Word generation using markov chains

words.rb

prefix = (ARGV[0] || 2).to_i
words = (ARGV[1] || 5).to_i

corpus = `cat /usr/share/dict/words`.downcase
dict = Hash.new { |h,k| h[k] = [] }
corpus.split("\n").each do |word|
  chars = "#{word} ".chars.to_a
  until chars.length < (prefix + 1) do
    dict[chars.first(prefix).join] << chars[prefix]
    chars.shift
  end
end

sentence = []
words.times do
  word = ''
  loop do
    word += dict.keys.sample if word.empty?
    word += dict[word.chars.to_a.last(prefix).join].sample
    break if word.chars.to_a.last == ' '
  end
  sentence << word
end

puts sentence.join