A Different Level Text Protection Mechanism With Differential Privacy

A lot of work has been done to address privacy issues Lyu et al. (2020); Anil et al. (2021); Dupuy et al. With the widespread application of differential privacy (2022); Li et al. (2021) to train language models using in text protection, however, the current text differential privacy (DP)Dwork et al. (2006), which is cleaning mechanism based on metric local differential considered the standard for privacy-preserving computing.