Using a Markov chain to generate text

A Markov chain is a system that predicts future outcomes of a system given current conditions. We can train a Markov chain on a corpus of data to generate new text by following the states.

A graphical representation of a chain is shown in the following figure:

Node E has a 70% probability to end up on node A, and a 30% probability to remain in place

Getting ready

Install the markov-chain library using cabal as follows:

$ cabal install markov-chain

Download a big corpus of text, and name it big.txt. In this recipe, we will be using the text downloaded from http://norvig.com/big.txt.

How to do it…

1. Import the following packages:

   import Data.MarkovChain
   import System.Random (mkStdGen)

2. Train a Markov chain on a big input of text and then run it as follows:

   main = do
   rawText <- readFile "big.txt"
   let g = mkStdGen 100
   putStrLn $ "Character by character: \n"
   putStrLn $ take 100 $ run 3 rawText 0 g
   putStrLn $ "\nWord by word: \n"
   putStrLn $ unwords $ take 100 $ run...