TOWARDS AN AUTOMATIC TURING TEST: LEARNING TO EVALUATE DIALOGUE RESPONSES - View it on GitHub