RE: A question on the vocabulary for 'persons' - ACL level of granularity?

--Dan, 

> Xiaoshu Wang wrote:
> >>> IMHO, inadequate separation of ontology's domain will have some 
> >>> serious side effects in the long run.  Aside from wasted
> >> bandwidth and
> >>> computation to handle the unnecessary statement, but when more 
> >>> ontologies are shared, the chance for incur conflict will
> >> increase and
> >>> makes the sharing ontology impossible.
> >> In my opinion, we should stop to care too much about single, 
> >> delimited ontologies. When I want to reuse some parts of 
> FOAF while 
> >> leaving out the ridiculous parts ('geekcodes', 'dna-checksums' and 
> >> other jokes), I would simply extract the classes and 
> properties that 
> >> I need and add them to my ontology / software. As I see it, all of 
> >> these statements should be seen as a global graph of RDF nodes and 
> >> arcs. How these are represented locally inside .rdf /.owl files or 
> >> through SPARQL endpoints is secondary.
> > 
> > I wish it could be that simple when you handle the task to 
> machine.  
> > Show me how you can only import the foaf:Person without 
> fetching the 
> > foaf:geekcodes as well?
 
> You might for example use SPARQL to take a subset. There are 
> already inline annotations indicating property and class 
> (im)maturity levels.

Of course, SPARQL can help to get a subgraph.  But it is much more an
"application" kind of semantics, where ontology should be just "declarative"
semantics. I am not sure how it helps ontology modulization. 

> Regarding fitness for semweb-lifesci purposes, my take is 
> that FOAF should be something that is useful for describing 
> the members of the community and (at a high level, alongside 
> of course Dublin Core, SKOS thesauri, and specialist 
> vocabularies/ontologies) their professional activities and 
> collaborative outputs - weblogs, online presentations, 
> publications etc. If it needs to change to achieve this, 
> we'll look at that.
> 
> Its usage within scientific and professional datasets 
> directly is something else. I would not expect to see the 
> FOAF namespace directly used, for example, in clinical trials 
> data. I wouldn't expect to see vCard either. In such contexts 
> it often makes sense to define one's own vocabulary, even if 
> it involves duplicating common namespaces (Dublin Core, 
> FOAF/vCard), to keep some local control and predictability. 
> It might be that someone writes OWL mappings (sameAs / 
> subclass etc) to link together different but essentially 
> identical notions of 'Person'.
> Whether such assertions are actually made within the 
> documents found at the relevant namespace URIs is again a 
> separate set of decisions.

I actually think that each ontology designer should think beyond its own
community  because in the long run, we don't know how things will be
interact with each other.  As Drew has mentioned before that we don't want
to: "(a) misuse an ontology; (b) duplicate what someone else has already
done."  If, for instance, every scientific discipline or subdisciplines
develop its own Person specification, we will then be forced to try to do
ontology merging and mapping, which, in my opinion, is no easier than
database/xml merging/mapping.  We then take away the advantage of open world
approach.

But, if we all have carefully modulized our ontology, for instance, if we
break FOAF into two parts: a Foaf_Person and Foaf_online.  With the
foaf1:Person ontology focus on the commonly shared property such as name,
email, fax etc., I don't see why the life-science community would not want
to extend it given the popularity of foaf.  Or on the other hand, if
everyonelse except foaf is using another Person ontology, foaf can realign
its to this more popular Person ontology without breaking the stability
associated with Foaf_online.    

Cheers,

Xiaoshu

Received on Monday, 18 September 2006 03:24:51 UTC