Application of Data Encryption in Chinese Named Entity Recognition

Jikun Dong, Kaifang Long, Hui Yu, Weizhi Xu

Abstract: Recently, with the continuous development of deep learning, there has been a significant improvement in the performance of named entity recognition tasks. However, privacy and confidentiality concerns in specific fields, such as biomedical and military, limit the availability of data for training deep neural networks. To address the issues of data leakage and the disclosure of sensitive data in these domains, we propose an encryption learning framework. For the first time, we employ multiple encryption algorithms to encrypt the training data in the named entity recognition task, training the deep neural network with the encrypted data. Our experiments, conducted on six Chinese datasets, including three self-constructed datasets, demonstrate that the encryption method achieves satisfactory results. In fact, the performance of some models trained with encrypted data even surpasses that of the unencrypted method, highlighting the effectiveness of the introduced encryption method and partially resolving the problem of data leakage.