Re: blog: semantic dissonance in uniprot

Sad to be drawn in.

On 29 Mar 2009, at 17:15, Oliver Ruebenacker wrote:

>     Hello Matthias, All,
>
> On Sun, Mar 29, 2009 at 11:13 AM, Matthias Samwald <samwald@gmx.at>  
> wrote:
>> Oliver wrote:
>>>  As I understand it, an owl:Class is simply something intended to be
>>> instantiated. I declare something a class if and only if I intend
>>> there to be instances.
>>
>> This is how you might choose to use OWL, but it is important to  
>> emphasize
>> that many ontologies (including the OBO ontologies and parts of the
>> Neurocommons Knowledge Base / Banff HCLS demo) encode a lot of useful
>> information just by using classes and property restrictions, without
>> instances.
>
>  Isn't that the typical way, that ontologies define classes and
> properties and users of these ontologies instantiate these classes?
[snip]

Nope. It's "a" way, but it's hardly typical and the way you talk about  
it is seriously misleading.

"Instantiating classes" suggests something akin to what one does in an  
object oriented programming language. I.e., it suggests that  
individuals are "created" from templates (aka classes). While OWL  
Classes are used this way in KA systems, it requires careful thought  
(and the intervention, typically of a "sanctioning" mechanism which  
indicates which parts of the description are salient for the KA).

So, that's just not a helpful way to think about things in the owl  
context. I myself do use the "TBox=schema; ABox=data" analogy  
sometimes, but I fear that its utility is limited and risk of  
misinterpretation very high.

Second, there's lots of ways to use ontologies with out having to use  
logical constants (i.e., individuals). Alignment of database schemas  
comes to mind. There you might never lift the database data into the  
ontology, but merely use information from the alignment to rewrite  
queries.

That's not to say that anyone writes class descriptions intending them  
to be necessarily empty (i.e., unsatisfiable). Just that instance  
retrieval is one task among many.

Cheers,
Bijan.

Received on Monday, 30 March 2009 11:35:45 UTC