MIT CSAIL's TextFooler generates adversarial text to strengthen natural language models

Researchers say they've developed a framework -- TextFooler -- that generates adversarial examples capable of fooling leading natural language models.