Re: ISSUE 77 and postcoordination [and ISSUE-40!]

Antoine,

> I don't see a real opposition here: (1) concept in a KOS alone can have 
> a meaning that makes them accessible, just by their linguistic 
> information and semantic positioning in a network. (2) Meaning can also 
> be given by their being used to describe documents. Or to borrow from 
> nice semantic theories, I don't see why having extensional meaning would 
> prohibit intensional one :-p

I also personally don't see the reason why SKOS would not meet 
functional requirements for management and application of both 
scenarios. I expected it would do so.

The only thing that appears to me slightly alarming is tendency to mix 
requirements.
So for instance if there is no available solution for some requirements 
in expressing intensional meaning (pre-coordination - syntactic 
relationship) - less efforts would be made to solve this as this is 
assumed to be resolved through the possibility of extensional meaning 
(coordination of terms in resources, or post-coordination in retrieval).

This way the exploitation of SKOS is restricted in areas of legacy data 
which would benefit the most of its application

> [notice: afaik, the latter aspect can be quite important, even in 
> "traditional" KOS world (isn't it related to literary warrant?) Or I 
> hope so, because in my project we do concept mapping based on the books 
> these concepts are indexed with ;-) ]

Literary warrant data is very important and is normally fed back into 
primary KOS through controlled process of maintenance and development. 
Because of this data coming from literary warrant and the one from 
primary KOS is usually kept separately. In some IR scenarios literary 
warrant is more important in others less. Depends of the authority 
behind the agency producing data - so normally one would like to have 
good control over where subject data comes from.

> The question for 2 is more about whether we should offer a specific 
> property for linking concept to resources (skos:subject) or instead rely 
> on other vocabularies that already enable this (dc:subject).

I would say - this depends whether there is a scenario of SKOS 
applications in which there will be no descriptive metadata such as DC.

If you can envisage repositories of resources which will not be 
interrogated through descriptive metadata  (title, creator, format, 
type, date, identifier etc. ) - but will only have SKOS data describing 
content attached directly to resources - then yes you would need part of 
SKOS that would act as descriptive metadata.


aida

Received on Monday, 17 March 2008 23:21:39 UTC