Re: [ontolog] UBL proposals for codesets?

[MDaconta@aol.com]

Hi Leo,

In a message dated 10/28/2002 10:08:00 AM US Mountain Standard Time, lobrst@mitre.org writes:
> One issue with your suggestion is that, hypothetically, a concept Uruguay then really has no connection to UR or URY (these are fictitious) but only to CODESET1 and CODESET2 or perhaps you mean to CODESET1[Uruguay's index],  CODESET2[Uruguay's index]? 


I would say that the relation is to CODESET1[Uruguay's index].  In other words,
that you have to follow a chain to understand the semantics.  And you can
go as far back in that chain as you need to find commonality in your
understanding.  Of course, if there is no chain, or if the chain stops 
prematurely then we do not have machine-processable semantics.  I guess
what I am stating is that you embody as little semantics as possible at 
each point with a link back (or index into) a wider body of semantics.

This is therefore just a syntactic relation. One problem: if there is no one-to-one 
> relation between a given concept and a codeset and so also many-to-many relations between 2 codesets. But if you have an ontology node, say "ball peen hammer" and a few codesets (i.e, think of a taxonomic codeset such as the UNSPSC and NAICS), you might have to have multiple mappings. There is obviously a semantic relation between codesets as a whole, and the individual codesets should be semantically codified too. I don't disagree with that. 


I would say that the idea of the many-to-many mappings is even more a reason
to make the mapping at the codeset level and not at the individual code level.
My argument of course hinges on the ability to understand the semantics of
an individual code, given the semantics of the codeset.


> Re Uruguay as a concept: although it can be considered an instance typically, there are often strong reasons to have it as a concept  (eg., in a belief space where there are multiple Uruguay instances, etc). 


I agree that Uruguay can both be an instance of a concept (like Country) and 
a concept itself.  But, given a code for Uruguay (let's say UR) ... does not mean
you get a link to all concepts related to that code.  I would say it is sufficient to
have a link to CODESET1[UR] which tells you that UR is a code for
Uruguay and Uruguay is a Country  ... and then the semantics of that codeset
may or may not point to ontologies telling you what a country is or about
the concept of Uruguay like (it is Jack's favorite country to visit).

Now, this may or may not specifically relate to UBL's proposal for 
codesets, but I think the idea of semantic chains warrants further
study.

Talk to you soon,

 - Mike
----------------------------------------------------
Michael C. Daconta
Director, Web & Technology Services
www.mcbrad.com