this program expects your directory to have the following files that stores
the lexicons into the following syntactic categories:
  "lexC"  complementizers
  "lexD"  determiners
  "lexJ"  adjectives
  "lexN"  nouns
  "lexP"  prepositions
  "lexV"  verbs
  "lexX"  unknowns
  "lexY"  adverbs

you should also have run your emails (or whatever text) through a statistical
package that builds bigrams and trigrams. an example is the CMU_Stat_Toolkit.
if you have done this and generated .arpa files, you can use the following to
"clean up" the formatting:

  cleanGrams.pl - splits given arpa file into a file containing only 
                  unigrams, bigrams, and trigrams
                - input: filename
                - output: filename-1, filename-2, filename-3
  then run unix "sort -r" (in reverse order) on the individual ngram files

now there are two programs. one generates phrases according to the trained
language model (genbo.pl) and the other just generates them randomly
(ranbo.pl).

  ranbo.pl      - generates random words taken from lexicon
                - input: POS
                - output: sentence

  genbo.pl      - takes the sorted trigrams and generates text according to the
                  most probable words in the trigrams (matching the lexicons)
                - input: POS
                - output: sentence