Re: Rich Annotations

On 23 Nov 2007, at 17:47, Jeremy Carroll wrote:

> Bijan Parsia wrote:
>> On 23 Nov 2007, at 15:36, Bijan Parsia wrote:
>> [snip]
>>> [1]
>> [snipped complaint about my proposal being called "wacky"]
>
> I suspect in my idiolect 'wacky' is more complimentary than in  
> standard british english, and more so again than in US english.

Ah, if that is so, then my apologies for taking it up the nose. A  
good demonstration as to why replying to such issues on list is a bad  
idea. I'd like to say that this was all planned as an object lesson  
but in truth I was just miffed and lost my cool. Oh well.

30 lashes with a twizzler at the face to face for me.

> I have quite a lot of admiration for the idea at a technical level.

Great. Let's proceed on that basis.

Let's start with annotation spaces which are 1) expressed in owl and  
2) all mayIgnore.

Suppose I have an OWL document O with annotations spaces A1 and A2.  
(Let's work at the functional syntax level for the moment, to elide  
encoding issues.) Let's presume a projection function P which simply  
takes a document O, and an annotation space A and returns the set of  
annotations in that space as axioms, i.e., as an OWL document. The  
*effect* of the proposal at an OWL DL level for an owl reasoner is  
that we have P(O, A1), P (O, A2), and O where all the annotations are  
simply ignored. A future query language (SPARQL/OWL for example)  
could treat O as a *dataset* with three graphs in it O - annotations,  
P(O, A1) and P(O, A2).

If you want to fullize this, you have choices. You could do the same  
thing as above, no harm, no foul. (Stick the annotations in literals,  
for example.)

You could also just take the union of the three. You might have a  
couple of extra assertions (e.g., that such a property belonged to an  
annotation space), but that's all.

I think the former technique is better because I think that the  
isolation between annotations and model is useful. But we can go  
either way.

I would argue that this is a small change because old ontologies  
(with annotations in the "Default space") can be interpreted exactly  
as before. Basically, it's a bit of syntactic sugar for multiple  
documents.

Cheers,
Bijan.

Received on Friday, 23 November 2007 18:10:54 UTC