And very interesting, making some considerations.
Grouping words one by one makes nonsense, but we have a text with the same stathistical properties of the original text. Grouping word two by two of three by three is the same, but we don't lost meaning (well, better, we don't lost too much meaning).

Using letters instead of words as units, I think (this is only a conjecture) we could "mimic" the sound of the original language, realizing a new kind of your Acme::Translator :). A kind that could be difficult to obtain even making Acme::Translator having Sex with Transcramble - Random text generator (brrr....).

Even more interesting experiments could be done using music instead of texts. Using something from the MIDI suite we could import a melody and use it to generate other similar (in the sense described by your program) melodies.

Well, just a thought, but could drive to interesting projects (I've always been interested in generative arts).

Quoting some random web source whose URL I didn't bother to note:

There's an example of a travesty generator implemented in Perl in the *first* edition of the manual by Larry Wall and Randal L. Schwartz, on page 326 (sadly, it's been deleted from the 2nd ed).

And here it is:

from the first Programming Perl (back when it was still pink)

#!/usr/bin/perl

# Usage travesty [files]

# First analyse the input.

while(<>) {
   
   # handle article headers and ">>>" quotes.

   next if /^\./;
   next if /^From / .. /^$/;
   next if /^Path: / .. /^$/;
   s/^\W+//;
   
   # Do each word.
  
   push(@ary,split(' '));
   while ($#ary > 1) {
      $a = $p;
      $p = $n;
      $w = shift(@ary);
      $n = $num{$w};
      if ($n eq '') {
         push(@word,$w);
         $n = pack('S',$#word);
         $num{$w} = $n;
      }
      $lookup{$a . $p} .= $n;
   }
}

# Now spew out the words, based on the frequencies.  If there
# is more than one possibility to choose from choose one
# randomly.

for (;;) {
   $n = $lookup{$a . $b};
   ($foo,$n) = each(lookup) if $n eq '';     # A bootstrap.
   $n = substr($n,int(rand(length($n))) & 0177776,2);
   $a = $p;
   $p = $n;
   ($w) = unpack('S',$n);
   $w = $word[$w];


   # See if word fits on line;
   $col += length($w) +1;
   if ($col >= 65) {
      $col = 0;
      print "\n";
   }
   else {
      print ' ';
   }
   print $w;

   # Paragraph every 10 sentences or so.

   if ($w =~ /\.$/) {
      if (rand() < .1) {
         print "\n";
         $col = 80;
      }
   }
}
[download]

p.s. These little gems are the reason I keep that old pink book around.

I tried this using a large text file from Writing of the Early Church Father", at Christian Classics Ethereal Library. The results were pretty amazing! Nice that you kept the "old book" around.

cool.

i did a very similar thing with markov chains on my weblog. whenever someone posts an entry, it updates the frequency count table (it all runs off a postgres db). then, from that, it generates a random post at every time the page is loaded. the idea was that it would be sort of a "hive-mind" of all the people who post there.

when i was writing it i had an idea that i still haven't had time to follow through on, but that your code is probably better suited for. i wanted to create a bot that would go through the discussions on slashdot, and for each thread, pick the highest rated posts, generate a frequency table from that and then post a random message to that discussion. then i wanted to watch how those posts get moderated. i'm pretty sure the markov script could get some pretty good karma.

anders pearson

The best use of it was when some crazy posts started appearing in the discussion group I hung out in. We couldn't figure out at first if this person was totally psycotic or just wrote so badly that nobody could figure out what she was trying to say! After a few attempts to communicate with her didn't work out, I used that program to mimic her writing style and replied to one. It drew a lengthy reply! We seem to have carried on a conversation, though I have no idea what it meant. It was fascenating in that the program included her pet tag-phrases and other affectations, as well as the overall flavor.

—John