- From: Bijan Parsia <bparsia@cs.man.ac.uk>
- Date: Sun, 13 Apr 2008 11:26:36 +0100
- To: Ivan Herman <ivan@w3.org>
- Cc: OWL Working Group WG <public-owl-wg@w3.org>
On Apr 12, 2008, at 11:07 AM, Ivan Herman wrote: > Hi Bijan, > > I was trying to find a good reason to shoot you, but I did not find > any:-( Sigh...:-) I must be losing my touch... > Seriously, I find the text really good as a start. Deep bow towards > Manchester... Thanks! > I was not sure how you wanted this text to evolve. I first tried to > use the {{Review}} macro but it seems that the style did not work > for this (why?) I think the style isn't site wide. Sandro, I think having both EdNote and Review being site wide would be a good idea. > so I was not sure how to do it. I ended up writing all my comments > in this mail.... I hope that is fine with you. Yep. > ---- Introduction ---- > > (I am not sure how to handle that, actually): with this numbering > scheme we have now, when I use the word 'OWL', do I mean 'OWL in > general, both OWL 1 or OWL 2', or do I mean OWL 1? I generally mean OWL 2 DL. This is just what happens, not policy. In the primer, we've not determined any policy yet. > I tend to think now in terms of the former, but it is not 100% > clear what you mean in the text (you specifically talk about OWL 1 > in the last section). It may be worth emphasizing that what you say > about Full/DL is valid both for OWL 1 and OWL 2, ie that OWL 1 has > adopted this fundamental "cut", and OWL 2 has adopted it, too. My belief, in general, is that OWL 2 will replace OWL 1, at least conceptually. So, in the primer, I try to localize all the backward looking discussion. In the OWL 1 section, OWL DL and Full are discussed as well. > You say (about OWL Full): "One can still define classes using > restrictions or constraint properties with property chains, and so > on.". That wasn't about OWL Full. > Why 'still'? As OWL Full is unrestricted, there is no 'still', so > to say... That sentence is rewritten. > (b.t.w., I do not think that remarks like "(subject to resource > constraints)" or "(modulo bugs in the implementation)." are > necessary, mainly if this text end up (as I think it should) as > part of an official document..., but that is a stylistic issue > only...) When writing text that may possibly strike some advocates as making a case for a dialect that they do not like, I think it's probably a good idea to act prophylactically against some obvious talking/ complaint points. Plus, the former is, I think, a useful point for end users (the latter is merely prophylactic...a response to a point raised loudly in earlier discussions). > ----- EL++ ----- > > I know the role of SNOMED and NCI played for EL++, but I am not > sure that should be part of such a profile description. First of > all, most of the people will have no idea what these ontologies > are, and, secondly, it gives the wrong impression that we define a > profile for a particular branch of industry. The point was to give an idea of the sorts of ontology EL++ is good for. I provided some additional characterization. We may want to replace the reference to the particular ontologies with the characterizations. > My reading of EL++ is that some of the property characterization > ([inverse]functional, symmetric) are also missing. I already mention that inverse is missing. > I think they are also part of the 'key missing features' that you > began to write. (Well, your easy key may change the importance of > inverse functional properties in future, although I have no idea > whether those would fit into EL++). > > ----- OWL-R ----- > > It is *really* not the way of shooting you:-), You *know* you want to! > but I do not understand what you mean by "all the OWL Fullish > features of RDFS"; but you say that yourself that it is not > helpful... If you mean reflection, well, at the moment, OWL-R does > not have that a.f.a.i.k. (that was the essence of the issue I had > on the axiomatic triples)... The only thing I see that may be > possible in OWL-R (well, OWL-R-Full, actually!) is the usage of the > same symbol for a data and object property. I may miss something... I was riffing off some stuff Jeremy has said and probably botching it. > Personally, I consider OWL-R (and at DL Lite, too, actually) as a > kind of a minimal set of extra ontology features that I can build > on top of RDFS, while keeping it still easy both to comprehend and > to implement. Well, on top of RDFS DL or RDFS Full :) > Ie, when you say "OWL-R is ideal for enriching RDF data", I think > it is not only a true statement but may be _the_ essential feature. Good! Now if I only knew what I meant by that... (Something like OWL-R isn't really intended for class expression inference.) > Although I would add something like ".. with a set of minimal > ontological statements going beyond RDFS." > > Also, I also do not understand why the following is true: > > "Compared with DL Lite, OWL-R works better when you have already > massaged your data into RDF and are working with it as RDF." > > (I would not use the word 'massaged' here, b.t.w.) I could very > well imagine to (1) have my data in RDF (2) add a minimal set of > OWL statements to further characterize my data and... end up in DL > Lite. That's certainly true. > Ie, I do not see that as a differentiating feature from DL Lite. > Both are a simple extension that works very well on top of RDF data. I was thinking that OWL-R works more naturally in ETL scenarios whereas DL Lite was designed to support federated query. But this may just be historical emphasis. > I guess this still reflect a difficulty that I still have in making > a clear high level (ie, user level) distinction between the usage > of OWL-R and DL Lite. Maybe your remark in OWL-R saying "when the > data must be massaged by additional rules" is the strongest > distinction here... I am curious to hear Zhe's opinion. Adding extra rules (e.g., DL Safe rules) to OWL-R doesn't change the complexity of inference; adding DL Safe rules (recursive ones for sure, at least) to DL-Lite does. > I think both in the case of DL Lite and OWL-R it is worth > emphasizing somewhere that they scale very well with the amount of > data. (RDF or otherwise:-) This is true of EL++ as well. > B.t.w., you said for DL-Lite: "DL Lite restricts class axioms > asymmetrically, that is, you can use constructs as the subclass > that you cannot use as the superclass.". Isn't that statement > correct for OWL-R, too? Yep, even moreso. But the steam, it ran out :) > As for when you ran out of steam: I wonder whether this document > should talk about the OWL-R-DL and OWL-R-Full differentiation at > all. According to section 4.4 of the profile document in almost all > the cases these two coincide in practical use, so the difference > between the two is really only mathematical, so to say. We could > actually say that on the OWL-R level the distinction between OWL > Full and OWL DL is mostly gone, and that is it... But I may have a > technical misunderstanding here. I don't have a good sense here. > Finally, as for your comments below on what to do with the text: I > agree that it should not be part of the profile document. However, > I am not concerned to have this text (plus some examples) being > part of the primer; actually, on the contrary. The current primer > (that I really like, do not take me wrong!) may not be easy to read > for the constituency of possible OWL-R and DL-Lite users. Ie, it > would be fairly difficult for them to understand what they can and > cannot do if the only thing we offer is the current primer and the > profile spec. Well, we can always scatter (hideable) comments throughout the primer. I.e., "This construct is unavailable in EL++". > Although your idea to put it as a separate document: that is worth > considering... A separate primer for EL++, yet another for OWL-R? Or make the current primer configurable a bit. > If we had infinite manpower, that might be the best option... But > even if we do that, I would definitely prefer to keep this at the WG. I'm not sure. Cheers, Bijan.
Received on Sunday, 13 April 2008 10:27:18 UTC