Re: reparsing the ambiguous

[Ilmari Karonen]

On Wed, 11 Jul 2001, David L. Nicol wrote:
> 
> Statistics break at the edges.  I meant something that
> will expand
> 
>   $$$name[5]{cheese}
> 
> into
 [snip]
>   my $RVAL;
>   eval {
>$RVAL = ${${${name}[5]}{cheese}}
>   }; # normal parse: the sixth element 
>  # in @name refers to a hash. Created like
>  #$H{cheese} = 'ducati'; $name[5] = \%H;

You do realize that the actual parsing of $$$name[5]{cheese} is

  ${${${name}}}[5]{cheese}

created like

  $H{cheese} = 'ducati';
  $A[5] = \%H;
  $S = \@A;
  $name = \$S;

right?

I'm not exactly sure what, if anything, this is an argument for.

-- 
Ilmari Karonen - http://www.sci.fi/~iltzu/
"This must be a use of the word 'obvious' of which I was previously unaware."
   -- Charles Martin in rec.arts.sf.science

Re: reparsing the ambiguous

[Dan Brian]

> I wonder how long (less than a year?) it will be until people are writing
> computer languages that know enough about context to select a parsing that
> Makes Sense when faced with an ambiguous construction.

Not long. My Linguana talk/paper @ TPC treats (in part) a natural language
programming language, taking sense ambiguity into account. Not terribly
sophisticated, and requiring gobs of statistical context data to work. But
it knows whether 'open' directs a file, URL, or network connection. It
wouldn't be (too) difficult to introduce context-dependent treatment of
variables. I doubt anyone is interested in downloading several hundred MB
of disambiguation statistics with a language distribution, however. :-)

Dan