SUO: RE: Starter Ontology Version 2
Does anyone besides Mike Uschold and myself think that we are being wildly optimistic by proposing to include several thousand concepts (each formally defined) in the initial SUO document? We could easily spend a month discussing each concept, judging from the e-mail so far.
I don't disagree that a generally useful upper-level ontology will ultimately have to contain several thousand concepts (classes and relations), but if we want to be successful in getting this out the door, I think we need to chop it up into much smaller pieces.
I propose that we divide up the SUO into numbered parts, and for our first release, just concentrate on getting Part 1 completed. I would suggest that Part 1 might include roughly our equivalent of what's in Cyc's first two Web pages, titled Fundamentals and Top Level. This would include our equivalents for the following concepts:
[Fundamentals:]
Thing, Individual, Collection, Predicate, isa, genls, disjointWith, covering, partitionedInto, Relationship, arity, genlPreds, genlInverse, resultIsa, arg1Isa, ..., [Logical Connectives], [Default Reasoning], and [Attributes], to name the highlights.
[Top Level:]
Situation, StaticSituation, Configuration, SituationType, Role, Intangible, IntangibleIndividual, ObjectType, StuffType, TemporalObjectType, TemporalStuffType, ExistingObjectType, ExistingStuffType, TemporalThing, SomethingExisting, PartiallyTangible, PartiallyIntangible, IntangibleExistingThing, TangibleThing, CompositeTangibleAndIntangibleObject, Event, subEvents, timeSlices, ..., subAbstrac, postEvents, postConfiguration, .., PhysicalEvent, CreationEvent, DestructionEvent, TransformationEvent, IntrinsicStateChangeEvent, GeneralizedTransfer, ..., MovementEvent, InformationTransferEvent, ..., Action, PurposefulAction, ..., SocialOccurrence, HumanActivity, MentalObject, ..., ScriptType, granuleOfTime, and granuleOfStuff, to name the highlights.
Even if we try to limit the scope to this set of concepts, the problem will be that more and more concepts might need to be sucked into the set when needed for the defining axioms, so it could all explode. Can we draw a boundary around some set of concepts like this without needing any outside concepts for the definitions? I hope so.
By the way, I prefer hyphens between the words of our terms, instead of these internal capitals that Cyc uses, so that spell checkers can deal with them, and word wrapping would go more smoothly.
John A. Thompson
Mathematics & Computing Technology
The Boeing Company