One of my upper-division English professors gave us an assignment this week where she asked us to write a paper imitating the style/form of Defoe’s Robinson Crusoe (if this sounds like high school to you, believe me, I’m right there with you).
I finally sat down to begin writing the paper tonight and after twenty mind-numbingly-time-wastingly frustrating minutes of trying to mimic the way Defoe’s writing sounds (boring, frankly) a thought ocurred to me.
The way a text sounds/feels to the reader is essentially an ordering problem — once you’ve got a grip on the lexicon, it’s a matter of recreating the author’s textual tendencies. What words tend to follow others? How often? What’s the distribution of the punctuation like? It’s more a math problem than anything else.
I found the full text of the novel online and dumped it into a synthesizer that runs through the text counting all the adjacencies, and used it to instantly spit out a few thousand words of nonsense text in the ‘style’ of Defoe. The semantics are all jumbled until you go through and swap out the nouns to make some logical connections, but the point is that it ‘feels’ exactly like Crusoe reads.
(This is based on treating the text as a Markov chain — incidentally, spammers use Markov to generate spam messages that sound more ‘human.’)
Beer time.