Re: Concept co-ordination

Hi,

Concept coordination is not about order but includes syntactic indexing. 
The Universal Decimal Classification (UDC) is a good example.

Christophe Dupriez wrote:

> This article discusses pre and post coordination:
> http://archive.nyu.edu/bitstream/2451/14173/1/IS-97-14.pdf
> and answers your question about "order".
> 
> I agree with you: coordination is more than boolean operation.
> But "coordinated with..." could be an operator to link a concept A to a 
> concept B.

What coordination means depends on the indexing language - it can be 
just a boolean and, an ordered weighting or a more complex graph of 
relations: Compare this examples of the coordinated terms "export", 
"France", and "Germany":

1. export from France to Germany
2. export from Germany to France
3. export from both Germany and France
4. export from France compared to export from Germany
5 ...

A *general* coordination of a given set of concepts only shows that this 
concepts play some roles in the subject of the document they desribe, 
but *which* roles they play (in the easiest case just dct:subject) is 
given by the specific syntax of the indexing language. Order and 
delimiters are just one example of such a syntax. An example of a 
coordinated concept in UDC is:

311:[622+669](485)

which stands for "statistics 310 of mining and metallurgy [622+669] in 
Sweden (485)". You need additional subclasses of skos:Concept and 
skos:subject (and maybe other classes and properties) to express such a 
concept in RDF. However I think that for each coordinated concept you 
can say that each part of it is a skos:narrower of the coordinated 
concept. If you don't know the details of a coordinating indexing 
language, you should at least understand

a:document dct:subject [
   skos:narrower
     udc:311 , udc:485 , [
       skos:narrower udc:662 , udc: 669
     ]
] .

You end up which blank nodes, but that's life if you deal with facted 
classifications which may include more possibilities of coordination 
than which may fit into the whole semantic web.

> I'm currently thinking about how the Bliss Classification might deal 
> with this issue, since it uses co-ordination extensively, and in 
> effect provides a template which allows users to do their own concept 
> co-ordination in addition to those included in the published schedules 
> [2]. 

Bliss is a faceted classfication - each facet can be modeled as a 
sub-property of dct:subject.

> I'm copying the LCSH list into this message, since I see that 
> "double-dash" concept co-ordination seems to be a common feature 
> within LCSH, e.g.:
>
> Glass beads--Italy
> and
> Glass beads--Italy--Murano

LCSH does only handle one of the simplest cases of coordination so you 
should not draw false general conclusions from this example.

> Is order significant, in the sense that the same concepts co-ordinated 
> in a different order might have a different meaning? 

As said above this depends on the indexing language and therefore should 
not be defined in SKOS. The aim of coordination support in SKOS is to 
make explicit the semantic that is implied by a coordinated notation. 
Order is not semantic (nor are special combining charcaters like '--', 
'[..]', ':' etc.) but syntax.

Cheers,
Jakob

[1] 
http://www.w3.org/TR/2009/NOTE-skos-primer-20090818/#secconceptcoordination
[2] http://www.blissclassification.org.uk/bcclass.htm

-- 
Jakob Voß <jakob.voss@gbv.de>, skype: nichtich
Verbundzentrale des GBV (VZG) / Common Library Network
Platz der Goettinger Sieben 1, 37073 Göttingen, Germany
+49 (0)551 39-10242, http://www.gbv.de

Received on Tuesday, 10 November 2009 09:36:43 UTC