Computational designs that apply DEEP LEARNING methods on extremely large data sets as a basis for predicting and constructing natural-sounding text.