Token-level and sequence-level loss smoothing for RNN language models

Jul 15, 2018 • Maha Elbayad, Laurent Besacier, Jakob Verbeek