markov-chain's Introduction

markov-chain

This library implements an n level markov chain in Clojure. It can be trained on any sequence of data and detect likeness or even generate it.

Usage

See examples/language.clj for how to use it to detect gibberish.

After training a 3 layer MC on individual chars of Great Expectations, Chapter V:

> (detect-likeness mc mc-depth "I was walking on the road.")
0.20905892931107461

> (detect-likeness mc mc-depth "a;lsjdfj lkeflh ashoe iofj")
0.05896670169096736

> (->> (generate-likeness mc 100) (reduce str ""))
"en to hoseliarefuld some wit'll. Joe an him ingitew and forns anceir
mened preandied try hat ineve a"

It also works for entire words, trained on Great Expectations, the whole book, without punctuation or caps:

> (detect-likeness mc mc-depth (clojure.string/split "i went to the room with them" #"\s+"))
0.04716500678653864

> (detect-likeness mc mc-depth (clojure.string/split "them room the i with to went" #"\s+"))
0.0045216139294960095

> (->> (markov/generate-likeness mc 100) (reduce str ""))
" a juryman in some unlucky hour i resolved a word further without
introducing estella s father would occasionally have some important
bearing on the forehead hardens the brain into a general air and flavor
about the same dim suggestion that i had tried hard at me but he really
is a pretty large experience of him in walking out of england without
laying it upon me i felt any tenderness in her wondering lament of
gracious asked my sister s sudden fancy for me this may lead to trouble
i know i did know it much i ve got you"

A little more training and it'll be ready to generate some high quality spam.