Apparently, I feel safer in the garage. :)

That is so cool.

Years ago, I wrote one of those to mangle BBS postings. Great fun!

Yes, what was once old is new again. Having livejournal offer freely available, easily parseable text dumps written by millions of different people, sure makes the source text problem easier!

Rock! On!

I wrote programs to do that back in high school, and asked everyone to send me their text files, so I could build up good percentages. I remember there being some pretty amusing results. I also gave it several large purity tests to build percentages from and gave it "Have you ever" as a seed string, and that gave some good results, I think.

I have noticed that it works better when used against a personal journal, rather than against a community. I think this is because a single person's writing style is often consistent across posts, but a community usually has posts written by many different people. The differences in writing style give the markov model less opportunity to choose alternate word chain paths.

Now, this doesn't mean that it's not still fun to run it against communities like badspam. :)

Yeah, that's a really good point. I've noticed it works well on journals of people who tend to use the same sorts of phrases over and over.

Awesome toy, by the way!

If your LJ is in one langauge, with only a few posts or quotes in other languages, you don't get good results when words in those other languages are chosen as the seed words. Because it has very few other words to pick next, so it ends up quoting large sections of posts. I think I only have one entry with french in it, which is lyrics from a song, and the randomizer picked a french word to start with so it quoted the entire song :) My LJ is mostly in Esperanto, and it actually took three or four reloads before I got a text which was in Esperanto.

The text generator is fantastic, I'm going to waste bloody ages playing with it! Well done!

It's great! Thank you so much.

Great toy!

Thank you so much! I am playing with already for two hours! It's hillarious. And some of the stuff that comes out is so creative and often times so smart and deep, I just can't believe it. Thank you!

It's very funny :-)
But do you use only the 20 last journal entries ?

Yes, the RSS feed of your journal from livejournal only includes the most recent entries. See http://www.livejournal.com/support/faqbrowse.bml?faqid=149

Genious!

Absolutely incredible, as it is becoming a next stage of Burroughs' cut-up technique. Thanks!

Nice toy. Really fun. :)
I wrote a thing like that few month before, and I'm interested to take a look at your code - can you show it to me? Please write to sergey(at)consal[dot]ru.

Very fun toy, thank you!
I would like to apply it to some "friends only" journal entries, so need to add some authentication.
Can you mail me code? PleeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeZZZ... :-))
"For personal use only!", of cource! :-)
If "yes" — xav13@atmusic.org

I added the ability to read friends-only posts, see this post to find out how.

The source code is available here.

Thanksthanksthanksthanks!!! It's cool! :-)

Now I'll try to add ability to mix various journals... May be funny... :-)

This is very, very cool! I did a trigram based system using the King James Bible as a corpus for a computational linguistics course. I got some great prophetic proclamations.

A lot of my friends are getting some great results from their journals, and I want to join in the fun... but it isn't working for my RSS feed. It says there's a parse error.

All I can figure is that there's something in my journal that your program doesn't like. My only thought is that I use XJournal, which inserts links to the iTunes Music Store for my currently playing music.

Any ideas why it won't work?

It looks like your guess is right - your RSS feed contains two é and a ö which were not properly UTF-8 encoded. These appear within song names, and I guess that's what XJournal inserted. Since I manually edited the cached copy of your journal, it will work for about the next hour or so until the cached copy expires. I could try to preprocess the RSS data to try to fix this sort of encoding error before the XML parser has a chance to complain. Perhaps I'll run some tests to see how prevalent this kind of encoding error is.

Hmm. Any chance this could be adapted so that it would take rss/atom feeds from locations over than LJ?

Sure, I posted a link to the source code a couple of comments up. It should be easy to modify to pull text from an alternate source.

it just cool!
it is difficuld to make some like this, is it?

Занимательно. Отлично написанно.