Re: OWL1.1 APis

This will answer only one part of the issue. Still working on the  
other half.

On 11 Dec 2006, at 10:24, Dave Reynolds wrote:

> Bijan Parsia wrote:
>> On Dec 6, 2006, at 10:52 AM, Dave Reynolds wrote:
>
>>> Had OWL 1.1 had some level of OWL/full compatibility (which I  
>>> guess would have required the RDF mapping to directly expose the  
>>> punning) then some Jena native reasoning support might have made  
>>> sense.
>>> However, given how big a break OWL 1.1 is from OWL then it's not  
>>> clear to me that we can do anything useful on that front in a  
>>> sufficiently backward compatible way.
>> I'm not sure I understand this line of reasoning. I presume that  
>> "Jena native reasoning support" refers to existing rule engines 
>> +rulesets bundled with Jena (correct me if I'm wrong, please).
>
> Correct.
>
>> The *engines* seem to me to be perfectly generic. The rulesets  
>> clearly aren't, being built for specific fragments. However, the  
>> OWL 1.1 tractable fragments document identifies a *number* of  
>> fragments of OWLDL/1.1 which are suitable for impelmentation over  
>> a datalog or relational engine (e.g., DLP, EL++, hornSHIQ for  
>> datalog, and DL Lite for relational). So, it would be *quite*  
>> straightforward to produce additional Jena reasoners which use the  
>> underlying rule engines for these fragments. Moreover, it would  
>> clearly be possible to extend some of these reasoners to deal with  
>> some of the extended features of OWL 1.1 in a "reasonable" manner.
>
> Oh sure, it would be perfectly possible to identify and implement  
> fragments of OWLDL/1.1 reasoning using the same engines. Those  
> would be new reasoners, fine. That's not the issue.

It was my issue :) I'm  not sure what the original poster wanted, but  
"native" suggests "bundled" but it also suggests "first class".  
Pellet is a first class reasoner in the Jena framework (when you  
don't use the DIG interface).

I think it's straightforward to adapt as "first class" support for  
OWL 1.1 as the current bundled reasoners provide for OWL. These might  
not be bundled, but I'd like to distinguish these aspects. For  
example, updating KAON2 to OWL 1.1 requires more work than Pellet, if  
just because 1) pellet was closer to supporting 1.1 and 2) the logic  
underlying OWL 1.1, SROIQ, was presented with a tableau algoirthm. To  
update KAON2, there needs to be some development of algorithms as  
well as implementation and optimization.

For the Jena bundled reasoners, it's pretty easy to give as good  
support for OWL 1.1 as is currently for OWL, just by adding some new  
rulesets. Such reasoners will benefit from any changes to the core  
engines as well. Plus, there will be Pellet. TopBraidComposer uses  
Jena and Pellet, so there is likely to continue to be good support  
for the blend.

The take away, I think, is that if you want excellent support for OWL  
1.1, Jena will still be quite viable, and in a fairly short time frame.

> The key phrase in the paragraph was "in a sufficiently backward  
> compatible way".
>
> What happens when someone applies an existing OWL full reasoner  
> (the Jena reasoners are all tiny fragments of OWL/full) to an OWLDL/ 
> 1.1 document? It seems to me that would be unsound, any use of  
> punning in the OWLDL/1.1 document would lead to incorrect entailments.

There will be more entailments, but that's the situation *now*.

> Up till now the semantic web stack has nested nicely. If I apply an  
> RDFS reasoner to an OWL document (whether DL or full) all the  
> deductions are sound.

Ok, now I'm getting into the second bit :) I don't think this is  
true, and not a sensible wish anyway. It certainly isn't the case for  
Simple Interpretations vs. RDFS Interpretations (i.e., some graphs  
are consistent on simple interpretations and inconsistent on RDFS  
ones...through in D-entailment and you get more such; by the time you  
hit OWL Lite, well, you see where it's going :))

> If I apply an OWL/full reasoner to an OWL/DL document the results  
> are sound.

Please explicate this a bit better. You certainly get entailments not  
sanctioned by the OWL DL semantics. OWL Full is a semantic extension!

> It seems to me that OWL/DL1.1 breaks this nesting, if I apply an  
> OWL-(full)-as-currently-specified reasoner to an OWLDL/1.1 document  
> the results are not sound.
>
> Is that right or am I missing something here?

One of us is :)

Cheers,
Bijan.

Received on Thursday, 14 December 2006 11:12:22 UTC