Erik Benson, of All Consuming Fame (sounds rather grand doesn't it?), took part in 2002 and 2003.
He then went on to self-publish his first attempt, and at the book website he had about 9 months of fun running all kinds of interesting stats on the book text as part of the process.
Among the fun had, a random sentence generator, based on the book text. The rules for generating the sentences are simple:
- You supply the starting word
- Each word is followed by a word that also appears after it somewhere in the original text
- The more times a word pair appears in the original text, the more likely it is used to be used in the sentence generator
Yes, mathematical readers, that's a discrete time Markov Chain. I tried seeding the generator with 'Remember' and got:
Remember not only did we discover that a handsome man loved dressing, loved combing his hair, loved preparing the angles and planes of his life, falling apart, trying to do so if necessary.
That's so good its hard to believe Benson isn't cheating.
All of these fun text tools deserve to be made generally available, and I think I'll add implementations here at classy.dk. Watch this space. If you know of a good existing resource (i.e. a reason not to build these tools) post a comment.