Re: Profiles intro

Hi Ivan,

On Thu, 10 Apr 2008, Ivan Herman wrote:
>
>> [...]
>>> [[[
>>> In some cases, the emphasis is on possibly very large concept hierarchies 
>>> and the corresponding queries, rather than on sophisticated 
>>> characterizations of, and with properties (roles). EL++ is a profile where 
>>> such queries can be responded in polynomial time.
>> 
>> Hmm. The term "concept hierarchies" sounds too weak for what you can
>> do with EL++. Indeed, you *can* relate classes in terms of properties
>> in EL++ (also in OWL-R, only not in DL Lite). EL++ is a true ontology
>> language and can do much more than describing hierarchies. For
>> example, it *does* have sophisticated facilities for talking about
>> roles.
>> 
>
> I am trying to find what the _emphasis_ is, and not to be exhaustive! 
> That is why I used the word 'emphasis'. 
> What would you put in place instead?

My "high level of abstraction" was intended to serve that purpose. 
That's of course rather, err, abstract. You seem to want an intuitive
description of the expressive power of EL++. Not easy. Things like
"positive" and "existential" come to mind. But that's logic parlance. 
Or do you want a description of its intended use?  That's much more
simple: large-scale ontologies like SNOMED and NCI.

>>> In other cases, the emphasis is to provide a minimal classification and 
>>> federation vocabulary over a possibly very large set of data (typically in 
>>> the form of RDF triplets), while maintaining efficient querying. Two such
>> 
>> DL Lite seems unrelated to RDF triplets. It is only OWL-R that is very
>> RDF-ish.
>> 
>
> Ouch. That hurts. This is a Semantic Web ontology, so if an ontology is 
> unrelated to RDF triplets, than what does it have to do with this group?

I guess I misphrased this when writing "unrelated". What I meant was that

- The inventors of DL Lite are not semantic web people, and did not
   build DL Lite on RDF in any way (which does not imply that DL Lite is
   useless or weird from an RDF perspective). Reasoning in DL Lite is
   completely different from standard reasoning approaches to RDF/RDFS.

- The design of OWL-R (i.e., the choice of included constructs) is,
   as far as I understand, as follows: start with RDF and RDFS and the
   rule-based reasoning that is standard in that area, and add as much
   OWL expressivity as possible while staying in the RDFS reasoning
   paradigm (see the ter Horst paper).

So DL Lite was not *made* for RDF. But if you *want* to, you can
clearly use DL Lite with RDF and get sth very useful out of that. DL
Lite is for use with relational databases and these can store triples.

> I do 
> not see why OWL-R would be more RDF-ish than DL Lite or vice versa.

I still think this is a valid view, see above. Whether we should
emphasize this view in any of our documents is another question. If
you as a W3C person are so strictly opposed, maybe not.

> We are talking about data and, possibly, lots of it. Those are typically  RDF 
> or RDF-able on the Semantic Web. That DL-Lite may be of interest outside of 
> the Semantic Web may be true, but is besides the point in this environment...

Apologies, but this seems like a very strong claim to me. There are
different communities in this WG who come from different backgrounds.
As I wrote in my previous mail, I have strong reservations against DL
Lite and OWL-R as *ontology languages*. But still I don't claim that
they are out of scope for this WG (which is about ontology languages),
and I understand that they are useful in a context different from the
one that I am interested in. Clearly, there are people using OWL in an
RDF context, and there are people using OWL in other contexts. I feel
that if we want to do a good job in this WG, we should try to appreciate
and cater for all meaningful uses of OWL (at least as long as they
don't interfere with the semantic web use).

>> [...]
>>> I also believe that the names of the profiles should somehow reflect the 
>>> high level characterization. Something like:
>>> 
>>> EL++     -> OWL HI (for hierarchies?)
>> 
>> I would be very unhappy with this. As noted above, this sounds very weak
>> given what EL++ can actually do.
>> 
>
> Yeah, I was not very sure about that. But what else?

No good idea as of now. Things like "large-scale ontologies" don't lend
themselves easily towards a name. The trouble is that the intended use 
and the characteristics of EL++ are in some sense close to that of OWL DL
(at least closer than in the cases of DL Lite and OWL-R). Will think about
it.

>>> DL Lite  -> OWL DB
>> 
>> Sounds reasonable. But is Zhe happy with this? One of the main reasons
>> d'etre for OWL-R seem to be the implementation in DBs with rules.
>> 
>
> But, well, OWL DB can be implemented without rules, right? So it is, in some 
> sense, more closely bound to databases than OWL-R.

I agree.

greetings,
 		Carsten

--
*      Carsten Lutz, Institut f"ur Theoretische Informatik, TU Dresden       *
*     Office phone:++49 351 46339171   mailto:lutz@tcs.inf.tu-dresden.de     *

Received on Thursday, 10 April 2008 16:33:26 UTC