Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

RE: SUO: CHG




>Hi All,
>
>	I agree that the representation of change is very important, and,
>now that I think about it, perhaps some aspects of it could be accomodated
>fairly easily within the merged ontology.  If we add the 'holds' predicate
>as described in, e.g., section 2.2 of Pat Hayes' paper "A Catalog of
>Temporal Theories", we can represent some cases in which
>relations/properties change over time.  For example, to borrow (and modify
>slightly) an example from Pat Hayes' paper, we can represent as follows the
>fact that Joe and Anne were married in 1995 but not in 2000.
>
>(and (holds (married Joe Anne) 1995) (not (holds (married Joe Anne) 2000)))
>
>The 'holds' predicate in conjunction with the temporal axioms already in the
>merged ontology should allow us to represent and inference over many cases
>where an individual alters its properties/relations from one time to
>another.  However, it probably isn't sufficient to cover all cases.  I'll
>investigate the possbility of incorporating something like (part of) the
>situation calculus to cover the remaining cases.  In any case, if no one
>objects, I'll go ahead and add the 'holds' predicate to the merged ontology.

Sorry to be the one, but I object. I mentioned the 'holds' idea in 
that catalog only because it was, indeed, a catalog, but I also gave 
good reasons why it wasnt a very good idea. First, it requires a 
predicate on entire expressions, which isnt legal in KIF. (Or else 
all the relations have to be reconfigured into functions: but 
functions to what?) Second, it isnt clear what it means to quantify 
through a 'holds'. Third, most seriously, it is severely limited in 
expressiveness. It basically allows one to describe changes by 
comparing two snapshots, but that isnt a good enough way to describe 
change, other than very simple changes. It wouldnt be enough to 
represent qualitative physical reasoning, for example.

A better choice would the 'slices of histories' idea, where 3-d 
entities are regarded as timeslices of 4-d histories, and one would 
write
(and (married (@Joe 1995) (@ Anne 1995) (not (married (@ Joe 2000)(@ 
Anne 2000))))
but this is admittedly somewhat cumbersome; so one can abbreviate it 
to a situation-calculus style:
(and (married Joe Anne 1995)(not (married Joe Anne 2000)))

Pat Hayes

---------------------------------------------------------------------
IHMC					(850)434 8903   home
40 South Alcaniz St.			(850)202 4416   office
Pensacola,  FL 32501			(850)202 4440   fax
phayes@ai.uwf.edu 
http://www.coginst.uwf.edu/~phayes