"Semantically Equivalent Adversarial Rules for Debugging NLP models."

Marco Túlio Ribeiro, Sameer Singh, Carlos Guestrin (2018)
a service of Schloss Dagstuhl - Leibniz Center for Informatics