tor 2010-10-14 klockan 17:40 -0400 skrev Thomas Baker:
> dct:AgentClass rdfs:subClassOf dct:AgentClass
This triple is completely harmless, though redundant.
>
> which seems like a harmless bit of redundancy because "All
> classes are subclasses of themselves" [2] -- a case of stating
> something explicitly that can be inferred.
>
> What is not clear to me, with a brief look at RDF Semantics,
> is whether we are actually losing information by not saying
> in 2010 what the term declaration said in 2008, i.e.:
>
> dct:AgentClass rdf:type rdfs:Class
> dct:AgentClass rdfs:subClassOf rdfs:Class
Both of these are important. The first says that there are instances of
AgentClass. Based on the definition, these could be things like
ex:Student - a class of agents or ex:Managers - a class of education
managers.
The second says that all instances of AgentClass are also classes. That
is, the student Bob cannot be an AgentClass, but the class of Students
(of which Bob is one) can be an AgentClass.
What remains to express is that all members of an instance of AgentClass
are Agents (equivalently: all instances of AgentClass are subclasses of
Agent). This isn't expressible in RDF Schema. And the notion is
incompatible with OWL DL, and requires OWL Full to be expressed...
>
> All classes are subclasses of rdfs:Resource, but are they
> also all subclasses of rdfs:Class (i.e., can this second
> triple simply be inferred)?
No, certainly not. That would mean that all resources are classes, which
is simply not true.
>
> Either way, this is clearly an error of transcription somewhere
> along the way and needs to be corrected. I can correct this
> with an errata note. I'm wondering how urgent this is, as
> I am currently very busy preparing for next week's DC-2010
> conference.
As we're talking about a single redundant but true assertion, I don't
see the hurry.
/Mikael
|