Noised Consistency Training for Text Summarization

Junnan Liu, Qianren Mao, Bang Liu +3

Neural abstractive summarization methods often require large quantities of labeled training data. However, labeling large amounts of summarization data is often prohibitive due to…