Can Question Generation Debias Question Answering Models? A Case Study on Question-Context Lexical Overlap
This addresses a specific bias issue in QA models for reading comprehension, offering an incremental improvement through a simple augmentation technique.
The paper tackles the problem of QA models exploiting dataset biases like question-context lexical overlap, which harms generalization to low-overlap questions, and finds that neural QG models amplify this bias; it proposes a synonym replacement-based data augmentation method that effectively mitigates the degradation with only 70k synthetic examples.
Question answering (QA) models for reading comprehension have been demonstrated to exploit unintended dataset biases such as question-context lexical overlap. This hinders QA models from generalizing to under-represented samples such as questions with low lexical overlap. Question generation (QG), a method for augmenting QA datasets, can be a solution for such performance degradation if QG can properly debias QA datasets. However, we discover that recent neural QG models are biased towards generating questions with high lexical overlap, which can amplify the dataset bias. Moreover, our analysis reveals that data augmentation with these QG models frequently impairs the performance on questions with low lexical overlap, while improving that on questions with high lexical overlap. To address this problem, we use a synonym replacement-based approach to augment questions with low lexical overlap. We demonstrate that the proposed data augmentation approach is simple yet effective to mitigate the degradation problem with only 70k synthetic examples. Our data is publicly available at https://github.com/KazutoshiShinoda/Synonym-Replacement.