[Top] [All Lists]

Re: [ontolog-forum] Foundation ontology, CYC, and Mapping

To: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: Rob Freeman <lists@xxxxxxxxxxxxxxxxxxx>
Date: Fri, 19 Feb 2010 12:00:16 +1300
Message-id: <7616afbc1002181500s52cf1c8cj8d68856ace7e7da2@xxxxxxxxxxxxxx>
David,    (01)

On Fri, Feb 19, 2010 at 5:08 AM, David Eddy <deddy@xxxxxxxxxxxxx> wrote:
> I would like to see the mathematical model that "proves" this:
> pound --> [mathematical proof] --> lb    (02)

You wouldn't want anything to assert that. That would be disastrous.
There must be lots of other ways of interpreting "pound". Quite apart
from physical violence, or artillery fire. Many times the equation you
would want would be:    (03)

pound --> , or #.    (04)

Even in weights and measures, the Chinese have their own measure (a
catty?) which is often called a "pound" in English.    (05)

But the perfidy of labels is an old subject.    (06)

The relevance of mathematics to the world is something else.    (07)

As I say, I actually have some sympathy with Pat's appeal to the "real
world". Though I wouldn't want to assert the observations of maths do
not apply, including the proven theoretical inconsistency of maths.
Rather my approach is to reject the idea of theory itself, complete
theory anyway. I think this belief that everything can be derived from
a single complete theory is what underlies both our problems with
labels, and the inconsistency of mathematical theories.    (08)

I would rely on observations of the world as the ultimate arbiter of
meaning (c.f. case-based reasoning), and not insist on any (single)
theory getting in the way of interpretation.    (09)

Actually, I was thinking about your problem David. I sketched some
ideas in my reply to you yesterday. Briefly I said I thought the way
you should approach interoperability between software was to find some
way to allow the computer to interpret labels based on examples of
use.    (010)

What was important was that the computer would refer to actual
examples of use, to decide how to interpret a label.    (011)

I was considering how you could do that practically.    (012)

The basic idea is that the computer does not care how data is labeled,
so long as it gets the kind of data it expects.    (013)

My suggestion is that the way to implement interoperability would be
to store actual data, all the actual values run through a program, and
use those to map interoperability.    (014)

It seems outrageous to suggest we store all the multiple values a
program is run on rather than some "simple" descriptive label. But
storage is cheap, and description is not simple, as we've been
discussing. Computers actually vastly prefer lots of detail. It is the
level they deal with best.    (015)

Google already stores every word we write.    (016)

Why not store every data value every program is run on too?    (017)

Many would be repeated. You could have a central "repository" of data
values, with links to programs which had operated on it. If any "new"
data was presented to the system, overlaps could be found and
"meaningful" suggestions made about other applicable processes or
information.    (018)

There would be no need for an exact match. Overlapping subsets would
suffice. Particular combinations of subsets identifying particular
uses. It might even be possible to "mine" the data for obvious
relationships and extend mappings in simple ways. Say for your Zip
Code example. Given the actual data items it would likely not be
difficult to identify different instantiations (perhaps with quite
different labels) and even strip off/add Post Codes and align
different formats for use in different systems. The data itself is
likely quite regular, much more regular than the way it is labeled.
Much could be done with the actual data to make up for the various
ways the labels can be interpreted by people.    (019)

Thinking of computer programs as examples of particular theories, a
collection of code anchored in actual data values might be as close to
an "infinite lattice of theories" as you are likely to get, with
different subsets, and combinations of subsets among the data, acting
as a mapping between different "theories".    (020)

Google has become a vast text repository company, storing every word
produced in the world, and has made a lot of money performing services
based on indexing that text. The more I think about it the more I feel
that a big data repository company, storing every data item run on
every computer program registered to it, would be capable of providing
many services of relevance to computer users. It would be more
powerful than Google in a way, because the data would index software
and the ways it can be used, not just texts.    (021)

Anyway, you would have interoperability implemented in a way which
made sense to computers (data) not the current half-baked efforts to
try and make humans adapt to computers by regimenting their use of
labels.    (022)

-Rob    (023)

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    (024)

<Prev in Thread] Current Thread [Next in Thread>