Example: Here is what comes out if you feed the Gettysburg address to Doctor Nerve's Markov Page:
Four score and seven years ago our poor power to dedicate -- we here to dedicate a larger sense, we can not perish from the unfinished work which they who fought here have a portion of devotion -- we can not perish from these honored dead we can not hallow -- and dead, who here gave their lives that nation might live. It is rather for the people, by the earth. Four score and dedicated, can not have come to be here have died in vain -- and dedicated here gave the proposition that we can long remember what we can not perish from the great civil war, testing whether that that from these dead we say here, but it can not dedicate a new nation, conceived in vain -- we can not consecrate -- that war. We are met on this continent, a new birth of devotion -- we say here, have consecrated it, far so nobly advanced. It is for us to the last full measure of the living, rather, to that field, as a great task remaining before us the unfinished work which they who fought here highly resolve that this ground. The world will little note, nor long remember what we can never forget what they did here. It is rather for us the people, by the earth. Four score and seven years ago our fathers brought forth on this nation, conceived and dead, who fought here gave the last full measure of devotion to the great battle-field of devotion -- we here dedicated to the people, by the living, rather, to the earth. Four score and seven years ago our poor power to add or any nation so nobly advanced. It is altogether fitting and dead, who fought here dedicated to be dedicated here highly
The idea is to defeat spam filters that reject content that is not mostly “normal text”. Markov-chain generated text looks meaningful to a simple spam filter. To tell the difference between Markov and real text takes natural language understanding. That’s still a hard problem.
Useful post, thank you for teaching me something new today.