W3C home > Mailing lists > Public > public-esw-thes@w3.org > January 2008

Re: [SKOS] The return of ISSUE-44 (was Re: TR : SKOS Reference Editor's Draft 23 December 2007)

From: Daniel Rubin <rubin@med.stanford.edu>
Date: Thu, 10 Jan 2008 10:21:18 -0800
Message-Id: <6.2.5.6.2.20080110101818.04793948@med.stanford.edu>
To: Bernard Vatant <bernard.vatant@mondeca.com>
Cc: Antoine Isaac <aisaac@few.vu.nl>,SKOS <public-esw-thes@w3.org>, SWD WG <public-swd-wg@w3.org>

At 09:48 AM 1/10/2008, Bernard Vatant wrote:

>Hi Daniel
>>>
>>>As is well stated in the introduction of the editor's draft, the 
>>>intended use of SKOS, not to support inference, but to support 
>>>indexing/classification, search and retrieval of resources.
>>
>>Classification is a type of inference, and indexing/retrieval will 
>>require inference at least along hierarchical relations for query 
>>expansion, so I don't see that we can easily keeps separate from 
>>the need to support inference (albeit simple inference)...
>Well, let me clarify. The kind of inference you mention can be 
>performed by different tools in various functional implementations 
>based on various interpretations of relations (hierarchical or not). 
>For example, for whatever reason, the query expansion might want to 
>follow the hierarchy upwards and/or downwards for a specific number 
>of steps (1, 2, n), or until it finds a reasonable number of 
>answers, or follow also the "related" links one step across, 
>whatever. IOW, different implementations will retrieve different 
>sets of answers for the same query.
>What I meant by inference is inference based on the formal 
>semantics, for which any tool implementing correctly the semantics 
>has to get the same results, e.g. a SPARQL request on a given RDF graph.
>
>Let me take an example. Suppose we have.
>
>a:X     skos:narrower   a:Z1
>a:Z1     skos:narrower   a:Z2
>a:Z1    skos:related   a:Z4
>a:Z2     skos:narrower   a:Z3
>
>:R1     skos:subject      a:Z1
>:R2     skos:subject      a:Z2
>:R3     skos:subject      a:Z3
>:R4     skos:subject      a:Z4
>
>And a query expansion downwards from a:X
>
>SELECT    ?r
>WHERE    {?r  skos:subject  ?c.
>                   a:X  skos:narrower  ?c.}
>
>Following SKOS semantics this will retrieve { :R1 } since 
>skos:narrower is not transitive. And it's OK.

 From my point of view, it does NOT make sense that skos:narrower and 
broader are not transitive.
And if applications can go ahead and make them transitive by 
expanding how they wish, that violates the asserted SKOS semantics. 
Unless I'm misunderstanding something here, this sounds like a 
formula for chaos.


>But nothing prevents an application to expand the query by whatever 
>specific mean it sees fit in the context
>- Expand recursively along skos:narrower,  retrieves {:R1, :R2, :R3}
>- Expand 2 steps along skos:narrower, retrieves {:R1, :R2}
>- Expand 2 steps along either skos:narrower or skos:related, 
>retrieves {:R1, :R2, :R4}
>
>etc ...
>
>In a nutshell, SKOS relations will support functional semantics of 
>the search engine, but does not enforce any specific semantics. And 
>I think it's OK.
>
>Bernard
>
>
>
>
>
>
><http://mondeca.wordpress.com/>
>
>
Received on Thursday, 10 January 2008 18:21:30 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:38:59 GMT