RE: philosophy of SWBPD (was Re: [OPEN] and/or [PORT] : a practical question)

At 10:28 +0200 4/1/04, NANNI Marco FTRD/DMI/SOP wrote:
Hello,


Jim Hendler wrotes

>In case anyone hasn't figured it out by now - I THINK IT SHOULD BE
>OUT OF SCOPE FOR THIS TASK FORCE TO WRITE ONTOLOGY ENGINEERING
>DOCUMENTS HAT ARE NOT RELATED TO THE SEMANTIC WEB as part of this
>Working Group.  If you'd like me to state it clearer, let me know
>  what to addd

Do i understand well what you want to say :
         You think that, according to the formal definition of the 
term "Ontology", building an ontology doesn't automatically mean that 
you are in the SW context ?


I think that also according to many informal definitions of ontology 
-- the Semantic Web is one particular place for ontologies


         If it is what you mean i completely agree with that because 
i think that , like Mr LAPALICE, we have been building Ontology 
since, as you say, 50 years without knowing it

well, we've been calling them ontologies (Ala Gruber) since the mid 
80s, although I first learned that term in my intro AI course in 1975 
-- and I suspect others on this group go back further.  Using a 
DL-like approach to ontologies probably dates back to KL-ONE by 
Brachman in the late 70s.



But you also write :

>  They ARE central to the design of OWL, in the sense that OWL is
>  specifically FOR the Web, and thus had to have a few things that
>  typical KR/O languages lack.


Do I have to understand that what i have written above is false if I 
use OWL (RDFS ?)? In other word if i use OWL/RDFS i'm automatically 
in the SW context (SWC) ? i think i can agree with that, but let me 
ask a more precise question :
                         - Do you think that a use case (i don't want 
to use the word application) where somebody uses OWL ontologies 
without REASONING TASKS (classification, individuals retrieval, 
etc...)is still a SW use case ?

Absolutely!  In fact, I think someone using RDF with no explicit 
ontology at all is not only a SW use case, but the most important 
ones out there at the moment -- so let's not get too 
ontology-centric, although this discussion has gone there (i.e. I 
haven't seen anyone on this group mention that RDFS ontologies are 
being widely deployed and are covered in the OEP scope)



                         - if yes :
                                 this raises a few  very correlated 
new questions :
                                 - can we make a clear distinction 
between an OWL ontology built outside the context of SWC and an OWL 
ontology in the SWC ?

I guess it would be possible for using OWL separately from the Web, 
but that is sort of like talking about using HTML separate from the 
Web -- why would you want to?   Seems a pretty borderline case.

                                 - Are we able to define two distinct 
guidelines, both for OWL but
                                                 - one for the more 
general OEC (which is clearly not our objective)
                                                 - one for the specific SWC ?
                                 - In other word, (it's always the 
same question but more precise i think) : what are the differences 
between SWC and OEC ?

Again - let's use the Web analogy -- people were building hypertext 
books long before Tim BL came along.  He saw an approach where one 
used languages and protocols to link these together across computers 
in a new way, and the Web was born.   Now, in a certain sense, we 
could say all Web applciations are Hypertext apps, but not all 
hypertext apps are Web apps -- and, hostory has shown, very little of 
the pre-Web hypertext stuff turned out to be the right best practices 
for the Web -- although certainly the people from that community who 
embraced the web were crucial in helping to identify good Web 
practices (and some still write articles today criticizing Tim's 
design and saying we could have done it better if we'd stuck with the 
earlier hypertext designs -- they claim the web might not be quite as 
large and society changing, but it would be designed more cleanly)



                                        
                         - if not :
                                 to what context does it belong ? the 
general Ontology engineering context (OEC) i suppose ? And in this 
case do you think that these contexts have such a little intersection 
in  terms of guidelines that there is no need for us to explore in 
details the OEC ?

I think the OEC stuff has been explored in hundreds of papers and 
books and is a very rich literature.  I don't see any advantage to my 
organization paying W3C fees so that we can participate in a 
traditional KR context - we can do that for free in our academic 
work.   We hope this WG will concentrate on the work that helps make 
it easier for people to understand what the Semantic Web is and how 
to use it to solve their real-world problems.


                         For me the direct consequence of this 
negative response is that the very "heavy" criteria (the only one 
perhaps ) to definitively distinguish the 2 contexts is  the fact we 
need/use or not some reasoning tasks.

Don't you think that by accepting this point of view, which is 
perhaps too much restrictive, we could have a simple "bodyguard" or 
(meta) guideline or whatever you want which could say to us :

                 All the advices, guidelines,...we are going to write 
MUST be thought keeping this following final objective in the mind : 
our outputs MUST help people to build, in a given context, the best 
(distributed) architecture (i.e ontologies could be only a - very 
important - part of it) to allow some very specific reasoning tasks.

I'm afraid that taking this point of view means that we have to kwow 
for the overall SWA lifecycle all the points which can have a real 
impact in REASONING capabilities. It's a hard work but perhaps that 
it is easier than the problem to say if this point or this point has 
to deal with OEC or SWC ?


You will have understood that, my personal point of view is to make 
such simplification in our approach. Not perhaps this one exactly 
which is, i must admit, very very restrictive (and perhaps false ? 
glurps!!!) but which has the merit to define precise criterias to 
select the point to study.

I sort of like the direction you're going, but I am not sure 
"reasoning tasks" captures it -- in particular, if I use a couple of 
inferences based on OWL in FOAF (for example, by making email 
addresses inversefunctional FOAF is able to tell information that is 
about the same person when gathered from different "Knows" relations 
)  is this "reasoning"?  It does seem to me to be Sem Web and it does 
use a little inferencing, but it is definitely not classification, 
etc.

Also, I don't think that something that uses a DL reasoner to 
classify data elements it is finding in RDF data would not count as a 
Sem Web application just because it uses a reasoner.

However, I would agree that soemthing that uses a reasoner to help 
create OWL ontologies is not inherently a DL tool in itself (i.e. it 
doesn't necessarily "embrace the Web nature of OWL") but that is sort 
of like saying a browser is not really part of the Web since it only 
displays the hypertext -- it's a true statement in some sense (and 
browser design is very different than Web page design) but it doesn't 
make that much sense to consider HTML design and use without some 
idea of browsers in mind

What I do agree with the most in the above is that we need to 
explicate the life cycle of





Thank you very much

best regards

Marco NANNI


-- 
Professor James Hendler			  http://www.cs.umd.edu/users/hendler 
Director, Semantic Web and Agent Technologies	  301-405-2696
Maryland Information and Network Dynamics Lab.	  301-405-6707 (Fax)
Univ of Maryland, College Park, MD 20742	  240-277-3388 (Cell)