Masking and purifying inputs for blocking textual adversarial attacks

Huan Zhang, Zhaoquan Gu, Hao Tan, Le Wang, Ziqi Zhu, Yushun Xie, Jianxin Li