- From: NANNI Marco FTRD/DMI/SOP <marco.nanni@francetelecom.com>
- Date: Thu, 15 Apr 2004 18:49:47 +0200
- To: "Jim Hendler" <hendler@cs.umd.edu>, <public-swbp-wg@w3.org>
- Message-ID: <BBBE5BAA3B351C488C415EA662EA88400B712D@ftrdmel2.rd.francetelecom.fr>
Hi, I quite agree with you. What i wanted to express with my comment was that when i see "the ontology is in OWL FULL", according to what i can read in the OWL guide "OWL Full is meant for users who want maximum expressiveness and the syntactic freedom of RDF with no computational guarantees. " people can think that it's, at first or at least, not a good thing to be in OWL FULL. I think of course that it is neither a bad thing nor a good one but i also think that people can't refrain from thinking that. What does "no computational guarantees" mean for a relative new comer in the SW world ? DL world and reasoning problems are not easy things. And i think, i'm convinced, that it's (or it could be) easier for people to choose the first Natacha, approach because it's more straightfroward and clean and by consequence to understand metamodeling than to : - generalize the other approachs in other use case, - and to understand DL and reasonning problems in general way without concrete help or examples What does "it is unlikely that any reasoning software will be able to support every feature of OWL " exactly mean ? . Does it mean that i will have core dumped when i will try to validate my ontology ? an infinite loop ? a reject message like : " unable to validate you ontology because ..." or some error/warning message ? in this last case if i decide to use my ontology in spite of these messages what kinds of problem will i have at running time according to my use case ? To be more precise : suppose that in the image use case of Natacha my job is only to receive images and to classify them, if all the images are annotated in such a way '(a good way) that i will never need to ask the reasonner to classify them because the name of the class explicitly appears in the annotation i don't care if i'm in OWL FULL. In this case i'm lucky. But, generally ideal world don't exist, and i will probably meet some page for wich the fact to be in OWL FULL will have perhaps heavy consequences for the integrity of my "running" application. (Note : behind this problem there is an other important and complicated question : do, ontology creators, have to build Ontologies very closed to their uses case or to a particular use of the ontology OR do they have to anticipate the maximum numbers of particulars uses ? In this last case i think that they'll always be obliged to consider very seriously the problem of having OWL Full ontologies or not) That's why I think that it could be fruitful, if we are able to do that, to give a list of the most current "possible" problems people can encounter with OWL FULL Ontologies. If you think it's useful, i think that it could be great to have two distinct things : 1) - for each point (e.g "classes as values as Natacha described) and for each solutions we could add to the Deborah proposal (for example) > my summarization of these is: > problem description (including background and observations could fit in here as well) > use case example > abstracted solution > owl solution > implications > references > if as in natasha's example, there are many solutions, then repeat abstracted solution, owl solution, implications, references > as appropriate. concrete examples of what OWL FULL implies when a solution put the ontology in the OWL Full world 2) - a more general chapter titled for example : "OWL Full : what does it imply ?" which could contains even the same examples of the listed points above I think it will be very helpful for readers to have these two different ways or two distinct entry points for these problems in the same document. This could be fruitfully completed with a good index where we could find entries like OWL Full - computing problems see : - patterns descriptions : page xx, yy - chapter "OWL Full : what does it imply" : page zz To start the work i could propose to try to take the Natacha uses case as a starting point and to show two different "instances" of it : one for wich to be OWL FULL is not a problem an the other for which you can have some problems. Thank you very much best regards marco -----Message d'origine----- De : public-swbp-wg-request@w3.org [mailto:public-swbp-wg-request@w3.org]De la part de Jim Hendler Envoye : jeudi 15 avril 2004 16:24 A : NANNI Marco FTRD/DMI/SOP; public-swbp-wg@w3.org Objet : RE: [OEP] "Classes as values" first draft At 12:52 +0200 4/15/04, NANNI Marco FTRD/DMI/SOP wrote: Perhaps that the fact to describe a blocking situation in the context of this use case due to the fact that we are in OWL FULL, could help people to better realize what to be in OWL FULL really means. My request comes from the fact that when i tried to find such a situation/example it was more difficult than i thought at first Sure, as long as for any "blocking condition" that someone claims is caused by being in OWL Full, we also include blocking conditions which result from being in OWL DL/Lite (there are many that I'm encoutering in my work these days) and also that we explore some new paradigms that are yet relatively sparse -- for example, Mike Dean had some nice examples (I think it was at a DAML meeting) of the idea of, essentially, taking the OWL DL subset of an OWL Full document and it using it for some reasoning (classification) tasks. Again, what I ask is that we remember we're in largely unexplored space and we need to be very careful of being judgmental -- I have met many people who believe in the future OWL DL will cease to exist and everyone will just use something called OWL, I have met many people who believe in the future OWL Full will cease to exist and everyone will just use something called OWL, and I've met many people who think it will continue as current - with perhaps more OWL profiles growing over time (for example, the Gene Ontology folks have been thinking about how to interact w/OWL, given they consider part-whole to be the most important kind of representation for the applications they run) -- in short, predicting the future is always difficult and we should be careful to embrace multiple views -- 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)
Received on Thursday, 15 April 2004 12:52:43 UTC