randomized perturbation defense

word substitution perturbation (1) deep nlp model (1) robustness enhancement (1) nlp model (1) word perturbation (1) word substitution (1) adversarial attack (1) deep learning (1) model robustness (1) adversarial robustness (1)