And he shall lay them unto the replacement text buffer.


We all know about Markov chains, right? You can "analyze" a chunk of text in the following rather mindless way:

  1. Make a table containing one instance of each word found in the text;

  2. For each entry in the table, make a list of each other word which follows that word, with a count for each one: If "bar" follows "foo" twice and "baz" follows "foo" once, "bar" would have an entry something like this:

    foo: bar 2, baz 1;

Once you've got that, you can pick an arbitrary word from the list, and then pick a random "next word" from the list of words that might follow the first word. You can weight that "random" choice by the frequencies you've got: "Foo" would have a 67% chance of being followed by "bar", and a 33% chance of being followed by "baz".

Those are your two first words. Repeat the process with the list of "next words" belonging to the second word, and the third, and so on ad infinitum -- or ad nauseam, more likely.

You could just be rigidly deterministic, and always choose the most frequent "next word", but there's a problem with that: In that case, any given word will always be followed by the same "next word". The first time any word recurs ("the", for example), you'll loop. If "the" is always followed by "dog", and "dog" is always followed by "eats", and "eats" is always followed by "the" . . . You see what I mean? When you hit "the", you'll get "dog" again, and so on, around and around: "the dog eats the dog eats the dog eats . . .", forever. And "forever is a mighty long time", as the man said. Markov chains have horribly limited entertainment value to begin with 1, and that'd make it even worse.


Well, anyway. I happened at one time to have just such a Markov chain text garbler around the house, so I fed it the man page for the Unix sed utility, mixed in with some choice portions of the books of Leviticus and Revelation from the King James Version of the Bible (as kindly provided by Project Gutenberg). After some minor cleanup, the results were illuminating to a remarkable degree, and are hereunto appended.


1:1 And I remember; and neither any more than the tithe of atonement shall therefore put him as an ear let all nonoverlapping instances of it; that maketh a solemn assembly; and worship the heave offering before him an earthquake. Such substrings are departed from thee: Thanks O Lord.

1:2 Replace each line reads after wizards to fulfil his angel which keep those miracles which shall sell unto a man; and purple and no label. This simplifies entering DOS style paths. Start the maximum value him unto him even that pertain unto him? Amen the actual editing of the brimstone.

1:3 Who is shaken of iron; and blemishes be displayed during compilation BUGS I abhor them, saying Command was before him that were breastplates as described below they tread under the street of burnt offering: The oblation unto thee thanks O Lord shall add a year after him? He washed the memorial of either a redemption for an opening command line for him for no guile: For one as described below.

1:4 Amen the rightmost character as honey in one manner of addresses permitted for him lay their iniquity in his angels standing on an hin shall dip his father's daughter in order before thee. An ephah of commands may he will draw out in birth and nation; And let him are departed from heaven great city for ever from even these unto an embedded newline to ed for thee an angel when any number.

1:5 Substitute for a number first line edits each row that do my statutes. Replace each function is selected for an address allowed as snow; and did belong. Apply the sword. 1 through 512 is upon whose head then their iniquity in ward that hear my feasts of commands may appear on a number.

1:6 An -f argument could not have been made: There is a sweet savour. And the compilation phase of the flour unleavened, mingled with the legs with the star above the ' is not allowed as the end of the altar: It at the string except the clearing of the D command also suppresses the priest shall be through a replacement text buffer overflow the next input processing.

1:7 Write the fire, of the following provisos: If there are used as shown in the priest shall pour oil upon the side of the character in the string. The listed in an offering be the burnt sacrifice of sed on a copy of the "quoted character" as explained in the command line of young pigeons. And thou bring his handful of 'rfile'. Place them on the priests, Aaron's sons the Sun computer.

1:8 P is ambiguous at most recent read of group } and he shall burn the altar: It shall bring it unto the same on standard output. That is not terminated properly sed: Garbled command line that function makes that the sacrifice of the buffer called a label for ye shall lay frankincense thereon. For example the liver, with the line number of input line of fowls, then he wash the 'replacement' for a 'set definition' or address selects the hold space.

1:9 Regular expressions can not be understandable by the kidneys, it upon the fat that is a "word" and "{m}" matches whatever the left. The -e and the herd, and kill it unto the standard output. On the g is made by fire, even when any number in an offering be immediately preceding a file name. It is equivalent to or a sacrifice of the herd, let him out of output files is made by deleting the priest shall bring it in the blood round about.

1:10 sed: Cannot open 'file' on last of functions, the firstfruits, ye shall wash the current text. Change lines not divide it of a thing most holy of fine flour with all on either the oil thereof, and cast it upon the last of his inwards. At the options on s command is on the script to nine such substrings in 'labels' and run cumulatively across files, or more than the current line of the is an 'RE', and cut it beside the altar: But his pieces. And he shall lay them unto the replacement text buffer. And he shall take thereout his head, and t [label] branch to nine such substrings are executed in BSD.



1 On the other hand, I did once run some Microsoft marketing copy through a Markov chain garbler and post it on Slashdot. The first five responses thought it was for real. I am not making this up.