Re: Glossary of terms relating to thesauri and faceted classifica tion

Starting at the end and working backwards...

On Mon, Jan 17, 2005 at 02:25:05PM -0500, Charles McCathieNevile wrote:
> Satisfied? :-) *Really* satisfied? :-) Not at all satisfied?
..
> Chaals (who doesn't really need a lot of beer. But if I get it, sharing it
> with a bunch of people would be a good thing to do with it...)

Hmm, beer aside, _somewhat_ satisfied... :)

> Another similarity is mathematics. "1" isn't a thing, it is a representation
> of a thing. We can discuss, manipulate and use the thing it represents to
> make people land at Heathrow instead of crash into the Thames, or to convince
> people that England has a chance to win the Ashes series. When we do this, we
> mostly don't need to know the exact relationship between "1" and what it
> represents, and don't even want to know - we are too busy fudging the figures
> and it doesn't change anything. But when we try to explain "6" to the people
> in Australia whose inherited number system only goes up to 3 or 4, it is much
> more obvious that there is some relationship, and useful to have a good story
> about what it is.

I do take the point and it is entirely possible that my
problems are just metaphysical...  

Onward...

> I think RDF's use of URIs is similar to the way a language provides words (or
> symbols, or movement patterns, or sound patterns) as identifiers for astract
> or concrete things. 

Okay, we agree that RDF uses URIs to identify things.
I'm wondering here how the term "term" maps onto that.

At various points, RDF and SPARQL specs seem to say that an
RDF term _is_ a URI and an RDF vocabulary (of RDF terms) _is_
a set of URIs.  In terms of the analogy to natural-language
words, this seems to be like saying that the string of letters
"r-o-s-e" _is_ the word; that the dictionary entry for that
string _defines_ that string; and that the dictionary as a
whole is a set of defined strings.

It's quite possible that I am splitting hairs here, but I am
more comfortable with the notion that there is an ineffable,
notional entity, a conceptual thing -- "the word" -- that is
denoted, represented, identified -- "spelled" -- by the string
"r-o-s-e".  One is not actually describing the string itself;
one is relating both the string and the definition to that
conceptual entity.

Indeed, this is what SKOS Core seems to be doing when it
carefully avoids implying that a "thesaurus term" _is_
a lexical label as opposed to the concept underlying the
lexical label.  (And it does this in part by wisely avoiding
the term "thesaurus term" at all!)

> Hmmm. I think that the URI references "defined" in the SKOS specification are
> RDF terms - that is, identifiers for some concepts.  Having identified the
> concepts, they can be further described in the SKOS spec. In other words,
> your preferred option.

I'm not so sure...  The draft DCMI Abstract Model [1] defines
"term" to be "The generic name for a property..., vocabulary
encoding scheme, syntax encoding scheme, or concept taken from
a controlled vocabulary (concept space)".  Then it defines
"term URI" as "The generic name for a URI reference that
identifies a term".  In other words, it makes a distinction
between a modeling entity and the identifier for that modeling
entity.

In contrast, the definitions of "RDF term" in RDF Semantics
and the SPARQL draft imply that the identifier _is_ the
modeling entity (not an identifier _for_ the modeling entity).
What you are saying, then, implies to me that an "RDF term"
(a URI) is not quite the same as a "DCMI term" (a conceptual
entity identified by a URI).

What, then, is a term in the SKOS vocabulary?  Reading on...:

> The SKOS vocabulary then, is a set of such entities...

Do you mean to say that a SKOS vocabulary is a set of URIs?

Would this mean that the DCMI use of "term" is at odds with
the RDF/SKOS use of "term"??

>                                     Since RDF uses URIs as a way to identify
> the things it relates, it is an easy shorthand in many cases to consider that
> the URIs are themselves the things.

Or is the RDF/SPARQL/SKOS way of putting things simply an
example of such a shorthand?  And that in reality, an RDF
vocabulary really _is_ a set of terms identified by URIs.
Assure me this is the case and I will provide you beer by
the case!

>                                     In many cases the distinction doesn't
> matter, but in some it does.

Or does it even really matter?  Convince me that it really,
really does not matter and I will provide two cases!  :)

Tom

[1] http://dublincore.org/documents/abstract-model/

-- 
Dr. Thomas Baker                        Thomas.Baker@izb.fraunhofer.de
Institutszentrum Schloss Birlinghoven         mobile +49-160-9664-2129
Fraunhofer-Gesellschaft                          work +49-30-8109-9027
53754 Sankt Augustin, Germany                    fax +49-2241-144-2352
Personal email: thbaker79@alumni.amherst.edu

Received on Monday, 17 January 2005 21:16:31 UTC