Towards Interpretable Natural Language Understanding with Explanations as Latent Variables

Generating natural language explanations has shown to be effective in not only offering interpretable explanations but also providing additional information and supervision for prediction. However, existing approaches usually require a large set of human annotated explanations for training while collecting a large set of explanations is not only time consuming but also expensive. In this paper, we develop a general framework for interpretable natural language understanding that requires only a small set of human annotated explanations for training. Our framework treats natural language explanations as latent variables that model the underlying reasoning process of a neural model. To optimize it, we develop a variational EM framework where an explanation generation module and an explanationaugmented prediction module are alternatively optimized and mutually enhance each other. Moreover, we further propose an explanation-based self-training method under this framework for semi-supervised learning. It alternates between assigning pseudo-labels to unlabeled data and generating new explanations to iteratively improve each other. Experiments on two natural language understanding tasks demonstrate that our framework can not only make effective predictions in both supervised and semi-supervised settings, but also generate good natural language explanations. 2 * Equal contribution, with order determined by rolling a dice. Work was done during internship at Mila. 2 Code is available at https://github.com/JamesHujy/ELV.git 34th Conference on Neural Information Processing Systems (NeurIPS 2020), Vancouver, Canada.