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

Re: Response to John Sowa's concerns on KIF, CG, and NCITS/L8




>
>John claims that "although the KIF and CG notations look different, their
>semantics are identical".  Honestly, looking at the KIF and CG *documents*
>(not what might have been intended), I could not convince myself that the
>two documents described identical semantics ... they look substantially
>different.
>
>I continued further with John's illustrations of CGs, many of them use the
>relationship "has".  Unfortunately, the "has" relationship (used many times
>throughout his book) is an ambiguous one ... I started thinking: Why are
>these CGs useful if, as demonstrated by these illustrations, they are
>ambiguous, don't afford machine interpretation, and they can only be
>properly understood by humans?  Why do we need an information technology
>standard for this?
>
>Later on in the book (pages 429-430), there is an illustration of a parse
>tree ... I was hopeful with this illustration because I know that many
>people can agree on semantics in this area.

Greetings. As a newcomer to this list (though not to these issues) I 
wonder if your use of 'semantics' here may differ from the usage 
among logicians. This word has many technical meanings, not all 
compatible, even within computer science. In the logical sense, a 
semantics is provided by a model theory of some kind. Now, 
superficially different syntactic forms may have identical model 
theories, so the way the theory 'looks' is irrelevant; and a model 
theory does not specify the precise meanings of all the nonlogical 
symbols used by a formalism.  Thus a CG (or a set of axioms in KIF, 
or indeed almost any other KR language) may indeed contain symbols 
which are 'ambiguous' in the sense that their exact meaning is not 
determined by the graph (or axioms), yet which have a complete 
semantics in the model-theoretic sense.

Forgive me if this is all obvious.  But if it is not, it might help 
if you could outline (or point us to an outline) of what you mean by 
a notion of "semantics" which would be suitable for standards 
documentation, as your usage seems to differ from that used in the 
logical-semantics community. For example:

> If there are many
>ambiguities or defects, then the document will have significant problems ==>
>there will be little interest approving the document as a standard
>("consensus-building phase", second phase) whether or not the document
>fast-tracked.

I have my own niggles with CG's, but they do seem to be quite 
well-worked out as far as ambiguities and 'defects' are concerned, ie 
they are indeed well-defined and have provable mathematical 
properties, both syntactic and semantic (in the logical sense).  They 
are an alternative notation for logic, in fact, and are just as 
amenable to machine processing as any other KR formalism with a 
similar semantics. So I wonder what 'defects' you are referring to.

Best wishes

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