Markov chainer working. Lovely.
Now it's just a matter of grafting the Engrishification onto this in some fashion; need to sort out where the punctuation modifications occur, send them all down to the last tier of changes, and move full-on word modification to the highest tier.
(This seems to make sense for now: Highest-level abstractions first, then the next-smallest things like letter changes and mispellings, then the lowest-level punctuational stuff.)
