Text this: Rationalization for explainable NLP: a survey