Content Generating Algorithms - Where to Start?

[NeoDevin]

I want to learn more about content generating algorithms (examples: here (http://www.abc.net.au/science/news/stories/2007/1842077.htm) and here (http://gizmodo.com/379388/computer-writes-200000-books-man-takes-credit)), but I'm not sure where to begin. I have basic programming knowledge, but I figure I can learn any more advanced programming requirements as I go.

Can someone recommend some reading to start with on the subject?

Thanks in advance.

[MATLABdude]

SCIgen the syntactically correct CS technobabble generator (for those moments when you wonder whether certain authors just randomly strung buzz-words together):
http://pdos.csail.mit.edu/scigen/

[junglebeast]

Does anyone else find it hilarious that the book generating program uses automation and copy-pasting in Word instead of just generating a latex source? Sounds like wayyy more of a pain than it has to be...

[Borek]

Markov chains can be used.

As he spoke, it was not merely possible, but I wouldn't really be true? Could they enter any of the month, which he lived. 'Why shouldn't we be married in the daytime, and now suddenly she slapped it away from it without discussion, as if she were relieved to hear something really interesting to tell them apart.' Well, then, ask Mrs. Hudson has been galloping all night, from eight in the governors began to reassure Aouda, telling her it no more, and, full of the house for information about the arrowheads themselves. I was weak. Yes, they had found no words for this. This is a nurse here who may some day citizens of Alexandria filled him with the sad news they were always very careful, in accordance with your short term survival scheme you have, rather than fall into the present. He paused. Very likely I have. Then, for pity's sake, Help! The way he stood hid behind a desk on the wall of Chasma Borealis and around six thirty a sudden thought, and so he held his sway. These ruffians, united by our master's wife, who had been unduly anxious concerning the destination of the crab's cerebral portion.

I don't remember what was feed into the generator - some combination of SF books probably.

[NeoDevin]

Noone has any suggestions on where to start learning about it?

[DavidSnider]

NeoDevin: Check out http://en.wikipedia.org/wiki/Markov_chains

[NeoDevin]

Thanks DavidSnider, that was an interesting read. I'm looking up more information now, hopefully this will lead me where I want to go. If anyone has a more introductory level and in depth source on markov chains, I would be grateful.

[Borek]

The basic idea behind the text above is:

1. use some original text to calculate probabilities that given word is followed by other word
2. generate text that retains these probabilities (that is - start with any word, draw the next one using probablities calculated in step one, use this word as a base for the next draw and so on)

If I recall correctly text presented was generated using three words sequences, but it is just a simple extension of the idea.

Edit: Hm, Chasma Borealis suggests Red Mars was used as the input :smile: I really don't remember, I wrote the code several years ago.

[DavidSnider]

I've heard rumors that social science papers have been generated this way and accepted into peer reviewed journals =)