Pat, Ron, Jawit, and Ali, (01)
I think that we can compromise. (02)
PC> That and the other applications you mention are examples of
> where semantics of information in a *narrow* domain have been
> carefully specified and are agreed to by those who develop systems
> to use that information. That tactic preceded the computer era
> with forms that people fill out, with the semantics of the fields
> well-defined. (03)
We can all agree on that. The only change I would make is to
elaborate the phrase "preceded the computer era" with the phrase
"preceded and formed the foundation for the computer era". The
primary reason why IBM beat Univac is that IBM provided a seamless
migration path from the ubiquitous punch card machinery (which was
first applied to the 1890 census) to modern computers. (04)
PC> I don’t doubt that such applications will continue to be developed.
> But your credit card won't do any of the millions of other tasks
> people want to do with information beyond serving as ID and a method
> of payment. (05)
I agree. Those are the kinds of applications that keep popping up
on the WWW. But I want to emphasize that this is an area of very
rapid development, it's in an extreme state of flux, and nobody
has a clue about what should be standardized or how. (06)
PC> Are you suggesting that we freeze all attempts at machine
> interoperability at their current level? (07)
Certainly not. We must encourage research as far as possible.
But the worst conceivable thing to do is to freeze some untested
upper level into a standard. You can only standardize *after*
the research has led to a successful project. At the moment,
we don't even know the best research directions to explore. (08)
PC> But those narrow domains in which information is well-defined
> cannot share information with each other without painstaking new
> one-by-one agreements on how the information in one domain relates
> to the information in the other domain. (09)
Indeed. That is why we should have a *framework* that can support
an open-ended range of ontologies, enable relations between them to
be defined or discovered, and support all possible generalizations,
specializations, and interrelationships. (010)
PC> What the common foundation ontology provides is a means to allow
> all specialized domains to define their information semantics in
> a way that allow them to automatically interoperate with no new
> agreements on semantics painfully hammered out when information
> transfer is desired. (011)
That is where pie in the sky starts dribbling down. The present
tense of the verb 'provides' is wrong. There is no such thing
at present. The closest approximation is Cyc, but Cyc can't
even come close to meeting your requirements. There is zero
evidence that anything you have suggested can do any better. (012)
PC> Once again, please keep in mind the difference between broad
> general semantic interoperability and interoperability in narrow
> domains. We have many examples of the latter, and as yet no
> examples of the former, because we have many examples of agreement
> on semantics in narrow domains and not yet a broad agreement on
> the semantics of multiple domains. (013)
I have been emphasizing that distinction very strongly. The narrow
domains have proved to be extremely valuable, but the general goal
has *never* been approximated, even in Cyc after 25 years of very
hard, very expensive work by very good people. (014)
PC> The point of the suggested FO project is that it is perfectly
> feasible to have such broad semantic interoperability... (015)
The word 'perfectly' demonstrates that there is no evidence.
When you have a proof, you don't need adjectives and adverbs
to make it convincing. (016)
PC> I am suggesting that:
>
> (1) we can enable interoperability among machines on a broader
> level than what has been done thus far, with a common FO;
>
> (2) there will be very large economic benefits from doing that;
>
> (3) there is therefore a powerful case for adequate funding for
> a project to accomplish broad semantic interoperability. (017)
Please replace the word 'suggesting' with 'hoping' because there
is no positive evidence for those claims and lots of negative
evidence (e.g. Cyc) that such goals are much harder to reach
than anyone had expected. (018)
Proposed compromise: (019)
1. We design a repository of ontologies together with a facility
for discovering and representing all generalization and
specialization relationships among ontologies. (020)
2. As part of that project we include all upper-level ontologies
that anyone has ever invented. (021)
3. We encourage you and Azamat and any others who have great visions
for the future to work on your foundations and to demonstrate
their value in relating the narrow ontologies. (022)
4. At every stage of development, developers with practical problems
can choose whichever contributions are appropriate for their
current project(s). Their results will validate some directions
and show that others are less promising. (023)
RW> Will I be able to buy more stuff in more places? Will my rates
> and transaction fees decrease in any noticeable way? Not optimistic. (024)
You have good reasons not being optimistic. So do I. But I believe
that there is value in bringing all the ontologies together in a
single repository and showing how they're related. (025)
My recommendation accommodates all the existing projects, and
it allows Pat and Azamat to continue their work. If they can
demonstrate success, that would be wonderful. But anybody who
needs to do something today can pick up one of the existing
ontologies that is adequate for their specialty. (026)
JK> Saying that the upper levels can be thrown away cannot mean that
> the distinctions that are made do not matter, so I assume that
> you mean that these upper level distinctions can't be used as
> the sole antecedents for very many inferences. (027)
In my earlier note, I said that the distinctions were more important
than any particular categories derived from them. There are many
different upper levels that can be derived by applying the FCA
methods to different combinations of distinctions. (028)
Perhaps some upper levels are better than others, but that is a
research question. It's not something that can or should be
decided a priori by freezing some upper level in a standard. (029)
AH> I would further conjecture that the intersection of agreement
> among such a diverse group of people would be so weak as to be
> of limited use. As you yourself note, the FO would bifurcate
> as extensions to the more specific needs of users. Here's the
> crux - the interoperability we seek isn't gained so much by
> connecting our ontologies to these (necessarily) weak upper
> concepts, but in establishing the links between the extensions
> people actually use. (030)
I agree. (031)
AH> I really don't see how searching for a unique foundation
> ontology helps at all, since in the end we will still need
> to generate these mappings for the extensions. (032)
That may be true. But the advantage of an infinite lattice is
that there is room for everybody. Pat and Azamat can work on
their generalizations, and other people can develop their
specializations. All of this can go on concurrently. (033)
If and when anybody makes a major breakthrough, they can announce
it to the world. If they have a better ontology or mousetrap,
people will beat a path to their door. (034)
John (035)
_________________________________________________________________
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 (036)
|