[SCL] Comments on Pat's SCL document

[John F. Sowa]

Pat,

I'm working on a CGIF definition with two parts:

  1. EBNF definition of the CGIF syntax.

  2. CGIF <-> SCL correspondences to define
     the semantics.

A version of Part I is attached (but I plan to make
further revisions and extensions) later this week,
so don't consider this one definitive.  The main
addition will be a representation of seqvars.
I am adopting the simplest and quickest way of
adding seqvars to CGIF:  copy the KIF syntax.

Some comments on your SCL document:

  1. The definition of "ontology" as any set of SCL
     sentences bothers me.  More precisely, I don't
     care if an ontology is defined as a set of
     sentences, but I very much object to saying
     that any set of sentences is an ontology --
     because one might want to say that a particular
     set of sentences is a theory, hypothesis,
     specification, declaration, or a set of
     assertions, assumptions, constraints, rules,
     judgments, ....

  2. I suggest that we adopt a more neutral term for a
     set of sentences.  If nobody can think of a better
     one-word synonym, I recommend "set of sentences"
     with the acronym SoS (or "collection of sentences"
     -- CoS).

  3. As you have observed, there are strong partisans
     for both typeless and typed languages.  You prefer
     typeless, and I prefer typed.  But we have compromised
     on a typeless core with restricted quantifiers as an
     option and with strongly typed languages supported as
     subsets.  To emphasize that point, I suggest that we
     show both options in our examples -- if we don't,
     there are going to be people who will complain.

  4. For every example in the document, I suggest that we
     show (1) an English statement, (2) a statement using
     restricted quantifiers, and (3) a typeless statement.
     The first example would therefore be written:

     English:  "Every boy kissed some girl."

     Typed:    (Ax:Boy)(Ey:Girl)kissed(x,y)

     Typeless: (Ax)(Boy(x) -> (Ey)(Girl(y) & Kissed(x,y)))

  5. For the CG example, the version you gave is correct,
     but it is possible to have shorter statements, both for
     the typed and the typeless versions:

     Typed:    (Kissed [Boy @every] [Girl])

     Typeless: [If: (Boy [*x]) [Then: (Girl [*y]) (Kissed ?x ?y)]]

     The if-then construction in CGs is actually defined as
     a nest of two negations ~[ ... ~[ ... ]], therefore, you
     can use existential quantifiers in both the antecedent
     and consequent -- exactly as in English, "If there is
     a boy x, then there is a girl y and x kissed y."

  6. I doubt that we should include URIs in the SCL core.
     Instead, I suggest that we have a recommended ontology
     with a dyadic relation URI, which would state that the
     SCL term in the first argument denotes the same entity
     specified by the URI label in the second argument.

  7. As I said many times before, I believe we should adopt
     the Z "toolkit" as a recommended ontology for sets,
     sequences, bags, and integers.  Since that toolkit has
     already been specified as part of an ISO standard, we
     can just refer to that standard by its official name:

     ISO/IEC IS 13568 Information technology — Z formal
     specification notation — Syntax, type system and semantics

     I have a PDF copy of the Final Draft of that standard.
     Send me a note, and I'll send a copy, if you wish.

I have lots of minor comments, but these are enough for
the time being.

On Monday and Tuesday, I'm going to a two-day DARPA review
of the FrameNet project (in Berkeley).  I'll send more
info and comments on Wednesday.

John

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://philebus.tamu.edu/pipermail/scl/attachments/20031207/f0712821/CGIF.htm