of task,
which could be very general. The http interface, for example,
was designed for accessing web pages, which is a specialized
type of task. But since the content of a web page may be
unrestricted, a large family of new kinds of subtasks have come
into existence, each of which may create specialized interfaces. (013)
> MW: It is commonplace for individual applications to have
> different ontologies (but hardly upper ontologies). For
> an organization to have and use an upper ontology as the basis
> for sewing them together is invaluable, and provides the basis
> for a common language for that business. (014)
Yes, but is the business-wide ontology really an upper ontology
in the sense of Cyc or other major AI systems? For example, they
might have categories such as Employee and Customer with Person
as a supertype, but database administrators rarely, if every,
define Person as a member of the species Homo Sapiens, a mammal,
vertebrate, etc. (015)
In fact, there are many databases for which the highest level
types in the ontology are Number and CharacterString. But
somehow, they manage to muddle along for years and interact
successfully with other databases that have a more realistic
ontology (or none at all). Things like this make me question
the need for an upper level as a prerequisite for communication. (016)
I realize that an upper level is important for reasoning, but
even a totally unrealistic one with CharacterString as the
topmost type can work for years -- it can even interoperate
with other systems that do inferences or computations with
totally different ontologies. (017)
BS> NCOR has an evaluation committee which is charged precisely
> with doing that. See the NCOR wiki at http://ontologist.org.
> Also the NCBO (http://ncbo.us) has a parallel effort for
> biomedical ontologies. (018)
Committees can serve a useful role in reviewing proposals,
but I was asking some fundamental questions about the criteria
that anyone, individual or committee, should apply. (019)
Doug Lenat is certainly a serious contributor to the field,
and he has on many occasions said that the middle levels are
far more important than the upper level. Over the years, I've
seen many different upper levels developed and used, most of
which are more or less adequate for the purpose of supporting
inferences at the middle levels. That observation lends some
credence to Lenat's point, and it raises very serious questions
about the nature of and the relationships among the levels. (020)
BS> The documentation of some of the alternatives DOES NOT
> USE CAPITAL LETTERS IN THE STYLE OF WEBSITES DEVOTED
> TO THE KENNEDY ASSASSINATION. (021)
I agree that heavy use of upper-case letters is often a sign
of a crank, but it's not always clear who's right. I admit
that there are points about Azmat's upper level that I find
troublesome, but I have never seen any upper level that does
not seem troublesome in one or more ways. If you don't like
EIS, then look at the systems from the MIT Media Lab, which
is a respectable institution that avoids upper-case letters. (022)
BS> A small one is desirable (and I think unavoidable), capturing
> for instance the distinction between monadic types and relations. (023)
I agree with Barry and Arun that the distinction between types
and monadic relations is useful, but different people have different
criteria for distinguishing them. (024)
BS> The idea is to do this [evaluation] scientifically, and
> by a neutral body with no axe to grind. (025)
No human being, except perhaps a newborn infant, is ever neutral.
Before accepting anything that has been done "scientifically",
I'd like to see a clear statement of the science that is employed. (026)
John (027)
_________________________________________________________________
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 (028)
|