Towards Interpretable Natural Language Understanding with Explanations as Latent Variables

NeurIPS 2020