Print

Print


On Mon, 22 Sep 2003, Rachel Heery wrote:

> On Mon, 22 Sep 2003, Andy Powell wrote:
>
> >
> > I'm not sure what 'people' you are referring to here?  As far as I know,
> > the term 'related metadata' is *only* used in my draft document?  Within
> > that document, it is used as defined (I think).
>
> Is the question rather as to whether it is worth addressing within the
> model linking to 'external metadata' about the same resource? It seems to
> me this sort of linking might be quite useful for 'modularising'
> descriptive metadata.

Yes, I agree that this is an interesting question.  I dunno what the
answer is! :-)  I'm torn between saying

EITHER

- this kind of modularisation should be handled 'outside' the DCMI
metadata model by the kind of Warwick framework-like packaging
functionality ofered by the OAI-PMH, METS, etc.

OR

- the abstract model includes the notion of linking to metadata about a
related resource, therefore (for consistency if nothing else) it should
include a mechanism for linking to additional metadata about the resource.

I think I lean slightly towards the latter of these two views.  However,
in modelling terms there appear to be two places to make such a linkage:

- as 'related metadata' associated with the dc:identifier element (or one
  of its element refinements).  But, as Ann pointed out, this will require
  a slight redefinition of 'related metadata', to something like

    A metadata record that describes the value of an element or element
    refinement

- as dcterms:hasDescription (a new element refinement for dc:relation) -
  note that such a property may be proposed by the collection description
  working group anyway in due course

Again, I tend to lean towards the latter (though with not much
conviction).

To come back to the unanswered question asked in a previous thread...
'what is a value?'.  It now seems to me that a value is

  the physical or conceptual entity that is associated with a property
  when it is used to describe a resource

So the 'value' of dc:creator when it is used to describe this email is me
- the physical entity.

A 'value' of dc:type when it is used to describe this email is the
conceptual entity that represents textual resources.

In both cases, the 'value entity' may have a 'value string' and a 'value
URI'. (Actually, I don't think I have a URI yet, but I'm sure I could pick
one up somewhere! ;-) ).

However, digging too deeply into this uncovers some horrible holes in the
definitions of the 15 elements.  For example, the value of dc:identifier
is not 'the resource' (a physical or conceptual entity) but 'a reference
to the resource' (a conceptual entity).  (I.e. in RDF, the value of
dc:identifier should never be a resource).  The same is true of
dc:relation ('A reference to a related resource') and dc:source.  This
seems to run counter to other definitions in DCMES - e.g. dc:creator,
which is defined to be 'An entity...' (rather than 'A reference to an
entitiy...').  This is probably a little unfortunate.

Similarly, the definition of dc:rights is at odds with the other
definitions because it effectively defines the value to be either a
'rights statement' or a 'link to a rights statement'.  None of the other
definitions allow for the explicit possibility of linking to the value.

In the abstract model, it would be nice to skirt over some of these issues
and assume that the value of dc:rights is a 'rights statement' (a
conceptual entitiy) and the values of dc:relation, dc:source and
dc:identifier are 'resources' (physical or conceptual entities).  Both
cases, the 'references to the resources' and the 'link to a rights
statement', should be handled by the model (using 'value URI' and 'related
metadata' respectively), not hard-coded into the definitions.

Anyway, enough for now...

Andy.

