Towards Interpreting BERT for Reading Comprehension Based QA
This work provides incremental insights into BERT's internal mechanisms for researchers in NLP interpretability, focusing on a specific domain of reading comprehension QA.
The paper tackled the problem of interpreting how BERT achieves near human-level performance in reading comprehension-based question answering, by analyzing layer roles using Integrated Gradients and observing that initial layers handle query-passage interaction while later layers focus on contextual understanding and answer prediction, with specific insights into quantifier questions.
BERT and its variants have achieved state-of-the-art performance in various NLP tasks. Since then, various works have been proposed to analyze the linguistic information being captured in BERT. However, the current works do not provide an insight into how BERT is able to achieve near human-level performance on the task of Reading Comprehension based Question Answering. In this work, we attempt to interpret BERT for RCQA. Since BERT layers do not have predefined roles, we define a layer's role or functionality using Integrated Gradients. Based on the defined roles, we perform a preliminary analysis across all layers. We observed that the initial layers focus on query-passage interaction, whereas later layers focus more on contextual understanding and enhancing the answer prediction. Specifically for quantifier questions (how much/how many), we notice that BERT focuses on confusing words (i.e., on other numerical quantities in the passage) in the later layers, but still manages to predict the answer correctly. The fine-tuning and analysis scripts will be publicly available at https://github.com/iitmnlp/BERT-Analysis-RCQA .