While existing approaches for question answering (QA) attain state-of-the-art (SOTA) performance when trained with large amounts of data, the ability to learn in a sample-efficient manner is a necessity in data-limited domains. I this talk, I will introduce our recent research on automatic question generation (QG) and question answering. Specifically, creating a high-quality question answering dataset for different domains is an extremely time-consuming and laborious task. To solve this issue, we propose to reduce the labeling cost and improve learning efficiency by automatically generating large-scale, high-quality QA pairs, evaluating the quality of the generated data and the ability of the trained model, and designing efficient strategies to collaboratively improving correlated tasks, such as question answering and question generation.

Recording: https://drive.google.com/file/d/1s6YHEkxBtfXerdH0McVCaXtK2PTZtnQr/view?usp=sharing