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

Jakob

> No, postcoordination is also found when creating a vocabulary and when 
> indexing resources.

Could you elaborate what you actually mean by it?

As Leonard and Stella pointed out pre- post- coordination is an 
attribute we give to vocabulary with respect to how it is processed at 
the point of retrieval.

Coordination of terms in the process of indexing (or vocabulary 
building) is called pre-coordination - and it implies syntactic 
relationships. Example: "history of philosophy"

Coordination of terms in the process of retrieval is called 
post-coordination and implies logical relationships (AND, OR, NOT) 
Example: "history" [AND/OR/NOT] "philosophy"

The idea of POST-coordination in vocabulary creation does not make 
sense. That would mean that one would have to anticipate and publish all 
possible combinations of concepts.

One can make a distinction between the fact that some pre-coordination 
may come from the vocabulary itself (built in) - the other 
pre-coordination comes in the process of applications. Classifications 
are excellent examples of this: Bliss and UDC will provide some 
combinations ready made other combinations of concepts will be made as 
needed. How many pre-coordinations will be offered in Bliss or UDC is an 
editorial decision and is completely arbitrary. Because systems are 
synthetic - users may add as much as they want or may use less than offered.

Practice so far shows that for formats used as 'carriers' of KOS for the 
purpose of information retrieval - it is completely irrelevant whether 
pre-coordiantion comes from the original vocabulary or whether 
pre-coordination is created in the process of indexing.


>> If SKOS is for publishing/exchanging vocabularies, it does not need 
>  > to do anything about postcoordination.
> 
> No, as soon as SKOS is used for more then indexing one resource with one 
> concept, we need a way to express postcoordination.

how about the following view

1. We don't use SKOS to index - we use vocabulary to do so.
The fact that vocabulary happen to be formatted in SKOS is irrelevant 
for the process of indexing and information retrieval.

2. We use resource metadata (e.g. Dublin Core) to establish link between 
resource and vocabulary. We do not use SKOS to represent this link ***

3. We use information retrieval system to make sense of links between 1. 
resource 2. metadata describing it 3. vocabulary used to populate metadata.

One resource is often indexed by more than one term from a single scheme 
but that is not likely to be a problem if we have layers above properly 
managed.


*** I think this may be the point of my not being able to see the 
advantage of using SKOS for anything else than as a machine processable 
carrier of a KOS. From what Antoine said earlier and Jakob implies now - 
I think there may be different interpretations here.

Aida

Received on Sunday, 16 March 2008 19:19:29 UTC