Whatcha lookin' at? DeepLIFTing BERT's Attention in Question Answering
This provides interpretability insights for NLP researchers working with transformer models, though it is incremental as it applies existing methods to analyze BERT.
The paper investigates why BERT achieves superior question-answering results by applying DeepLIFT to analyze attention shifts and clustering patterns to compare with human reasoning.
There has been great success recently in tackling challenging NLP tasks by neural networks which have been pre-trained and fine-tuned on large amounts of task data. In this paper, we investigate one such model, BERT for question-answering, with the aim to analyze why it is able to achieve significantly better results than other models. We run DeepLIFT on the model predictions and test the outcomes to monitor shift in the attention values for input. We also cluster the results to analyze any possible patterns similar to human reasoning depending on the kind of input paragraph and question the model is trying to answer.