She tastes like a comet

[HappyScrappy]

My entertainment of late has been writing a program that will read in all the radiohead lyrics and then form a markov matrix and then randomly write songs based on that.

At first I was using a single letter sampling rate and only The Bends album, but that outputted a bunch of gibberish.
Then I bumped it up to sample four positions at a time, and also started to include songs from Pablo Honey, and it is at least readable now.

Here is a sample of the 4 position scan before I write one for a two and three position scan. From writing these in the past, I'm pretty sure that the 3 pos scan will work the best.

-------------------
When you're strong enough the bar with you home sky high

My uncle Bill
My Belisha beacon
Talking to?
i'm gonna melt down.

Rows of houses all be the trained and pain
A lack stab them in
You how that you home.

the thing would happen
I wish, I wish I can see it's the rules
But stand fade out, it's the black of rubber plant
In the fire,
and I could never chaotic this
Just lying in the fire.
You do it everythings,
To blow whole
A total waste of houses and bawl
You have and sulk, something caught in the times you sulk, somethink off
He's beady gonna melt down on your head shrinkers
They love me dry
Don't leave me dream
Nice dream
Nice dream, nice dream, nice dream, nice dream

If you try at working out

She tastes like a comet
Suck, suck yourself

Don't yourself
Just crumbles and stay awake ching me my girls into position
-------------------

 

[Taps]

What the fuck are you talking about? lol

 

[2 ton hoss]

HAVE YOU NEVER HEARD OF A MARKOV MATRIX? LOL

 

[HappyScrappy]

here is one with a 3 character sampling rate.

hmmm, gonna now try to have it do something like a 3+1+3+1 sort of thing since the straight 3 isn't so great.
first will try a dual.

----------------------------
Disinfected, eating you.
whateverything
My uncle Bill yourself again and no-one else
The hell to yourself again saying uncle
Be a hole
My uncle
And i just as special,

but I'm a world
You me dream

I can breaming to?
i know that something out come help but it's beautiful world
I beautiful weird
You don't leave me, made out working hell nothere

She like their blue heady gone is broken

You and fade ones,
could still stand it wears
The head shring up in crush it belong out again

I don't leave happy,
what recognition,
To blow me a come
Limb the doing part and bar with my girlfriend, was back of feel it on
The hight from wouldn't befor so
-------------------------------

 

[2 ton hoss]

HIGHER SAMPLING RATES ARE ALMOST ALWAYS BETTER.

BUT YOU HAVE TO MAKE SURE YOU HAVE ENOUGH TRAINING INPUT.

 

[HappyScrappy]

yeah, it is an XML file that I'm just manually populating now.
I can add a stripper script later to scrape it off of the page that I'm getting it from now, but there aren't that many songs, so it isn't too bad.
it gets smarter as the data grows.

the problem with too high a sampling rate is that you get too much repetition. I built in a weighted random, which adds to that - and as you say, a larger sampling base should help. that is why I want to add a 3+1+3... or 4+1+4... b/c that way the 1 char spot will have a larger number of possible followers that if it was a 4 to 4 comparison.

I'm off to bed now. got it working enough that it is readable, good enough for tonight.

 

[Pamela]

Good Night!

 

[smallmovesal]

that's pretty cool

 

[slickdadd]

lol. "She tastes like a comet, Suck, Suck yourself" awesome.

 

[WODIN]

Most people your age HS are dropping acid or E...I see where this would be a pointless endevour on your part.

Very solid work.