On Thu, Aug 13, 2009 at 05:37:43AM -0700, Karen Coyle wrote:
> Thanks, Pete. The example that I've had in my head is: a thing
> identified as "Karen's car" belongs to classes "vehicles" and "things
> owned by Karen". Now that I've gotten that far, I'm trying to figure out
> WHY and WHEN I would make use of classes in my metadata declarations and
> my metadata instances. Classes make some sense to me as an organizing
> principle, but in nearly all cases that I can come up with I end up
> preferring to use a different approach. So, right now I'm trying to
> express some "person" information, and I need to produce birth and death
> dates. There's the BIO approach:
>
> Class:Birth
> property:date
>
> But for reasons that I have trouble articulating, I would rather have a
> specific property for birth date, like
> <http://dbpedia.org/ontology/birthdate>"dbpedia-owl.birthdate". I think
> what I'm puzzling over is what happens to the class when you want to
> express your metadata as triples?
Hi Karen,
The lightbulb went on for me when Alistair Miles described the
declaration of domain and range classes as a "license to make
inferences".
In the example above, "dbpedia:birthdate" has a declared range
of "dbpedia-owl:Person" [1]. This means that if you come across
metadata stating:
X dbpedia:birthday "2008-03-01"
and you know the declared domain of "dbpedia:birthday" (because
you looked it up at [1]), then you have a license to infer that
X is a member of the class "dbpedia-owl:Person". In other words,
the statement above infers:
X rdf:type dbpedia-owl:Person
An inference engine will actually add that statement to your data.
The example you give above seems to be saying that "date" is a
property of a member of the class of "births". It is in fact
not inconceivable to conceptualize a birthdate that way: a birth
is an event, and the birth event can have a date and perhaps a
location and maybe even a "birther" and a "birthee" (as it
were). That is in fact the gist of the example in Appendix E of
a paper about the ABC Harmony [2], which models the narrative:
On June 14 2001 at the Wesley Hospital, an 8lb 11oz baby
girl was delivered to parents Jill and John Smith. The
obstetrician at the delivery was Jane Jekyl and the midwife
was Carl Hyde.
For expressing "person" information, a property like "birthdate"
is more appropriate because it gives license to infer that the
subject is a person.
Tom
[1] http://dbpedia.org/ontology/birthdate
[2] http://metadata.net/harmony/JODI_Final.pdf
>
> kc
>
> Pete Johnston wrote:
> >Jon said:
> >
> >
> >>I assume the same applies to the DCAM.
> >>
> >
> >Yes. DCAM's range and domain _are_ those of RDF Schema.
> >
> >So, as Jon says, if a property is the subject of multiple range
> >assertions then the value can be inferred to be a member of each of the
> >classes.
> >
> >I sometimes struggle to come up with realistic examples, but one which
> >is sometimes used is that of a property "hasMother" with a domain
> >assertion for the class of Humans and two range assertions, one for the
> >class of Humans and a second for the class of Females (where that class
> >of Females also includes non-humans).
> >
> >Pete
> >
> >
> >
>
> --
> -----------------------------------
> Karen Coyle / Digital Library Consultant
> [log in to unmask] http://www.kcoyle.net
> ph.: 510-540-7596 skype: kcoylenet
> fx.: 510-848-3913
> mo.: 510-435-8234
> ------------------------------------
--
Thomas Baker <[log in to unmask]>
|