Xiang Lin, Simeng Han, Shafiq R. Joty: Straight to the Gradient: Learning to Use Novel Tokens for Neural Text Generation. ICML 2021: 6642-6653