RE: [OEP and PORT] "Classes as values": comments on usage of dc:s ubject

Hi Natasha, all,

> On the other hand, if we are trying 
> to illustrate a point, such as when and how to use classes as 
> property 
> values, it may be good to limit the content of the example to things 
> that are absolutely necessary to understanding the point and not have 
> the reader have to wander elsewhere to look for things that 
> don't make 
> the example any clearer. 

In general I am a strong advocate of making these notes as simple and
accessible as possible.  So focussing on the bare essentials is I think a
very good idea.  

> Or do they? So, in this particular 
> case, would 
> referring to the SKOS vocabulary make an example more clear 
> than using 
> the local parentSubject property? What do you think?

I think if approach 3 is going to stay in this note, then using the SKOS
vocab is a good idea, mainly because if everyone uses the same set of
properties to organise their subject hierarchies (rather than everyone
inventing their own), it makes sharing them even easier.  Also if there is a
clear spec on how they should be used, this helps to promote consistent use
and application.  

My current personal view on using dc:subject is that it should only be
applied as in approach 3.  For other approaches, other properties should be
used (e.g. foaf:topic and foaf:primaryTopic).  This could lead to a clear
recommendation on which approach should be used with which property.  Then,
when you come across a dc:subject property or a foaf:topic property, you
would know exactly what to expect.  

Why approach 3 and the dc:subject property?  Approach 3 is consistent with
the DCQ [1] spec for dc:subject, while the others are arguably not.  Also,
approach 3 represents a specific style of doing what is traditionally called
'subject-based indexing', where the 'subjects' are essentially concepts,
identified via their labels and any explanatory notes or definitions, but
are not necessarily classes or individuals.  I've met a lot of people for
whom subject-based indexing is a major part of their responsibilities, but
are confused about how to use the semantic web to do it better.  Issuing a
clear and unambiguous recommendation about the use of dc:subject, that is in
line with current specs, would be warmly welcomed I suspect.  It would also
go a long way to preventing mixed and inconsistent usage within larger
organisations such as libraries, where inconsistency is currently a major
barrier to metadata sharing. 

For many of the people I work with, because interoperability is such a major
concern, it is true to say that they are not looking for many different ways
to do something, they want just one (or maybe two at a stretch :). 

Incidentally SKOS-Core was designed to be consistent with the current DCQ
spec for usage of dc:subject.

Anyway, my opinion on this is still not at all set, I thought I'd try to get
this out there to see what you all thought.

Alistair.

[1] http://dublincore.org/documents/dcq-rdf-xml/ see section 2.3.1 and 2.3.9



> 
> I really don't have an answer...
> 
> Natasha
> 
> PS. I do agree with Bernard though that referring to specific (or at 
> least realistic book titles) is better than having individuals named 
> MyLionBook.
>

Received on Wednesday, 5 May 2004 13:21:54 UTC