OASIS Mailing List ArchivesView the OASIS mailing list archive below
or browse/search using MarkMail.

 


Help: OASIS Mailing Lists Help | MarkMail Help

regrep-semantic message

[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]


Subject: RE: [regrep-semantic] [UDEF]


I very much agree with Evan's analysis.  It is very hard to express an
ontology with single tree that let along one that doesn't have typed
relationships.  It becomes even more difficult when one tries to take the
tree cross industry and international.

-----Original Message-----
From: ewallace@cme.nist.gov [mailto:ewallace@cme.nist.gov]
Sent: Friday, February 06, 2004 5:59 PM
To: carlmattocks@checkmi.com
Cc: regrep-semantic@lists.oasis-open.org
Subject: Re: [regrep-semantic] [UDEF]




"Carl Mattocks" <carlmattocks@checkmi.com> wrote:

>This is interesting. I want to now more..
>
>Zach:
>
>Please expand on the notion of 'UDEF semantic identifiers'.
>
>Evan:
>
>Please elaborate on 'lattices of these relationships '.
>

I meant networks rather than strict trees.  A simple example network
is a class with multiple inheritance.

There are also horizontal relationships like
synonyms and properties.  Think about a design model of a racecar which
describes different component systems.  All of these components have
a partOf relation to the car.  Something like a transmission often
plays at least two different roles in a hierarchy of component systems
in a racecar.  It is partOf the drivetrain and may be partOf the load
bearing structural system.   Twisting all these properties and
relationships into a strict hierarchy leads to awkward models such as
the UDEF Object tree.

I didn't mean to imply that supporting lattices was unusual for modeling
languages.  It isn't.  I was arguing that such expressiveness is necessary
for useful semantic models.

>Everyone :
>
>Please consider if the Semantic Web could leverage "concepts ... denoted
>by the paths from these nodes to the root rather than the node itself"

To a certain extent they already do.  I was trying to simplify a finer
distinction.  The path back to the root through subtype relations in an
RDFS or OWL model of course has implications on a class and instances
(individuals) of that class.  Just the implications you would expect if
you have programmed in an Object Orient programming language.  If
Racecar is a subtypeOf Car is a subtypeOf Vehicle, then any Racecar
instance is also a Car and a Vehicle instance and inherits the
characteristics of those supertypes.

By constrast, the relations in the UDEF Object tree do not have any
explicitly defined implications.  It's only when you have followed the
path that you might be able to infer what the relations might have been
along each connection in the path.  This makes the tree hard to navigate
when looking for a specific concept.  It also can lead to related or
similar concepts being located quite far apart in the tree.

-Evan



[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]