Αποστολή με SMS: On guaranteed optimal robust explanations for NLP models