[
Lists Home |
Date Index |
Thread Index
]
Dare Obasanjo wrote:
>For this to work (a) every description of a person must use the same data model & (b) there needs to exist a mapping from your applications data model to that of the unknown schema available somewhere. This seems fairly optimistic to me and highly unlikely in the geenral case in practice.
>
>
a) This is *explicity* untrue -- read the OWL requirements doc.
http://www.w3.org/TR/2004/REC-webont-req-20040210/
see 3.3 Ontology interoperability
"Different ontologies may model the same concept in different ways..."
as well as requirements:
R2,R3,R4,R11
b) mapping from "data model" to "schema" is done via URIs as well as the
OWL operators:
owl:sameAs
owl:equivalentClass
owl:equivalentProperty
The sum of this is that different models/ontologies of a "Person" can
interoperate. This is one of the huge benefits of *not* having a fixed
schema with represented as a fixed set of relational columns in which to
represent a person. Different ontologies might define different
properties of a person (each ontology defines the properties a
particular application might be interested in). *Somewhere else* an
inferencing engine can declare two "people" to the the same -- that is,
for example, one person represented in a credit card transaction
(identified via a signature and CC number) as the _same as_ another
representation of the _same_ person, perhaps identified by a different
credit card transaction, or perhaps as identified by a tax return
(social security number)... use your imagination. I am not saying that
there aren't processing issues *possible*, only that one does not need
to use a fixed "schema".
>
>Semantic Web proponents tend to gloss over these points whenever describing the Semantic Web utopia.
>
>
>
If you read the WebOnt use cases and requirements document, that is
explicitly not the case -- so perhaps what you say is true for "Semantic
Web proponents" who haven't been involved with the actual development of
semantic web standards.
Jonathan
|