Re: Resources with more than one type

I would say that a thing could be said to have more than one type in three
circumstances, and it might be useful to distinguish them.

1) When a more general or more specialized type could be substituted - that
is, remaining within a type's lineage within an ontology.  I think this is
mainly what Tim B.L. is talking about in his post.  It would seem that one
would want to be able to express whether such a type could be so replaced,
but this doesn't necessarily require an ability to assign more than one
type.  That information is already inherrent in the ontology (if there is
one, of course).

2) When a type is the descendent of more than one ancestor type.  Is a
flying car an airplane or a car?  It might have descended from both.

3) When it is being considered as being in several different ontologies at
the same time. In this case, it is unlikely that the given type labels in
one ontology will encompass exactly the same categories as in the other.
For example, an oyster might be considered as being in an biological
classification ontology, or in an ontology constructed for the restaurant
trade.  One might try to say "A [restaurant-trade type A] oyster can be one
of the set [biology type 25, biology type 30,...]".  The distinction between
ontologies would presumably be handled by namespaces, but it seems that
limitation to a single type would require some special provision or syntax
for handling such statements.  I don't yet know enough about DAML+OIL to
know how it would be handled.

Cheers,

Tom P

[Tim Berners-Lee]
    [Ian Dickinson]
> > and an API could comfortable support a get/set semantics for *the*
> > type.  If, OTOH, rdf:type should always be thought of as multi-valued,
> then
> > the semantics for the API will have to be modelled on
> > adding/removing/listing the types plural, with an attendant increase in
> > complexity.
>
> Yes, but there you are.  Presumably you just want the answer to
> the query for triples  (rdf:type,  A, x) returning all x.
>
> The axioms for the DAML Properties will typically give more types
> when they are applied to many datas, by inferring supertypes etc.
>
> > So, my question is whether unique (transitive, unambiguous) datatypes,
> with
> > multiple rdf:type's, are the exception or the rule.
>
> I would say the rule.
>

Received on Thursday, 19 July 2001 08:17:15 UTC