Few-Shot Question Answering by Pretraining Span Selection
This work is significant for researchers and practitioners working on question answering, as it provides a method to achieve strong performance with limited labeled data, addressing a common bottleneck in real-world applications.
This paper addresses the poor performance of standard pretrained models in few-shot question answering (QA) settings, where only a few hundred training examples are available. They propose a new pretraining scheme called recurring span selection, which significantly improves performance, achieving 72.7 F1 on SQuAD with only 128 training examples.
In several question answering benchmarks, pretrained models have reached human parity through fine-tuning on an order of 100,000 annotated questions and answers. We explore the more realistic few-shot setting, where only a few hundred training examples are available, and observe that standard models perform poorly, highlighting the discrepancy between current pretraining objectives and question answering. We propose a new pretraining scheme tailored for question answering: recurring span selection. Given a passage with multiple sets of recurring spans, we mask in each set all recurring spans but one, and ask the model to select the correct span in the passage for each masked span. Masked spans are replaced with a special token, viewed as a question representation, that is later used during fine-tuning to select the answer span. The resulting model obtains surprisingly good results on multiple benchmarks (e.g., 72.7 F1 on SQuAD with only 128 training examples), while maintaining competitive performance in the high-resource setting.