Bill Stewart (bill.stewart@pobox.com)
Fri, 27 Mar 1998 13:30:03 -0800
At 03:42 AM 3/27/98 +0100, nobody at zoom.nu wrote
>Next I want to write a parody generator. Given a base document, this
>extracts the frequence of letter combinations and then uses it to produce
>another document which has the same combinations of letters. Depending
>on the lengths of the letter strings used, the resulting document can
>often look superficially like real text.
Not sure about a source for it, but Peter Wayner's paper on Mimic Functions
describes how to use arbitrary context-free grammars to encode binary data
(easy) and decode it using a parser (a bit harder, but parsers are standard
technology, though I've forgotten most of what I did with them 10 years ago :-)
This will get you a much better parody than letter combinations, and can pass
arbitrary machine-enforced tests for unencryptedness, though a human would
probably catch on quickly enough.
One of his examples generated fake baseball-game reports
<pitchername> winds up. He throws.
It's a curve ball. High and inside. <catchername> throws it back.
He throws it again - fastball. <battername> swings <hard> and misses.
Note that if you want to keep track of which strike it is, or Who's on first,
you either need to write long hairy productions for your grammar,
or else you're out of the context-free ballpark and if I remember right,
you're no longer able to reconstruct cleanly.
Some of the standard Mad Libs or Bureaucratic Buzzword Bingo things
fit well in this environment. Yow! I'm feeding Buzzwords to a Yacc!
So you can proactively leverage your synergies into
<sentence> :: "Let's" <adverb> <verb> <noun-phrase>! ;
<noun-phrase> :: <possessive pronoun> <noun phrase> |
<adjective> <noun phrase> |
<noun> ;
and encode a few bits with each grammatical production as well
as in the choice of nouns, verbs, adverbs, etc.
Thanks!
Bill
Bill Stewart, bill.stewart@pobox.com
PGP Fingerprint D454 E202 CBC8 40BF 3C85 B884 0ABE 4639
The following archive was created by hippie-mail 7.98617-22 on Fri Aug 21 1998 - 17:16:22 ADT