Re: blog: semantic dissonance in uniprot

On 30 Mar 2009, at 16:38, Oliver Ruebenacker wrote:

>      Hello Bijan, All,
>
> On Mon, Mar 30, 2009 at 11:31 AM, Bijan Parsia  
> <bparsia@cs.man.ac.uk> wrote:
>> On 30 Mar 2009, at 16:12, Oliver Ruebenacker wrote:
>>>  Can you name any popular ontology that does not primarily declare
>>> classes and properties?
>>
>> I don't have to, since I'm not claiming that.
>
>   Then it is not clear to me what you are claiming.

I'm claiming that classes in OWL are not typically intended to be  
"instantiated" by users (in OWL).

It could be that the publisher/modeller instantiates them in their  
modeling (e.g., FMA).

It could be that the class hierachy *is* the intended output (most  
controlled vocabularies, e.g., NCI Thesaurus, SNOMED).

It could be that the data is never going to be in an ontology or  
asserted to be instances of classes in any way.

It could be that users will simply extend the class hierarchy.

It could be that you don't say that it's a member of a named class,  
but you only want to infer it.

Of course, people *do* use OWL as a data KB. It might be that that  
shall come to dominate.

Even there, a lot of interesting uses will no neatly fit in that  
mental model. I think it's unhelpful.

Cheers,
Bijan.