TY - JOUR
AU - 
AB - MSP: Multi-Stage Prompting for Making Pre-trained Language Models Better Translators 1,3,4 6 1,3,4 1,2,3,4,5 Zhixing Tan , Xiangwen Zhang , Shuo Wang , and Yang Liu Department of Computer Science and Technology, Tsinghua University, Beijing, China Institute for AI Industry Research, Tsinghua University, Beijing, China Institute for Artificial Intelligence, Tsinghua University, Beijing, China Beijing National Research Center for Information Science and Technology International Innovation Center of Tsinghua University, Shanghai, China Kuaishou Tech, Co. Abstract and Knowles, 2017). While neural machine trans- lation (NMT) (Sutskever et al., 2014; Bahdanau Prompting has recently been shown as a promis- et al., 2015; Vaswani et al., 2017) is the current ing approach for applying pre-trained language de facto approach for machine translation, using models to perform downstream tasks. We pre-trained LMs as translators via prompting is ap- present Multi-Stage Prompting, a simple and pealing in several aspects. For example, for the automatic approach for leveraging pre-trained method described in this paper, supporting a new language models to translation tasks. To better mitigate the discrepancy between pre-training translation direction with a pre-trained LM occu- and translation, MSP divides the translation pies disk spaces below 20M, which is much smaller process via pre-trained language models 
TI - MSP: Multi-Stage Prompting for Making Pre-trained Language Models Better Translators
JF - Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
DO - 10.18653/v1/2022.acl-long.424
DA - 2022-01-01
UR - https://www.deepdyve.com/lp/unpaywall/msp-multi-stage-prompting-for-making-pre-trained-language-models-N44uQvfj1T
DP - DeepDyve
ER -