What do Models Learn from Question Answering Datasets?

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, pages 2429–2438, November 16–20, 2020. c⃝2020 Association for Computational Linguistics 2429 What do Models Learn from Question Answering Datasets? Priyanka Sen Amazon Alexa sepriyan@amazon.com Amir Saffari Amazon Alexa amsafari@amazon.com Abstract While models have reached superhuman performance on popular question answering (QA) datasets such as SQuAD, they have yet to outperform humans on the task of question answering itself. In this paper, we investigate if models are learning reading comprehension from QA datasets by evaluating BERT-based models across five datasets. We evaluate mod- els on their generalizability to out-of-domain examples, responses to missing or incorrect data, and ability to handle question variations. We find that no single dataset is robust to all of our experiments and identify shortcomings in both datasets and evaluation methods. Following our analysis, we make recommen- dations for building future QA datasets that better evaluate the task of question answering through reading comprehension. We also release code to convert QA datasets to a shared format for easier experimentation at https: //github.com/amazon-research/ qa-dataset-converter. 1 Introduction Question answering (QA) through reading compre- hension has seen considerable progress in recent years. This progress is in large part due to the re- lease of large language models like BERT (Devlin et al., 2019) and new datasets that have introduced impossible questions (Rajpurkar et al., 2018), big- ger scales (Kwiatkowski et al., 2019), and context (Choi et al., 2018; Reddy et al., 2019) to ques- tion answering. At the time of writing this paper, models have outperformed human baselines on the widely-used SQuAD 1.1 and SQuAD 2.0 datasets, and more challenging datasets such as QuAC have models less than 7 F1 points away. Despite these increases in F1 scores, we are still far from saying question answering is a solved problem. Concerns have been raised over how challeng- ing QA datasets are. Previous work has found that simple heuristics can give good performance on SQuAD 1.1 (Weissenborn et al., 2017), and suc- cessful SQuAD models lack robustness by giving inconsistent answers (Ribeiro et al., 2019) or be- ing vulnerable to adversarial attacks (Jia and Liang, 2017; Wallace et al., 2019). If state-of-the-art mod- els are excelling at test sets but not solving the un- derlying task of reading comprehension, then our test sets are flawed. We need to better understand what models learn from QA datasets. In this work, we ask three questions: (1) Does performance on individual QA datasets generalize to new datasets? (2) Do models need to learn reading comprehen- sion for QA datasets?, and (3) Can QA models handle variations in questions? To answer these questions, we evaluate BERT models trained on five QA datasets using simple generalization and robustness probes. We find that (1) Model performance does not generalize well outside of heuristics like question-context overlaps, (2) Removing or corrupting dataset examples does not always harm model performance, showing that models can rely on simpler methods than reading comprehension to answer questions, and (3) No dataset …

Content truncated to 500 words