Prompt-Based Learning in Natural Language Processing from 2020 to 2022, a Survey

Kai Wang, Jia Du, Baofeng Li, Yu Qin, Jianguo Tang, Yanquan Zhou, Lei Li

Abstract: Prompt-based learning is a new paradigm of natural language processing by adding “prompt information” to input and adjusting downstream tasks to text generation tasks without significantly changing the structure and parameters of the pre-training language model. In this paper, the improvement of prompt-based learning in recent years is systematically sorted out along several dimensions, to display a clear vein for the development and improvement of prompt-based learning that has been done for beginners of prompt-based learning, help and enlighten prompt-based learning work in the future.