RE: Aside - Re: Question: DAML cardinality restrictions

> > From: Jan Grant [mailto:Jan.Grant@bristol.ac.uk]
>[DAML examples elided]
> > I can't help feeling a little worried by these examples. Yes, I know
> > they're "only" examples. But the model of the world they present is
> > broken - will DAML be good for describing the real world or just
> > mathematical arenas and EDI? Is it even wise to try the former - or
> > should the examples be rewritten to be less contentious?
>
>Probably, but I think there are problems when entirely removing contention
>from examples such as these.
>
>[Sanity warning: these ideas are no doubt unfinished, and are contributed so
>that people with much more experience than me can bat them around some more
>or tell me where the answers are]

The issues you and Jon Grant raise are basic to the very idea of 
ontology description and knowledge representation. They deserve a 
book-length answer (and there are many book-length documents which 
discuss them, if you want to get into this stuff in detail) but I 
will try to sketch some replies.

First, the use of an ontology language (such as DAML+OIL) does not 
imply any endorsement of any particular ontological decisions which 
could be expressed in that language. DAML enables one to assert that 
people have feet, but it also enables someone to make a different 
assertion, or to refuse to commit themselves about the relationship 
between feet and people. Peter is right that it is probably 
impossible to say anything that someone wouldnt think is mistaken 
(even in mathematics, by the way); DAML does not presume that its use 
will provide a universal recipe for ensuring agreement. What it does 
try to provide is a way in which one set of ontological assumptions 
can be made precise, so that the question of whether or not you agree 
with it can at least be sharpened. It can help make negotiations more 
focussed, but it cannot magically produce consensus.

It would be misleading to only use mathematical examples because the 
intended use of DAML+OIL is not primarily mathematical in nature; it 
is intended to be used to represent claims about the world. The best 
that any such language can do is to represent the meaning of those 
claims as sharply and accurately as possible. That the claims 
themselves are contentious should come as no surprise.

>One problem as a language or formalism designer is that it's very difficult
>to use the thing until you've written it, and examples tend to be dreamed up
>on the spur of the moment and then stick.

True, true.

> With SMK and GRAIL, for example,
>we were fond of using "compound composite fracture of the left humerus" (or
>"compound fracture of the middle eyebrow" to illustrate what could happen
>without constraints).  These examples do tend to take on a life of their
>own, and are frequently in entirely arbitrary domains because it's easier to
>talk about than naming the concepts Cn, the roles Rn, and the datatypes Tn.
>So one source of contention is simply that there are more useful things to
>spend time on (from the designer's point of view :-).  This can be reduced
>by the designers expending more effort on the examples, but the third and
>fourth categories of contention (below) may then creep in.
>
>A second problem, especially for concrete datatypes, is the unit problem.
>What do you measure shoe size in?  Inches?  Abstract units (see below)?  Is
>that speed in miles per hour, kilometres per hour, metres per second, or
>furlongs per fortnight?  So a second source of contention is from
>underspecified examples; this can be reduced by being more explicit about
>your units, but I don't see anything in XML Schema that aids this [a pet
>peeve of mine].

Interestingly enough we had extended discussions of just this point. 
There are many ways of giving numerical shoe sizes, indeed. But where 
should these distinctions be encoded? It doesn't seem reasonable to 
expect that every country's shoe sizing conventions should be 
considered a distinct datatype. What does seem reasonable is that a 
fully fledged shoe sizing ontology might distinguish a particular 
size, a system of size coding, and the numerical value of a size in a 
system. But, to repeat, this sort of complexity belongs in the 
ontology, not in the datatypes. DAML+OIL would have no trouble 
describing distinctions like these if you want them described.

>A third and more subtle problem is assumptions that standards are the same
>worldwide.  Is that abstract unit representing shoe size UK/US/Japan, male
>or female?  So a third source of contention is implicit localisation; this
>can be reduced by describing mathematical arenas and EDI, by sufficient
>experience on the part of the modeller to know that differences exist, or by
>the whole world standardising on the same units :-).

The example was intended to be a simple one in an introductory 
document, not a final world-wide univerally acceptable theory of shoe 
sizing. This kind of simplicity might be intended for use in an 
ontology built for keeping records in one shoe store chain in 
Florida, for example, which only sells beach sandals and only uses 
one system of sizing.  It isnt intended to claim that a more 
universal shoe sizing ontology might not be more complex; it 
obviously would. There already are elaborate ontologies (not written 
in DAML+OIL yet) embodying general theories of measuring scales, 
scale-change conventions and so on.

>A fourth problem (well-known to many of the list members) is that defining
>an ontology may make explicit the assumptions of the team creating the
>model, and those assumptions may not be universally applicable.

That is not a problem. On the contrary, that is one reason why 
ontology languages are useful: they make assumptions clear and 
explicit, so one knows what one is agreeing or disagreeing with. The 
existence of an ontology does not in any way restrain the freedom of 
other people to write different ontologies which reflect a different 
point of view. We expect that ontologies will contradict each other 
from time to time.

> Not
>everyone wears shoes, or even has legs.  A travel ontology may include names
>of countries; what do you do when there are named regions that are not
>universally recognised as independent nation states?  How do you model the
>ownership of the West Bank?

Who is 'you' ? Some people will do one thing, others another.

> And what happens when organisations that would
>naturally have differing views aim to use the same ontology?

Aiming to use the same ontology might be one way to move towards a 
mutual understanding. I am more concerened about the situation where 
each is using their own ontology, and the ontologies are 
incompatible. Which is the normal case in the current world, of 
course.

> So a fourth
>source of contention may be experience or ideology; this can also be reduced
>by describing mathematical arenas and EDI.

The potential sources of such disagreements are too numerous to 
catalog. We do not want or expect to reduce such contention. We only 
aim to make it clearer.

Speaking now personally, I can forsee a future in which providing 
translations between different ontological naming conventions will be 
widely needed, and that people will make large amounts of $$ by 
providing such services.

>I think the least contentious models would therefore be those that described
>universally agreed areas that have an existing formal definition in another
>form --- maths and products (excepting units) seem to fit well here.
>Descriptions of the real world are always filtered through the perceptions
>and models of the describer; DAML certainly seems capable of acting as a
>notation for those descriptions, but I think much of the description will
>have to come from the people requiring the descriptions rather than those
>defining the notation.

Of course they will. That is what DAML+OIL is intended for. The 
notation makes almost no assumptions for itself. Please do not make 
the error of confusing the language of DAML+OIL with the content of 
any particular DAML+OIL ontology (including the ones that we use to 
illustrate the language.)

Pat Hayes

---------------------------------------------------------------------
IHMC					(850)434 8903   home
40 South Alcaniz St.			(850)202 4416   office
Pensacola,  FL 32501			(850)202 4440   fax
phayes@ai.uwf.edu 
http://www.coginst.uwf.edu/~phayes

Received on Friday, 30 March 2001 13:46:38 UTC