[
Lists Home |
Date Index |
Thread Index
]
As has been pointed out:
1. The problem is not with the ontology technology. The problem
is with rules applied to membership that are then assumed
to have a meaningful relationship to real world entities,
particularly in the the co-occurrence constraints (business rules).
2. That the scope of application and the abstract domain
are not necessarily the same.
3. That the best approach to implementing these can be
to create an extensible enumeration of the relationship
types given that the domain of application may vary
significantly by locale, that is, the ontology itself
may have to be very weak to have a global scope.
Regardless of the purity of the abstractions of the
ontological instance or its metalanguage, the social
dimensions of the application domain are a powerful
selector of the technology applied, where the
rules and relationships are declared and what they
are locale by locale.
In the example provided, the problem of a single viewpoint
is clear: at the Constitutional level, at the Federal level,
at the state level, at the agency level, at the
individual level, the ontological distinctions
and the co-occurrence values are distinctly different.
len
From: Pete Kirkham [mailto:pete.kirkham@baesystems.com]
(a little late as I've been away for a long weekend)
Irene Polikoff:
> Well, this seem to go back to the question of whether "husband" should
> be treated as a class - a subclass of males or as a relationship (object
> property).
In the ontology, the class describes a set of things with common properties,
this
should not be confused with a type in an information model/schema or a class
in an
OO implementation that encapsulates the data that represents the information
that
is a viewpoint on the abstract concept that is described by the ontological
classification.
So the ontology may have husband as a class- as there is a 'meaningful' set
of
human individuals that share the properties maleness and marriage, and an
information model for on application of that ontology may only have the
capability
of representing husbands with female spouses. All information models of real
world
situations appear doomed to incompleteness.
David Megginson:
> I'd say forget about classes and make your objects big bags of properties
> (attributes and relationships). The main advantages of classes for
> object-oriented programming -- code reuse and type-safe polymorphism --
> don't even apply to simple information representation, so why pay the
costs
> of a class structure if you don't receive any benefits?
The benefit of classifying concepts gives a shorthand to help us reason
about
things- it's easier to agree what a husband is, then talk about what
information
the application needs to represent about or process when dealing with
husbands than
it is to always talk about 'individuals who are in a legally married state
who are
also male and human'.
> There exists an entity A.
> A is human.
> A is male.
To me, at the ontological level, these are classes, irrespective of whether
or how
you choose to implement the data level in static typed OO software.
Given all statements may have a temporal limit, there's no problem with one
web
page from 1997 saying 'Fred is a husband' and another from 1993 saying 'Fred
is not
a husband'; OWL specifically notes that data in web documents may change and
be
inconsistent. If not, then nothing would be able to be classified, as
nothing real
is permanent. In OWL, even if you define only the relations, I can define a
class
based solely on the presence of those relations, and declare that your
statements
indicate a classification. AFAICT this is the main interoperation mechanism
that
OWL uses- saying that from your information model {x: x is a Human & x is a
Male &
x is married to y} is equivalent to my {x: x marital status = husband }
class, if
my information model only represents marital status as an enumeration. Such
mappings apply only for the information mapped, so tomorrow an individual
may be in
a different classification.
It's always possible to construct classes from relationships in the
ontology, and
this seems to be done best by what is meaningful in the domain.
In the information model, the types will be based on what is useful to the
applications that are using the information. It may be useful to limit
husbands to
current heterosexual relationships for one application, for another (e.g. a
pension
scheme), the instances considered as members of the husband type may include
deceased partners.
The software constructs that encapsulate the data may end up being very
different
to the ontological classifications and the information model types, simply
because
the software has to be written a certain way to work. Hence, in a static OO
language, you end up with David's implementation, irrespective of the
concepts
embodied. But capturing the classifications that make up the ontology
shouldn't be
constrained by the limits of one programming paradigm that may be used to
implement
one application in that domain.
Pete
********************************************************************
This email and any attachments are confidential to the intended
recipient and may also be privileged. If you are not the intended
recipient please delete it from your system and notify the sender.
You should not copy it or use it for any purpose nor disclose or
distribute its contents to any other person.
********************************************************************
-----------------------------------------------------------------
The xml-dev list is sponsored by XML.org <http://www.xml.org>, an
initiative of OASIS <http://www.oasis-open.org>
The list archives are at http://lists.xml.org/archives/xml-dev/
To subscribe or unsubscribe from this list use the subscription
manager: <http://www.oasis-open.org/mlmanage/index.php>
|
