fast-ish-markov

:car: a decently fast and synchronous markov for when u just need some garbage text

MARKOV TIPS

• cleanup yr own dang lines yrself! lowercase, splitting-on/removing line-ending punctuation, etc.
• best effects gained if u "rate" the output somehow. i.e, generate a thousand lines, run them through some algorithm that returns a "score"
  • u know like, maybe u keep generating markov lines OVER AND OVER until u find ones that rhyme?
  • or maybe u do POS analysis on a known text and generate markov lines till u have a match?
  • get wild!
• a wordfilter is also recommeneded. dariusk/wordfilter, jimkang/iscool, or coleww/this-is-probably-ok-to-say are all good choices.
• another neat trick is to check the markov output against the source lines with a levehnstein distance thing, to make sure it's not just repeating "real text"
• hey! since u already have this garbled text, why not run it through further text processors? word-vomit, poetic-vomit, diacriticize, new-slang, to-unicode, etc.
• make multiple chains and have them dip into each other. if u provide a large num, the chain will run till it hits a stop, at which point u slice off the last n words and try another chain. Mixing a ton of corpuses isn't always advisable, but having several chains that each mix a few corpii and each flow into one another might be more interestinger.
• use markovs as a component in a template, i.e, only generate short bursts of words based on keys for a template. "and they said '{generate some stuff}' and so on and so on etc."
• only seed the markov with lines of a known format, i.e, "I WISH I WAS A _____________", then u can fill based on "I WISH" and let the pseudo-magic flourish

EXAMPLE

var markov = require('fast-ish-markov')
var m = markov(["i like cats", "i like dogs"], 2)
m.fill(m.pick(), 3)
// will be either "i like cats" or "i like dogs". yay probability!