On Mar 26, 2009, at 10:58 AM, Phillip Lord wrote:
Pat Hayes <pha...@ihmc.us> writes:
We can consider this as
problematic even with a very simple example.
Let's assume we have two databases with information about Carbon.
meaning, I presume, the element with atomic number 14.
I was thinking of the carbon with atomic number 6.
Sorry, wrong terminology. My bad. But its the element we are talking
about, is the point.
Maybe,
but what happens if one is talking about the structure of Carbon and
it's location in the periodic table, while the other is talking
about
Carbon with the isotopic mix that we have in living organisms on
earth?
So what? They can be saying different things about the same
element. Any
isotopic mix of carbon is still carbon.
Different isotopic mixes have different properties. Atomic masses,
melting points and so on.
Right. But they can still be mixes of the same element. We are talking
about two different things: elements, and particular isotopic mixes.
They have different identity criteria and different properties; and
they are closely related, but not identical to one another. This is
really just an example of the solution to the vague-sameAs issue in an
earlier email. In order to treat different isotopes as 'identical', we
invent a (slightly) more general concept, called in this case an
'element', map the isotopes to that, and use (actual) identity on the
elements rather than the isotopes. This isnt historically accurate, of
course, but its the same logic at work.
In biology, we have the same problem. Is porcine insulin the same as
human insulin? Is "real" human insulin the same as recombinant
human insulin? Well, the answer to all of these is no
Fine, you just answered the basic ontological question.
, even though most
biologists will tell you that real and recombinant insulin are the
same
because they have the same primary sequence; a medic will tell you
otherwise, because they have different effects. Why? Don't know.
A deep question, but not a killer for ontology use.
It's not a deep question, just one to which we don't have an answer.
If you make the distinctions that you might need some of the time,
all
of the time, then you are going to end up with a very complicated
model.
Yes, you no doubt are. Tough. Its a complicated world.
Yes. And on of those complications is that we have to engineer for
usability as well as accuracy.
In my experience, they often go together. Inaccuracy makes things seem
easier to use at first, but they rapidly break, giving rise to a lot
more work than if one had done it right in the first place.
Formal ontologies are
often, perhaps always, more complicated than the informal
'knowledge' they
set out to formalize. They are obliged to make finer, more
persnickety,
distinctions between things.
Hence the evolutionary biologist says all the insulins are the same.
I don't care what the anyone says, that is wrong. They are
indistinguishable
for certain purposes, but if anyone can distinguish them at all,
they are not
the _same_.
I think that position is defensible, but unusable.
Sorry, this particular point is not open for negotiation. That
distinguishable things are not identical is a necessary, logical
truth. If you want to deny this, I have no idea what you are saying.
You aren't talking English.
All these examples can be handled by making fussy distinctions
between kinds
of thing at different granularities: carbon molecules, carbon
isotopes,
carbon the element; and then having mappings between them. I
don't know much
about insulin, but it sounds from the above that the same trick
would work.
It is tedious and hair-splitting to set this up, but once in place
its fairly
easy to use: you just choose the terminology corresponding to the
'level' you
wish to be talk ing about. sameAs works OK at each level, but you
can't be
careless in using it across levels.
If this makes you want to groan, I'm sorry. But ontology
engineering is rather
like programming.
Actually, I quite like programming. I also know how to split things
out
in the way you describe.
It requires an unusual attention to detail and a willingness to write
a lot of boring stuff, because its for computers to use, and they are
as dumb as dirt and have to have every little thing explained to them
carefully. And yup, its complicated. Until AI succeeds, it will
always
be complicated.
I'd quite enjoy it if you could patronise me a little more please.
Well, sorry, but (1) this is for public consumption so we expect a
little rhetoric, and (2) you are the one who seems to be wanting a
magic bullet.
The only solution (which is too complicated) I can
think of is to do what we do when we have this problem in
programming;
you use a pluggable notion of equality, by using some sort of
comparitor
function or object. I don't think that this is an issue for OWL
myself;
I think it's something we may need to build on top of OWL.
It belongs in your ontology for carbon and insulin, not in OWL.
Is that not what my last sentance says?
Yes, I was intending to express agreement with you here.
Pat
Phil
------------------------------------------------------------
IHMC (850)434 8903 or (650)494 3973
40 South Alcaniz St. (850)202 4416 office
Pensacola (850)202 4440 fax
FL 32502 (850)291 0667 mobile
phayesAT-SIGNihmc.us http://www.ihmc.us/users/phayes