> It is dealt with explicitly by the METS document mark-up
> http://www.loc.gov/standards/mets/METSOverview.v2.html
>
> <quote>
> External Descriptive Metadata (mdRef): an mdRef element provides a URI
> which may be used in retrieving the external metadata. For example, the
> following metadata reference points to the finding aid for a particular
> digital library object:
> </quote>
>
> Rachel
>
>
>
>
> > > (a) More metadata about a resource that there is no appropriate DC
> > > property for. Where metadata is encoded as XML the DC in XML
> > > guidelines give examples of this as 'mixed' metadata.
> > >
> > > (b) The same metadata as in the DC property fields, but encoded
> > > using metadata properties from a different namespace.
> >
> > I'm not aware of 'related metadata' being used to mean either of these -
> > not in the abstract model draft at least.
> >
> > > (c) A combination of (a) and (b).
> > >
> > > (d) Metadata about a related resource.
> > >
> > > (a, b, c) are in fact encodings of application profiles.
> > >
> > > I'm a bit unsure about (d). It implies I could include in a DC record
> > > for a journal article metadata about other articles that the article
> > > references, using dcterms:references. Doesn't this break the
> > > principle (1-1?) that a DC record for a resource should not include
> > > metadata for another resource?
> >
> > No, it does not do this.  Quite the opposite.  One of the purposes of the
> > model is to make it clear that 'related metadata' is *not* part of the
> > abstract 'record' and that all the properties in the record *must* be
> > attributes of the resource being described.  Hence the abstract model
> > completely endorses the 1:1 principle.
> >
> > Now, it may be the case that a particular encoding of instance metadata
> > (e.g. the RDF/XML encoding) may embed descriptions of multiple resources
> > within a single instance document - but that is a syntax issue.  But by
> > looking at the way a particular syntax implements the abstract model, it
> > is possible to see that the 1:1 principle is preserved in the abstract
> > model.
> >
> > > Or is this document an attempt ot
> > > relax that rule, reflecting that in reality people are creating records
> > > that include related metadata?
> >
> > Again, no.  The abstract model tries to clearly separate out the
> > descriptions of different resources when more than one resource is being
> > described.  That is the whole purpose of introducing the notion of
> > 'related metadata'.
> >
> > > Later in the Abstract Model, DCSV is defined as a labelled string,
> > > and that it should be treated as related metadata. This seems to
> > > be using 'related metadata' as in (a,b,c) in some cases. Period,
> > > Box and Point are giving more, or more precise, metadata about
> > > the resource itself not about a related resource.
> >
> > Not really... period, box and point give 'related metadata' about three
> > related resources - a 'period in time', a 'box in space' and a
> > 'point in space'.  These three things might be fairly abstract concepts,
> > but they are nonetheless resources.
> >
> > > The now defunct
> > > DCMICite was providing more metadata about the resource, for
> > > which there was no appropriate DC property, not metadata for a
> > > related resource.
> >
> > I agree that the DCMICite proposal does raise a slight problem with my
> > definition of 'related metadata'.  dc:identifier is something of a special
> > case.  Because the value URI for dc:identifier is the URI of the resource
> > being described, if there is 'related metadata' associated with that
> > property, then is is metadata about the resource, and not metadata about a
> > related resource.  This is also true for any element refinements of
> > dc:identifier.  I need to think a bit more about this.  I'm not sure how
> > best to handle it.
> >
> > > Though the Vcard example is metadata about a
> > > related resource.
> >
> > Correct.
> >
> > > So does the Abstrat Model document also need to include
> > > something about mixed metadata / application profile?
> >
> > I think it probably does - but not for the reasons you outline above :-).
> >
> > My current plan is to re-draft the document with an initial section that
> > covers an abstract model for DCMI metadata (though I might label this an
> > abstract model for DC[MI] application profiles).  This model will then be
> > refined into abstract models for qualified and simple DC.  I think that
> > will make the model cover what you are suggesting here.
> >
> > Andy
> > --
> > Distributed Systems, UKOLN, University of Bath, Bath, BA2 7AY, UK
> > http://www.ukoln.ac.uk/ukoln/staff/a.powell       +44 1225 383933
> > Resource Discovery Network http://www.rdn.ac.uk/
> >
>
>
> ---------------------------------------------------------------------------
> Rachel Heery
> UKOLN
> University of Bath                              tel: +44 (0)1225 826724
> Bath, BA2 7AY, UK                               fax: +44 (0)1225 826838
> http://www.ukoln.ac.uk/
>

Andy
--
Distributed Systems, UKOLN, University of Bath, Bath, BA2 7AY, UK
http://www.ukoln.ac.uk/ukoln/staff/a.powell       +44 1225 383933
Resource Discovery Network http://www.rdn.ac.uk/