Re: SemWeb terminology page

On Tue, Nov 2, 2010 at 4:49 PM, Karen Coyle <kcoyle@kcoyle.net> wrote:
> The term "vocabularies" gets used for a lot of different things in semantic
> web discussions. However, as this thread shows, there isn't an obvious set
> of clear terms to use in its place.
>
> Depending on who you are talking to, the things that DCMI calls "value
> vocabularies" are "controlled lists of terms" or "authority lists" or "pick
> lists." Although 'value vocabulary' is a clear distinction to adherents of
> DCAM, I have not heard that phrase used by any other communities. When I
> talk to librarians, I often use the phrase "controlled list" in my
> explanation. It would be good to get the 'value vocabulary' concept
> disseminated broadly.

Yup. "Value vocabulary" seems clear, but isn't in wide use yet.

I also hear "code list" from Geo people lately (as well as moves to
encode these in SKOS btw).


> The analogy to properties is "data elements" in the traditional IT world. In
> fact, the MARC documentation refers to the fields and subfields as data
> elements. For that reason, "metadata element" and "metadata element set"
> seem to resonate with folks who are already somewhat familiar with a data
> processing model. However, I worry that people will assume that a property
> is the same as a data element.
>
> The terms "property," "value" and "statement" have no meaning for folks in
> the library world. These are new concepts, and should be introduced as
> representing a new way of creating and using metadata. I think it is
> legitimate to say that MARC does not have properties (in the semweb sense),
> and there are no statements in a MARC record as it is coded today. The
> advantage here is that librarians can move to new concepts and a new
> vocabulary about those concepts, which I think will help keep them from
> dragging the old ideas along with them into the semantic web.
>
> Therefore (after all of that), I would vote for using 'value vocabularies'
> and 'properties' ('set of properties' for something like foaf or dcterms?),
> but explain them in terms of controlled lists and data elements, emphasizing
> the differences.

Just a nitpic: most/many RDF vocabularies (DC, SKOS, FOAF at least)
describe classes of thing as well as property terms; Agent, Image,
Document etc. The fancier ones (in OWL often) also define some other
bits and pieces too, eg. instances of a class (to use as a controlled
value...), or to express rules. So 'set of properties' captures 2/3 of
what FOAF or DC or SKOS define. 'Set of property and class terms'
captures pretty much everything, unless a vocabulary is making heavy
use of OWL.

> Yep, easier said than done.

Can't argue there :)

Dan

Received on Tuesday, 2 November 2010 16:01:20 UTC