deep nlp model

word substitution perturbation (1) randomized perturbation defense (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)