How does that work then?

(Bazongaloid, Fri 10 Jul 2009, 16:02, 1 reply, 17 years ago)

You want the long boring answer?
Yeah why not, can't be arsed to do any more actual work this week :-)

It scrapes board, talk, ot history pages and grabs the name, time and text from each post.

Then it feeds it into a db, recording the probability of one word following another, and also records owners, first words etc

The output bit writes sentences by picking a random owner, then a start word, then trying to follow it by picking random numbers and the corresponding word... so words commonly used together will often follow each other, but the sentence itself skews around in a nicely trippy way.

When you refresh home, it appends a random number of new twunts and displays what it's got.

Apart from the fluff for searches, voting etc, that's about it really :-)

Strangely, the people who post most end up sounding most brain dead, and the people who mainly lurk end up with their twunts being almost their original text.
(MrOli is ugly, but in the morning you will be sober, Fri 10 Jul 2009, 16:25, Reply)

If you look closely it doesn't work terribly well.
Though Dave Trouser's entry is hilarious
(Bazongaloid, Fri 10 Jul 2009, 16:36, Reply)

Yeah,
might be keeping a bit too much text, which makes it write gibberish and fewer funnies. There's a fine balance. I'll trim it out a bit.
(MrOli is ugly, but in the morning you will be sober, Fri 10 Jul 2009, 16:44, Reply)