A Closer Look into the Robustness of Neural Dependency Parsers Using Better Adversarial Examples

33975

Knowledge Triples