- From: Ian Horrocks <horrocks@cs.man.ac.uk>
- Date: Sat, 4 Jun 2005 07:43:36 +0100
- To: Paul Gearon <gearon@itee.uq.edu.au>
- Cc: public-owl-dev@w3.org, Jim Hendler <hendler@cs.umd.edu>
On 3 Jun 2005, at 14:53, Paul Gearon wrote: > > Going back a couple of months now... > > On 07/04/2005, at 11:14 PM, Jim Hendler wrote: >> (NOTE: This is a private message, it has no link to the Sem Web >> activity, or anything else -- I write this wearing my professor hat!) >> >> In recent days I've been attending a lot of meetings where I have >> been approached by people talking about not just the limitations of >> the current OWL (something I've been hearing about for a long time >> :-)) but actually talking about proposed technical solutions. I >> think it would be good to start to collect some of these and to think >> a bit about those things that we might want to see go into some >> future OWL version >> There's a wide variety of these things going from simple extensions >> to OWL (such as adding qualified restrictions, having an >> owl:allDisjoint , etc.) to adding some standard ways of doing common >> things in other KR langauges (part-whole, bounded transitivity, >> probability models) or going beyond to new concepts in Sem Web (new >> models of partial import, named ontology segments, etc.) >> I'd like to hear what people are working on, or what people need - >> this way we'll have these ideas on record for the eventual next >> generation of OWL technology. > > This may appear quite naïve, but I've noted that OWL permits many of > the accessibility relationships of modal logic (transitivity, > reflexivity, symmetry) but not relationships that are Euclidean. What > I mean, is to say for a relationship R1: > R1(A,B) > R1(A,C) > => R2(B,C) > > For some relationship R2. This would seem useful in a case like the > property "parentOf" if it were to relate to "siblingOf". > > At the moment this sort of thing is done with rule systems, but it > seems to me that the way relationships relate to each other is also a > part of an ontology. Rule systems are often very specific, while > ontologies are much more general, so I think it's better to express a > concept in an ontology, if possible. > > The above example relies on R2 being symmetric, but if such a > construct were describable in OWL then perhaps non-symmetric versions > could also be expressed? This would create the possibility of > describing more complex relationships, like "uncleOf". > > As I said, this is probably naïve, but I've been wondering about > expressing relationships between properties for a while. I'd > appreciate any feedback on the problems this sort of thing would > cause. Paul, Sorry to be self publicising, but I think you will find at least some of the answers in a paper that Ulrike Sattler and I wrote that deals with adding complex role inclusion axioms to the logics that underly OWL [1]. A shorter conference version is available on-line at [2]. The short answer is that extension in this direction is possible, but that care is needed in order to preserve decidability. Ian [1] Ian Horrocks and Ulrike Sattler. Decidability of SHIQ with complex role inclusion axioms. Artificial Intelligence, 160(1-2):79-104, December 2004. [2] http://www.cs.man.ac.uk/~horrocks/Publications/download/2003/ HoSa03a.pdf > > Regards, > Paul Gearon > > >
Received on Saturday, 4 June 2005 18:41:16 UTC