Few-Shot Question Answering by Pretraining Span Selection

TL;DR

This paper introduces a span selection pretraining method for question answering that significantly improves few-shot performance, achieving high accuracy with limited training data.

Contribution

It proposes a novel span selection pretraining scheme tailored for question answering, enhancing few-shot learning capabilities.

Findings

Achieves 72.7 F1 on SQuAD with only 128 training examples.

Outperforms standard models in few-shot settings.

Maintains competitive performance in high-resource scenarios.

Abstract

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.