model_numerals

var utterances = ["one","two","three","four","five","six","seven","eight","nine","ten"]

var utterancePrior = function() { return uniformDraw(utterances) } 

var states = _.range(1, 11)

var statePrior = function() { return uniformDraw(states) } 

var cost = function(utterance){
  utterance == "null" ? 0 :
  1
}

var nohalos = {
  "one" : 1,
  "two" : 1,
  "three" : 1,
  "four" : 1,
  "five" : 1,
  "six" : 1,
  "seven" : 1,
  "eight" : 1,
  "nine" : 1,
  "ten" : 1
}

var literalMeanings = {
  "one" : function(state) { return state >= 1; },
  "two" : function(state) { return state >= 2; },
  "three" : function(state) { return state >= 3; },
  "four" : function(state) { return state >= 4; },
  "five" : function(state) { return state >= 5; },
  "six" : function(state) { return state >= 6; },
  "seven" : function(state) { return state >= 7; },
  "eight" : function(state) { return state >= 8; },
  "nine" : function(state) { return state >= 9; },
  "ten" : function(state) { return state >= 10; },
}; 

var followsthat = function(exp1, exp2) {
  var f = literalMeanings[exp1]
  var g = literalMeanings[exp2]
  var m = function() {
    var s = statePrior()
    //add the desired assumption:
    condition(f(s))
    return {'follows': g(s)}
  };
  var dist = Infer({}, m)
  return dist 
};

var literalListener = cache(function(utt) {
  return Infer({model: function(){
    var state = statePrior()
    var meaning = literalMeanings[utt]
    condition(meaning(state))
    return state
  }})
})

// pragmatic speaker
var speaker = cache(function(state) {
  return Infer({model: function(){
    var utt = utterancePrior()
    factor(literalListener(utt).score(state)-cost(utt))
    return utt;
  }})
})

var partialTruth = cache(function(utt){
  var meaning = literalMeanings[utt]
  var consistentWorlds = Infer(function(){
    var state = statePrior()
    condition(meaning(state))
    return state
  })
  return Infer({model: function(){
    var uttprime = utterancePrior()
    var state = sample(consistentWorlds)
    condition(Math.exp(followsthat(uttprime, utt).score({"follows": true}))==1)
    condition(Math.exp(followsthat(utt, uttprime).score({"follows": true}))!=1)
    factor(speaker(state).score(uttprime))
    return uttprime;
  }})
})

var impSpeaker = cache(function(state,alpha,beta,halos) { 
  return Infer(function() {
    var utt = utterancePrior()
    var uttprime = utterancePrior()
    var uttprimeMeaning = literalMeanings[uttprime]
    condition(uttprimeMeaning(state))
    var speakerProb = Math.exp(speaker(state).score(utt))
    if(state == 10 | state == 0) {
      factor(Math.log(speakerProb))
    } else {
      var partialTruthProb = halos[utt] * Math.exp(partialTruth(uttprime).score(utt))
      factor(alpha * (Math.log(speakerProb + beta * partialTruthProb)))
    }
    return(utt)
  })    
})

var pragmaticListener = function(utterance,alpha,beta,halos){
  Infer({model: function(){
    var state = statePrior()
    observe(impSpeaker(state,alpha,beta,halos),utterance)
    return state
  }})
}