Distribution Matching for Rationalization

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 available1.

Authors

• Yongfeng Huang The Chinese University of Hong Kong
• Yujun Chen School of Computer Science and Technology, East China Normal University Chongqing Institute, East China Normal University
• Yulun Du
• Zhilin Yang

Keywords

No keywords are indexed for this paper.