This is interesting: a Markov chain-based model of Sarah Palin’s interview answers. It uses her existing public statements to construct a computer model of what words/phrases would be likely to follow one another, then uses the model to generate random answers. It’s relatively coherent. Ask a question! Some ideas.