- From: Alan Ruttenberg <alanruttenberg@gmail.com>
- Date: Tue, 29 Apr 2008 18:25:09 -0400
- To: Ian Horrocks <ian.horrocks@comlab.ox.ac.uk>
- Cc: OWL Working Group WG <public-owl-wg@w3.org>, Rees Jonathan <jar@creativecommons.org>, Chris Mungall <cjm@fruitfly.org>
On Apr 28, 2008, at 11:07 AM, Ian Horrocks wrote: > Last week it was agreed in principle that we should add these > properties, but we need to find suitable names for them. > Suggestions in an email please! > > Ian Here's some responses from a query I sent out. Jonathan's tickle my fancy the most. UR or universallyRelatedTo works too. Not sure what the bottom equiv for this is. NR? for notRelatedTo? I can see the merit of the proposal to leave them ungrammatical as a warning, as Chris suggests, though that wouldn't be my preference. -Alan From Jonathan Rees: > Just thinking... > > Start with any universal unary predicate and put "co" in the front. > > E.g. if everything exists (sorry, I don't know whether it does), > then A coexistsWith Y > > Along the same lines: > > A isThingWith B > > -Jonathan Responding, Chris Mungall > This to me implies some kind of temporal overlap; I guess OWL would > want to try and be more neutral here. > > E.g. > > assassinationOfJFK UR cambrianExplosion > > is presumably OK w.r.t the intended semantics of UR (since > 'happens_after' would be an owl:subPropertyOf of this) > > thingProperty is horrible; topProperty is at least consistent with > DL terminology (and probably with 'bottomProperty', for which I > can't think of a remotely sensible name). Ah, but I guess owl uses > thing/nothing rather than top/bottom. > > But I think there is an asymmetry here contraindicating use of the > same terminology. ABox is to TBox as RBox is to...? The actual > relation-instances (facts in owl) as such go in the ABox. > > If it's necessary for the formalism to name these properties, then > I'm happy with UniversalProperty or TopProperty as these names give > a clue that these are mathematical abstractions and need not > necessarily have ontological correlates. I don't like relatedTo for > this reason - colloquially at least, many things are unrelated. > > Do they need named at all? Can they not just exist in the abstract > syntax in the spec? I guess the use case here is that some > reasoners may infer that a relation declared in an RBox use role > chains cannot possibly exist, and would place it as a subProperty > of NothingProperty. Hmm. Jonathan, again: > Personally I prefer names for which the naive reading is both > grammatical and correct. Thus it is best for "properties" if the > name is something that you can put a singular noun phrase before > and after. > > I'm reminded of "The flowers that bloom in the spring / Have > nothing to do with the case". > > A hasSomethingOrNothingToDoWith B > > A doesOrDoesNotRelateTo B > > by construction neither of these can be false (assuming excluded > middle)... then the bottom would have to be > > A doesAndDoesNotRelateTo B > > Very sorry... I'll keep working on it... > -Jonathan
Received on Tuesday, 29 April 2008 22:25:48 UTC