- From: Ian Horrocks <horrocks@cs.man.ac.uk>
- Date: Fri, 25 May 2007 11:02:45 +0100
- To: Matt Williams <matthew.williams@cancer.org.uk>
- Cc: Owl Dev <public-owl-dev@w3.org>, Semantic Web <semantic-web@w3.org>
Hi Matt, It isn't completely clear whether you are asking about using some kind of rule engine to reason with OWL or extending OWL with rules; I will assume that it is the former. It is pretty easy to write some sound inference rules for OWL; what is difficult is guaranteeing completeness and termination. This is made more tricky by that fact that the semantics of rule systems generally assume a closed domain (the only individuals that exist are those that are explicitly mentioned in the ontology), whereas the semantics of OWL allows for the existence of (a possibly infinite number of) additional unnamed individuals -- in fact there exist OWL ontologies for which all models have domains of infinite size. Incompleteness may be a much more serious problem that it at first appears, because failure to derive a positive result is invariably interpreted as a negative result -- which is obviously incorrect in general. There may be applications where this incorrectness is not much of an issue, but there are also many where it is -- see [1] for an example where incomplete reasoning could have led to patients being mis-diagnosed. Moreover, given that several highly efficient and correct reasoners are available, one would presumably need a pretty compelling reason to want to develop/use an incorrect one. There are lots of papers on reasoning with OWL that you can read in order to get an idea of what is needed in order to guarantee correctness: [2] describes a tableau based method, and [3] describes a method based on a (highly non-trivial) reduction to disjunctive datalog rules. Regards, Ian [1] http://owl-workshop.man.ac.uk/acceptedPosition/submission_19.pdf [2] http://www.cs.man.ac.uk/~horrocks/Publications/download/2007/ HoSa07a.pdf [3] http://www.cs.man.ac.uk/~bmotik/publications/papers/hms07query- journal.pdf On 25 May 2007, at 09:54, Matt Williams wrote: > > Dear All, > > I was wondering if anyone can give me some precise pointers as to > why implementing OWL reasoning on rules is so hard? > > There seem to be lots of systems that do subsets of OWL as rules, > but I'm still unclear about what features in OWL don't work when > translated into rules. > > Thanks a lot, > > Matt > -- > http://acl.icnet.uk/~mw > http://adhominem.blogsome.com/ > +44 (0)7834 899570 >
Received on Friday, 25 May 2007 10:03:09 UTC