The task of rationalization aims to extract pieces of input text as rationales to justify neural network predictions on text classification tasks. By definition, rationales represent key text pieces used for prediction and thus should have similar classification feature distribution compared to the original input text. However, previous methods mainly focused on maximizing the mutual information between rationales and labels while neglecting the relationship between rationales and input text. To address this issue, we propose a novel rationalization method that matches the distributions of rationales and input text in both the feature space and output space. Empirically, the proposed distribution matching approach consistently outperforms previous methods by a large margin. Our data and code are available.

翻译：合理化的任务旨在提取一些输入文本,作为有理由对文本分类任务作出神经网络预测的理由,根据定义,理由代表用于预测的关键文本,因此,与原始输入文本相比,分类特点分布应当相似,然而,以往的方法主要侧重于最大限度地扩大理由与标签之间的相互信息,同时忽视理由与输入文本之间的关系。为解决这一问题,我们提议一种新的合理化方法,在特性空间和产出空间中,与理由与输入文本的分布相匹配。很典型的是,拟议的分配匹配方法始终以大差幅比以前的方法更优。我们的数据和代码是可用的。