Re: An annotation specifying a relation

On Thu, Jul 19, 2012 at 8:24 AM, Robert Casties
<casties@mpiwg-berlin.mpg.de> wrote:
> I want to be able to create annotations that are "about" a relation
> between two (or more) targets. How do I represent that in the OAC model?
>
> For example, I want to annotate that I think that sentence S1 in text T1
> has been influenced by sentence S2 in text T2.



> I can define an annotation with two targets but I can not state that the
> two have a specific (directed or undirected) relation.
>
> What would be the best way to model this?

I suggest checking out section 4 of the extension:

  http://www.openannotation.org/spec/extension/#Semantic

You would use a Specific Resource to identify the sentences.


Two ways:

a) oax:hasSemanticTag and multiple targets

:ann1 a oa:Annotation ;
  oa:hasTarget :originalSentence, :newSentence ;
  oax:hasSemanticTag ex:someRelation .

This would be an undirected relation as targets are unordered.

This semantic tag is intended mainly for classification.


b) oax:hasSemanticTag and body/target

:ann1 a oa:Annotation ;
  oa:hasTarget :originalSentence ;
  oa:hasBody :newSentence ;
  oax:hasSemanticTag ex:quoted .

This is appropriate if the newSentence is "somewhat about" the
originalSentence, for instance


You can think of this as almost as an reification of:
  :newSentence ex:quoted :originalSentence .

However that would just be your interpretation for ex:influenced as an
semantic tag. (The model don't put any requirements for the semantic
tag other than that it has to be a resource)



c) Multiple targets, named graph as body (my recommendation):

http://www.openannotation.org/spec/extension/#NamedGraph

{
:ann1 a oa:Annotation ;
  oa:hasTarget :originalSentence, :newSentence ;
  oa:hasBody <http://example.com/aboutNewSentence> .

<http://example.com/aboutNewSentence> a trig:Graph .
}

<http://example.com/aboutNewSentence> {
  :newSentence ex:quoted :originalSentence ;
      prov:hadOriginalSource :originalSentence .
}


The advantage of this approach is that you can add arbitrary details
to the relation, and can reuse existing vocabularies (like
prov:hadOriginalSource from the W3C provenance ontology (PROV) [1]).

You do not have to use named graph serialisations such as TriG to
represent this, <http://example.com/aboutNewSentence> could (SHOULD)
be a regular RDF resource (RDF/XML, Turtle, etc) containing the body
graph.



-- 
Stian Soiland-Reyes, myGrid team
School of Computer Science
The University of Manchester

Received on Friday, 20 July 2012 08:29:10 UTC