Walk a couple of n-gram dictionaries to generate a 'blurt'.

generate.gs

var fs=require('fs');

var opensbuf   = fs.readFileSync(process.argv[2]).toString().split('\n');
var middles = JSON.parse(fs.readFileSync(process.argv[3]).toString());
var closes = JSON.parse(fs.readFileSync(process.argv[4]).toString());

var window_length = 2;

var select_start=function(starts){
  var select_index = Math.floor(Math.random()*opensbuf.length);
  return select_index;
}

var get_key=function(string){
  var i = 0;
  var key = '';

  for (i=string.split(' ').length-(window_length);
       i<string.split(' ').length;
       i++) {
    if (key!='') {key+=' ';}
    key+=string.split(' ')[i];
  }
  return key;
}

var select_follow=function(string, dict){
  var key=get_key(string);
  try {
    var select_index = Math.floor(Math.random()*dict[key].length);
    return select_index;
  }
  catch (e) {throw e;}
}

var completed = false;
while (!completed) {
  try {
    var tweet='';
    var more_middles = Math.floor(Math.random()*4);
    tweet+=opensbuf[select_start(opensbuf)];
    while (more_middles--) {
      tweet+=' '+middles[get_key(tweet)][select_follow(tweet, middles)];
    }
    tweet+=' '+closes[get_key(tweet)][select_follow(tweet, closes)];
    completed = true;
 } catch (e){};
}
console.log(tweet);