SUO: RE: RE: SUO Ballot with 2 Questions
Thanks Matthew. My comments are below:
<snip>
Eric > > > > Circumstances have prevented me from following this
> > > > discussion very
> > > > closely, however. Did I miss the part where you make a
> > > > compelling case
> > > > for why we wouldn't want to use the 4D subset of OpenCyc or
> > > > of SUMO as a
> > > > 4D starter? I verified for myself that OpenCyc supports your
> > > > definition
> > > > of 4D, but I haven't for SUMO. I believe Adam said that it
does.
> > >
> > > MW: There are 3 key points.
> > >
> > > 1. I am not proposing that the LIS should be the only
> > starting point,
> > > just a framework. I also have in mind various work from the
> > literature
> > > to add into the pot in the axioms line - but in specific areas.
> > >
> > > 2. It is the only starter material that is ONLY 4D.
> >
> > [ELP] It sounds rather trivial to me to strip out the non-4D
> > axioms out
> > of OpenCyc.
>
> MW: Well there are no axioms in OpenCyc beyond structural ones,
[ELP] Right, I was using the term "axiom" a little broadly.
> but reading between the lines of Pierluigi's comments, I suspect
> most of it is 3D.
[ELP] If you will look at the OpenCyc axioms, you will discover that the
argument types of "timeSlices" are both of type "TemporalThing".
The predicate "timeSlices" inherits from "temporallySubsumes" and
"parts". So I'm seeing the merenomic temporal/spatial "regions" to
which you were alluding earlier.
Then please note the "TemporalThing" is right above "SomethingExisting"
and "SpatialThing-Localized" in the type hierarchy.
Is there something else it needs to be truly 4D throughout? Until
Pierluige weighs in here, we are dealing with hearsay wrt his claims of
incompleteness.
> We will see though I'm sure. 3D and 4D (should)
> separate out in the top few layers of a combined ontology, so if
> we could separate out those parts of OpenCYC it should be quite
> easy to do. However, the OpenCYC dataset contains some 5000 concepts
> which in Word works out at 2000 pages - a lot to trawl through.
[ELP] If you can show hoards of OpenCyc concept that you consider to be
4D that doesn't inherit from TemporalThing, I will grant you your point.
I'm concluding, perhaps naively, that 4D-ness can be backfilled easily
into any reasonable 3D or 3D + time ontological or inheritance-based
data model.
<snip>
Best
-Eric