[topicmapmail] XTM-RDF conversion

[Lars Marius Garshol]

Hi Darina,

* Dicheva, Darina
>
> We are currently implementing an XTM-RDF conversion for TM4L. In this
> relation, I would like to ask for an opinion on translating  TM
> associations.

This problem is in my opinion already solved. An algorithm for such  
conversion is given at

   http://www.ontopia.net/topicmaps/materials/tmrdf.html#N1024

and is implemented in the OKS Samplers, so you can try it out for  
yourself. (Just use "Export" and select RDF/XML.)

A set of PSIs that can be used to guide the conversion are defined at
   http://psi.ontopia.net/tm2rdf/

> The problem is that RDF is built on directional binary relations:
> Subject-Property-Object, while in TM a binary association  authorship
> has five elements:
>
> authorship(author: Shakespeare, work: Hamlet).

The natural translation of this in RDF is in my opinion:

   has-author(hamlet, shakespeare)

The roles are just an artifact of how the information is expressed in  
Topic Maps, and are not needed in RDF.

> Currently we are thinking of translating such an association as
>
> 	Author -- wrote -- Work
>
> but are wondering whether to include also the reverse part of the
> relationship in the translation:
>
> 	Work -- written-by -- Author
>
> How would including it, or not including it, affect RDF
> completeness/readability/inference capability?

You don't need to include it, since the question of what the inverse  
is is one that exists for every RDF statement. OWL lets you handle  
this (if you want) by explicitly stating that one property is the  
inverse of another, as in

   wrote owl:inverseOf written-by

> Clearly, if we want to represent both directions, we'd have to make  
> two statements per every association instance: [...]

Actually, if you use owl:inverseOf then software can infer the  
inverse statements. Not all software does, but in theory it could,  
and in practice it often does.

> Would that be undesirable?

I think so, yes. It's better to just state it once.

> There are, of course, other more thorny issues related to scope and
> variants, but piece-by-piece.

Yes. Scope and variants is harder. :)

--Lars M.