There is still an active debate on whether a Taxon (TaxonConcept) is a individual or a class (at least in my mind). From the OWL perspective it can't be both or it pushes us into OWL Full and possible undecideability.
Should we not do anything until we can settle this discussion (it is at the heart of modelling our domain) or should be simply agree that it may be modelled as either class or instance but that we still be defined it as existing?
I prefer the latter course. The 'correct' or optimal answer to the question depends on what the model if for and as that is incredibly vaguely specified it provides us with no grounds for testing proposed solutions. We could discuss till the cows come home and no one would ever be right or wrong and we would never have a way of ascertaining when we got to the end of the discuss. Life, after all, is short.
On the specific example that Bob mentions I wonder why we need the property at all. The notion of ownership is not restricted to biodiversity informatics.
dc:rightsHolder http://purl.org/dc/terms/rightsHolder "A person or organization owning or managing rights over the resource."
Would probably do the job and be understood by every application that 'understood' DC. Likewise a specimen is dc:isPartOf a collection (A related resource in which the described resource is physically or logically included.)
I am not aware of a way of suppressing assertions imported as part of an ontology. If I import TaxonConcepts that some one has helpfully tagged with dc:title assertions then they will be treated as instances by OWL not classes (classes aren't the subject or object of propeties only individuals are) unless I declare dc:title an OWL annotation property which will of course remove dc:title from all inference where it may be used elsewhere on real instances of things in my combined ontology. The only thing to do is to manually (or programmatically) remove all the offending triples on import. I'd like to know a better way of doing this.
But I should shut up as I am supposed to be on holiday.
All the best,
Roger
On 18 Oct 2009, at 15:38, Jonathan Rees wrote:
On Sun, Oct 18, 2009 at 10:02 AM, Roger Hyam rogerhyam@mac.com wrote:
Hi Bob,
This is why I believe we should not make assertions in the core ontology about range and domain of properties.
If some one wants to import these notions (terms) then they are forced to also import our (possibly flawed) world view.
If we were to create a perfect model of the reality of the biodiversity informatics domain in OWL (or XML Schema or UML or alphabetti spaghetti) due to Hyam's unoriginal law* it would be wrong next week or certainly by the next TDWG meeting.
There is a different view, adopted by many ontology projects: You do the best you can, and then if you find a mistake in the ontology, you fix it.
Without guidance such as domain and range, you increase the probability that curation will be inconsistent across projects. And inconsistency makes data integration harder down the road, pushing work that ought to be done once, at curation time, off onto hapless consumers, each of whom has to do cleaning and normalization separately.
The ontology is a significant point of leverage and it would be a shame not to use it.
The OWL reasoners are really helpful at findings bugs like this one. Public review helps too, as we have just seen.
If someone wants to ignore domain, range, subclass, etc. because they're afraid they contain mistakes, that's pretty easy to do. If someone wants to know how to use the ontology consistently with how others are using it, that's hard to do, if the information they need isn't in the ontology.
Jonathan Rees Science Commons
This is why we should perhaps have a list of terms (bound to URIs) and a separate set of models that are used for particular occasions. Ultimately the semantics of the data is governed by the query that is being asked i.e. it is context sensitive. "Give that XYZ, what are the ABC?"
- Hyam's unoriginal law states "The future is different - that is how
we tell it isn't the present."
Hope this helps,
Roger
On 17 Oct 2009, at 22:45, Bob Morris wrote:
http://rs.tdwg.org/ontology/voc/Collection.rdf#hasOwner specifies two classes in its range: Person and Institution,
The formal semantics of rdfs:range entails that the actual range of an object property is the intersection of the named rdfs:range classes.
I believe this all entails that the only owner of a Collection is something that is both a Person and an Institution.
Also, since these classes are not(?) defined as disjoint, the result is that any set of triples Roger rdfs:type Person C hasOwner Roger turns Roger into an Institution
Likewise Kew rdfs:type Institution C hasOwner Kew turns Kew into a Person
I could be wrong about all this, perhaps unless I too am an Institution, since, in my experience, Institutions rarely admit to being wrong. (Uh, oh, better not go there... :-) )
Bob Morris
p.s. This also suggests that the stylesheet human.xsl is somewhat recalcitrant, since the html rendering of http://rs.tdwg.org/ontology/voc/Collection.rdf only sees Institution as the range of hasOwner
-- Robert A. Morris Professor of Computer Science (nominally retired) UMASS-Boston 100 Morrissey Blvd Boston, MA 02125-3390 Associate, Harvard University Herberia email: ram@cs.umb.edu web: http://bdei.cs.umb.edu/ web: http://etaxonomy.org/FilteredPush http://www.cs.umb.edu/~ram phone (+1)617 287 6466 _______________________________________________ tdwg-tag mailing list tdwg-tag@lists.tdwg.org http://lists.tdwg.org/mailman/listinfo/tdwg-tag
tdwg-tag mailing list tdwg-tag@lists.tdwg.org http://lists.tdwg.org/mailman/listinfo/tdwg-tag
tdwg-tag mailing list tdwg-tag@lists.tdwg.org http://lists.tdwg.org/mailman/listinfo/tdwg-tag