[SCL] Re: (SCL): Seqvars revisited

pat hayes phayes at ai.uwf.edu
Thu Dec 19 22:53:12 CST 2002


>On Wed, Dec 18, 2002 at 10:05:50AM -0600, Pat Hayes wrote:
>>  I think that we can get the benefits [of seqvars] without the
>>  unworkable bad side of them if we take a little more trouble and treat
>>  them explicitly as sequences with a fixedpoint semantics a la LISP,
>>  which is after all what the old KIF obviously had in mind, since it
>>  used LISP *as* its syntax. This is what the implementers are asking
>>  for, and its what the RDF/OWL axis has done (in a horribly ugly way,
>>  but leave that aside for now.) I seem to see a convergence here
>>  between the demands of theory and those of practice, which is very
>>  encouraging.
>
>I'd hate to see them go cuz the semantics is so damn cool *sniff* but I
>suspect you are right.

Hey, remember this is SCL, not CL.

>And actually, I think some of the nice features
>of the semantics can be retained (see below).
>
>As for your model theory: Unfortunately, I am in the throes of grading
>graduate student papers and cannot comment at length or with the clarity
>I would otherwise want.  But let me just throw a couple of things out
>there.  First, I wasn't quite clear on your discussion about recursion
>and finitude.  Aren't you simply saying that the domain of an
>interpretation for the language is just the union of a set A of
>individuals and the set A* of all finite sequences over A?

Yes, but there really is a problem with that 'finite' , which isn't 
first-order defineable, and since we are getting kind of deliberately 
sloppy about the syntax/semantics boundary here, I think we have to 
approach the metatheory with slightly more formalist glasses on, so 
to speak.

>  That is,
>aren't we just making the range of the seqvars from the original SKIF
>(simple KIF) model theory first class citizens and welcoming them into
>the domain of the quantifiers?

I think it will come out that way, but if we take it more carefully 
we will be less confused about where exactly we are standing.  And 
maybe more important, I think that a similar route can be used to 
introduce other useful domains apart from the sequences....so there 
might be more  goodies in the jar.

>  Your discussion seems more complicated
>than is necessary.

Well, maybe. If it turns out to be needlessly complex, after all, 
then we can scrap it.

>
>Second, the syntactic problem that arises -- and which you wrestle
>valiantly with -- when you put sequences in the domain is how to manage
>the potential syntactic ambiguities that arise when you have seqvars --
>do you want to be talking about the sequences or their elements?

Proposal: the variables range OVER *syntactic* sequences. The atomic 
assertions USE the sequences.  In this view of the machinery, these 
are essentially substitutional quantifiers. This means that if you 
instantiate with a sequence containing a variable, then that variable 
is *free* , not quantified. So you only get 'schemas' by using these 
- let me call them - list variables. The quantifiers basically say 
'for all schemas...' and 'for some schema....', but I don't expect 
the latter to be of much utility.

Our current use of the universal seq. quantifier now looks like a 
two-stage quantification: it is the normal-universal *closure* of the 
substitution-universal-list quantification of a list variable. Ie it 
says, for all expressions got by substituting a (recursively defined) 
sequence of terms for ?x, the result of universally quantifying that 
expression is true. No wonder its a bitch for a proof theory, right?

But look, we shouldnt try to get this done by email, its late and 
tomorrow I have to go Xmas-travelling. I promise I will get these 
ideas written up and web-accessible by the first week in January, in 
several versions if they come out that way.

>   I have
>not been able to consider your proposal in detail, but let me just
>mention an approach I worked out a while back that appears to be an
>alternative to yours.  The idea is simply to use exactly the current CL
>semantics for sequence variables (modulo pushing sequences directly into
>the domain), so that a seqvar @x is always used to predicate things of
>the *elements* in the value of @x.  To talk about sequences explicitly
>the idea is simply to add a sequence-forming operator "seq" (or, more
>conveniently, perhaps, simple brackets "[...]").  Thus, in particular,
>(seq a b) (or [a b]) forms a term for the sequence <a,b> (seq @x) (or
>[@x]) yields a term explicitly denoting the sequence that @x takes as a
>value.
>
>Back in the days when the SUO list had a reasonable signal to noise
>ratio, I worked out a theory of sequences along these lines.  I include
>it below for what it's worth (perhaps only general disapprobation, but
>even showing what's wrong with the approach is useful).  Three things
>about it.  It is stronger than the theory of sequences you are
>suggesting (I think, from the quick read) as it allows nested sequences.

That is a mixed blessing, though. We need to look at this carefully. 
If sequence quantifiers distinguished subsequences from sequence 
elements they wouldnt work properly.

>Second, because it is convenient to talk about the length of sequences,
>it comes with a a simply theory of the integers, so I include that as
>well.  Finally, There is also a simple theory of classes and relations
>in the background, but I'll leave that out.
>
>-chris
>
>ps: An SCL list has been set up and should be operational later today.
>
>--
>
>  /\ ASCII ribbon | Chris Menzel -- http://philebus.tamu.edu/~cmenzel
>  \/   campaign   | Philosophy Dept, Texas A&M University
>  /\   against    | College Station, TX 77843-4237
>/  \ HTML email  | voice: 979.845-5660  fax: 979.845.0458

(Purely procedural point, but I think we ought to operate as far as 
possible by putting things like this on a website somewhere and 
emailing the uri, rather than by email inclusions, or else the emails 
are going to explode.)

>*****
>
>
>;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
>;;                          SEQUENCES                             ;;
>;;                     (Requires Integers)                        ;;
>;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
<snip>

Pat
-- 
---------------------------------------------------------------------
IHMC					(850)434 8903 or (650)494 3973   home
40 South Alcaniz St.			(850)202 4416   office
Pensacola              			(850)202 4440   fax
FL 32501           				(850)291 0667    cell
phayes at ai.uwf.edu	          http://www.coginst.uwf.edu/~phayes
s.pam at ai.uwf.edu   for spam




More information about the Scl mailing list