Re: "zero relations" in dataset mappings (has no matching entity in)

(I think it is time you had at least one reply :-) )

On 20 Jan 2011, at 13:10, Neubert Joachim wrote:

> When matching and mapping two datasets, it is common that - on both
> sides - you find entities which don't have a matching entity on the
> other side.
> When that non-matching was verified intellectually, it could be valuable
> to report this fact - especially to keep track of "false positives"
> (e.g. matching labels, but different concepts in SKOS systems).
This is indeed very valuable information.
I often say that it is more valuable to know that two things that you might have thought were the same are in fact different, than any number of obvious equivalences. (perhaps Donald Rumsfeld said something similar :-) )
There is a property for this:- owl:differentFrom, which indicates that two URI references refer to different individuals.
However, it is rarely, if ever, used; I am still waiting to find any in the wild so I can put them at !
There is also the owl:AllDifferent class which can be used to indicate a group of URI references all point to different individuals.
> Basically, this states a relation between an entity - e.g., a
> skos:Concept - and a set of entities - as defined e.g. by a
> skos:ConceptScheme or a void:Dataset.
> Are you aware of any pattern to express this in RDF?
> I consider coining something like
>  ext:noMatchingEntity rdfs:subPropertyOf skos:note .
> Since the date of the above mentioned verification should be reported,
> you could end up along the lines the following example
>  <> ext:noMatchingEntity 
>    [ rdf:value <> ;
>      dcterms:modified "2010-01-25"^^xsd:date ] .
> What do you think?
Making the more global assertion I would think is more problematic, as you need to introduce the idea of RDF documents, datasets, or whatever, as you have.
Mind you, as the Named Graph world is becoming more accepted, I guess that might be a viable way of saying what you want in a portable fashion.

I would guess that some people would be uncomfortable with this, but doing something is good - if you have gained some knowledge, then you need to model it in RDF, and be able to put it on the SEmWeb. If you just "remember" it in a DB or something, then we haven't got very far.

> Cheers, Joachim 

Hugh Glaser,  
              Intelligence, Agents, Multimedia
              School of Electronics and Computer Science,
              University of Southampton,
              Southampton SO17 1BJ
Work: +44 23 8059 3670, Fax: +44 23 8059 3045
Mobile: +44 78 9422 3822, Home: +44 23 8061 5652

Received on Saturday, 12 February 2011 19:55:47 UTC