Re: beyond 'formal' relations: describing relations between scientific and non-scientific material

Dennis, this is very interesting. I am not an expert specifically o
nthese but i see 2 ways

a) use a proper ontology, so keep the model in "triples". Which
ontology? there can be some to start from, definitely you'll have to
extend them however for your own purpose, which is totally allowed

b) do it the relational DB way in RDF, create a class called
"relationship defined by a wordnet term" with a property that is
"from" another that is "to" and another that is "kind".. from and to
goes from your subject to your object , the "kind" points to, say, a
wordnet entry or a dbpedia entry which defines the relationship.  (ok
in RDF you might have done with reification but this is less triples,
possibly more clear.)

somebody might be screaming about the b) approach as that of betraying
a bit the triple nature of RDF, but reality is that.. these sort of
more advances structures have totally legitimate and unavoidable uses
and people (clients) will have to live with them.
Matter of fact it is no problem whatsoever for somebody knowing this
structure to use it, e.g. in a sparql query.

Just to give you an example recently i was looking at RDF from
freebase. They have a structure called "participation of an actor in a
moving starring in a role" it is a structure like the one above which
ties together an actor, the movie and the role in which is portraied
in the movie (e.g. dr. evil).

We all would have loved a simple triple "actor -> movie" but the truth
is that the above structure is useful as it allows many other queries,
once understood and leveraged. My take is that smart semantic web
clients will simply more and more put rules to adapt to different data
structures as needed.

b) has the advantage that one could explore the relationship[ between
your describing terms exploring the wordnet words that are linked

again.. this is my very naive view, never really worked on this.

Giovanni


On Wed, Apr 1, 2009 at 9:33 AM, Dennis - UT <dv.eprints@gmail.com> wrote:
> Hi,
>
> We are currently working on a repository for OAI ORE resource maps
> (http://www.openarchives.org/ore/1.0/toc). In this system we are trying to
> describe relations between scientific publications and other material (both
> scientific and non-scientific). To do this we are planning to use several
> (RDF) vocabularies / ontologies.
>
> A question is: how to cope with diversity in scientific disciplines and
> communication on the one hand and standardizing relation descriptions when
> aggregating publications about a certain topic? Vocabularies now available
> (FOAF, DCterms, etc) mainly restrict to formal relations and do not include
> relations concerning the content in a more detailed way than for instance
> 'dc:subject'. This may be the consequence of the diversity in scientific
> semantics. Is there any literature/article about this issue?
>
> An example case is describing relations between scientific publications and
> their 'application'. For example: a publication proposes certain changes,
> government policy makers later decide to create actual policies based on
> this information. So far we didn’t find any existing solution to describe
> such relations. Suggestions on existing vocabularies to describe / annotate
> such relations are very welcome, thanks!
>
> Kind regards,
>
> Dennis
> University of Twente
>
>

Received on Wednesday, 1 April 2009 14:42:53 UTC