Re: [ontolog-forum] Fwd: Ontologies and individuals

[William Frank <williamf.frank@xxxxxxxxx>]

The below might fit somewhat with the way John Sowa wants to have only variables and relations, but is quite different from how I have usually seen any modelling theories, including FoL  applied. I have found difficult, practical problems in dealing with the day to day knowledge that people have, and expressing it formally, using syntactic categories dictated by a modelling paradigm, no matter what we CALL the categories: variables and relations, or blings and wahoos, without carrying more or less theoretical baggage.  I seem to find  baggage in the way most people **apply** FoL and OWL, whether required by the form to do so or not, lots more still in E-R modelling, and needing to travel with a whole host of porters for the standard UML semantics.   I may well be wrong, but seem to have found that not making any such distinctions the easiest way to avoid as much of this baggage as possible, while using an upper ontology as a way of classifying these thingies, though that UpperOntology of course not being of interest to the domain practitioners, but being a finer grained, more useful theory than the others mentioned. 

I seem to find that if one starts with a theory, the examples provided somehow wind up fitting it.  Easy peasy.  But collecting everyday knowledge and then analysing it is a different kettle of fish.  I have found that software designers always just throw away what does not fit their tools, 'hard coding' the rest.   Maybe I have been just been looking in the wrong places.  For example, here is what I learned at the paint store. The guy behind the counter did not think this was hard stuff, but it's been bothering me for 30 years, since I started to apply instead of theory:

Analysis Corpus for Paint Store Domain Ontology

All paints have a color

Colors are defined by manufacturers

All paints have a manufacturer

Rose Red is a Color defined by Dutch Boy

So, If the manufacturer of the paint is not Dutch Boy, the paint color cannot be Rose Red

Paint 347 is Rose Red

All paints have a gloss

Eggshell is a gloss

if the color of a paint is rose red, the gloss of the paint may not be flat

Paint 347 is eggshell

a 1 pint square can is a container

containers have sizes, 1 pint is a size

a paint product is a paint from a manufacturer in a container

paint products are identified by SKUs

SKU 1355 is the paint product:

          paint 347 in container 1 pint square can

if the color of a paint is rose red, the gloss of the paint may not be flat

SKUs have prices

This is a dented  SKU 1355 can.

a purchase is a transaction between the paint store a a customer, involving SKU quantities delivered by the store to the customer, and amounts paid by the customer to the store, where ....

purchase volumes are quanities of paints purchased by a given customer segment
each customer is in a customer segment
John Deer is a customer, and in the commercial customer segment
commerical customers are a customer segment 
Paint 347 is not often purchased by commercial customers  

***every one*** of these things mentioned, from dented cans to their skus to purchases to colors to being a gloss, except for syntactic and logical particles like 'has' 'is' and 'if' and 'every' seem most easily represented as thingies to me, thingies we give names to,  with the least teaching required, except for people who have to unlearn about attributes and entities etc.

On Sat, Dec 22, 2012 at 5:40 PM, John F Sowa <sowa@xxxxxxxxxxx> wrote:

On 12/22/2012 11:24 AM, Chris Partridge wrote:
> I had assumed he meant the individual-type distinction. I guess you might
> mean that here.  I thought he was not willing to assume *any* philosophical
> notion was sufficiently secure for *any* practical purpose

Did you read the McCarthy-Hayes paper from 1969?  Section 2.1 has the
title "What AI can learn from philosophy".

I believe that studying philosophy is good.  But when you're trying
to teach students how to implement an ontology, you have to focus
on translating English spec's to FOL.  The philosophical jargon is
irrelevant to that task:

  1. Instead of vague discussions about individuals and types, show how
     to map the spec's to variables and relations.

  2. A type is specified by a monadic relation P(x).  P is the type,
     and x is something of type P.

  3. Two more philosophical terms that should be banished from the
     tutorial are 'universal' and 'particular'.  All relations are
     universals, and their arguments are particulars.  But those
     terms are irrelevant.  Just talk about variables and relations.

  4. Another nightmare is the term 'abstract particular'.  If you mean
     the option of quantifying over relations and letting relation
     variables appear in argument position, then say so.

> I agree that one can argue things bottoms out with intuition; but
> I guess we definitely agree it should not start with raw intuition.

Everything starts with raw intuition.  But you have to do further
analysis, induction, deduction, testing, etc.  In any case, when you're
trying to teach students how to representing a subject in FOL, there is
no reason to mention the word 'intuition'.

> The goal is not that to find a perfect ontology, but one that meets the
> constraints set. The question is what constraints to set.

Of course.  Nobody disputes that point.

> A baroque top ontology seems to get in the way. But this may be
> a feature of the relative immaturity of the area.

Ontology is over two millennia old.  For implementing ontologies,
we've had FOL for over 130 years.

People have been applying ontologies to AI and databases since
the 1960s.  Cyc has had over a thousand person-years of R & D
in applied ontology since 1984.

You can't call that immature.  I admit that there are a lot of
students who are hopelessly confused.  But much of that confusion
is created by people who kick around vague philosophical jargon.
Just teach students how to translate English spec's to FOL.

John

_________________________________________________________________
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

-- 
William Frank

413/376-8167

_________________________________________________________________
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    (01)