[Orechem] Document ontology and instance prototypes

[Frey J.G.]

Is this consistent with the Experiment Ontology put up by Mark?
Jeremy



On 28/07/2009 19:38, "Carl Lagoze" <clagoze at gmail.com> wrote:

Hi all:

I just put up on the wiki at http://services.nsdl.org/trac/oreChem/wiki/DocumentOntology the results of a good bit of work I've done on a proposed document ontology for the project.  The text below is the guts of the wiki entry.  I realize the work is not yet complete, but I think it gives a reasonable working model of what the PSU folks should use to represent info extracted from documents (I used the spec sent by Lee GIles as my requirements document).  As I say in the wiki text, the model is ORE compatible, which I consider quite important.

I'd work more on this but am leaving for vacation on Thursday through August 16 (no email contact) and wanted to get this out.  Flame at will but I won't respond till return.

Now onto my N other deadlines before leaving.

Carl



PPrototype document ontology and instance <http://services.nsdl.org/trac/oreChem/wiki/DocumentOntology#Prototypedocumentontologyandinstance>
I've attached to this wiki page three files that show my proposal thus far:

 *   orechem.owl (probobly should be rdfs) - This is a set of vocabulary terms that I had to invent for the ontology. Virtually all are specializations (subclasses or subproperties) of terms from other vocabularies.
 *   orechem-article.xml - An rdf instance file illustrating a rather complete example of the use of orechem.owl and lots of other vocabularies. I models an article in a journal. To do that it uses vocabulary terms from lots of namespaces including:
    *   Dublin Core elements and terms
    *   foaf
    *   ore
    *   chemaxiom
    *   bibo
    *   vcard
 *   orechem-article.svg - A graphical display (From ISAVIZ) of orechem-article.xml. Lots of triples here so you'll have to zoom in.

You will note that the example is compatible with the ORE notion of an Aggregation (and describing Resource Map), modulo a few missing metadata elements. This was obviously a priority for me since we want to fit into this generic space.

Also, probably much to Nico's chagrin, this is all an RDF-based rather than OWL-based model. Sorry, but too much of the world (e.g., DC, ORE, bibo) does not fit in the OWL space, but in the more generic RDF space. We do loose inferencing ability, but personally I don't see that as important until triple stores scale-up to that task.