Exploring Distantly-Labeled Rationales in Neural Network Models
This work addresses the challenge of improving model performance with imperfect rationales in classification tasks, representing an incremental advancement in rationale-based neural network methods.
The paper tackles the problem of using distantly-labeled rationales in neural networks by proposing two auxiliary loss functions to better utilize these rationales, focusing on important non-rationale words and reducing redundancy in training. Experiments on two classification tasks show that the methods significantly outperform existing approaches.
Recent studies strive to incorporate various human rationales into neural networks to improve model performance, but few pay attention to the quality of the rationales. Most existing methods distribute their models' focus to distantly-labeled rationale words entirely and equally, while ignoring the potential important non-rationale words and not distinguishing the importance of different rationale words. In this paper, we propose two novel auxiliary loss functions to make better use of distantly-labeled rationales, which encourage models to maintain their focus on important words beyond labeled rationales (PINs) and alleviate redundant training on non-helpful rationales (NoIRs). Experiments on two representative classification tasks show that our proposed methods can push a classification model to effectively learn crucial clues from non-perfect rationales while maintaining the ability to spread its focus to other unlabeled important words, thus significantly outperform existing methods.