Re: Re[2]: Modelling Resumes in RDF(S) - work in progress

On 2002-06-05 23:56, "ext Uldis Bojars" <uldis.bojars@gmx.net> wrote:

> 
> PS> Choosing URIs as values allows one to localize the interface
> PS> using alternate labels in other languages. Literal values would
> PS> be less internationally oriented.
> 
> PS> E.g. 
> 
> PS>    <cv:Sex rdf:about="voc:blackeye.vsaa.lv/cv/sex/male">
> PS>       <rdfs:label xml:lang="en">Male</rdfs:label>
> PS>       <rdfs:label xml:lang="fi">Mies</rdfs:label>
> PS>       ...
> PS>    </cv:Sex>
> 
> And it allows one to query RDF statements based on URI values.
> 
> OK - how do I define my own vocabulary for describing things (sex in
> this case)? Create a class Gender and subclassed Male and Female?
> Or should Male and Female be instances of class Gender?

They would be instances of the class.

> Are there benefits to make Male, Female, Gender subclasses of WordNet
> classes? 

One benefit of making Gender a subclass of a WordNet class is that
any query searching for an instance of the WordNet class would also
find instances of its subclases (presuming the search engine provides
RDFS support).

> And I it necessary to have any RDF file for (male, female,
> gender) at all - maybe it is sufficient to know that URI for male will
> be voc:blackeye.vsaa.lv/cv/sex/male, although it does not point to a
> real web document.

I must warn you that I have very strong opinions about this
issue that are not necessarily held by a majority of folks...

I would consider it an error (or at least very poor practice)
to use a URL which is expected to resolve to a representation
of a web resource to denote an abstract resource such as a term
in a controlled vocabulary.

There is a difference between retrieving knowledge about an
abstract resource, and retrieving the abstract resource itself.

It is not possible to obtain an HTTP representation of a
vocabulary term. You might obtain an HTTP representation of
an RDF schema, which contains statements about that term,
but that is not at all the same thing.

A picture of a dog is not a dog. And one should not confuse
a URI that denotes the picture with a URI that denotes the
actual dog.

I consider this an essential, core principle of the semantic
web.

There are others, though, who would recommend using an
http: URL to denote your term and to put something there
such as an RDDL or RDF instance. While that may provide
some useful functionality, "not all that feels good in
the here and now is good for you in the long run..."

> PS> For birth place values, have a look at TGN:
> 
> PS> http://www.getty.edu/research/tools/vocabulary/tgn/
> 
> Thanks, I am looking at this
> 
>>> 
>>> _2_  About linking Person with a CV
> 
> PS> Rather, you want the aboutPerson property, as in
> 
> PS>    <cv:CV rdf:about="{URI denoting CV}">
> PS>       <cv:aboutPerson rdf:resource="{URI denoting person}"/>
> PS>       ...
> PS>    </cv:CV>
> 
> I guessed so :)
> 
>>> _3_ Structure of CV information
> 
> PS> Looks fine to me, though I didn't see how you plan to
> PS> relate the actual content with the CV_Entries.
> 
> CV_entry does not have any actual content.
> It is there are an abstract class that groups all different kinds of
> CV entries together. But it does not have any real functionality.
> 
> After I looked at this model more closely, I figured some more
> interesting classes appearing in the CV model:
> 
> Working history and courses both point to a Company (where one worked
> or who organized courses). So it may make sense to make another class
> company and reference to it from work history and/or educations.

OK. So you're looking at avoiding any free textual content. There
certainly will be advantages to that insofar as query precision is
concerned.

> Also - reference points to a person. So the class reference may have
> some special properties and then point to a Person class. Only one
> question about this:
> 
> If the Reference has the following properties: Type
> (personal/professional), name, surname, title, organization, phone,
> e-mail), the only thing that Person described in vCard3.0 does not
> cover is type of the reference. How do I describe it in RDFS?

Arbitrary qualification of statements in RDF is a bit burdensome.

I would myself define a number of subproperties of 'reference' which
carry the specialized semantics.

Cheers,

Patrick

--
               
Patrick Stickler              Phone: +358 50 483 9453
Senior Research Scientist     Fax:   +358 7180 35409
Nokia Research Center         Email: patrick.stickler@nokia.com

Received on Thursday, 6 June 2002 07:35:38 UTC