@inproceedings{li-etal-2023-ccl23,
    title = "{CCL}23-Eval 任务1系统报告:基于增量预训练与对抗学习的古籍命名实体识别(System Report for {CCL}23-Eval Task 1:::{G}u{NER} Based on Incremental Pretraining and Adversarial Learning)",
    author = "Li, Jianlong  and
      Yu, Youren  and
      Liu, Xueyang  and
      Zhu, Siwen",
    booktitle = "Proceedings of the 22nd Chinese National Conference on Computational Linguistics (Volume 3: Evaluations)",
    month = aug,
    year = "2023",
    address = "Harbin, China",
    publisher = "Chinese Information Processing Society of China",
    url = "https://aclanthology.org/2023.ccl-3.3",
    pages = "23--33",
    abstract = "{``}古籍命名实体识别是正确分析处理古汉语文本的基础步骤,也是深度挖掘、组织人文知识的重要前提。古汉语信息熵高、艰涩难懂,因此该领域技术研究进展缓慢。针对现有实体识别模型抗干扰能力差、实体边界识别不准确的问题,本文提出使用NEZHA-TCN与全局指针相结合的方式进行古籍命名实体识别。同时构建了一套古文数据集,该数据集包含正史中各种古籍文本,共87M,397,995条文本,用于NEZHA-TCN模型的增量预训练。在模型训练过程中,为了增强模型的抗干扰能力,引入快速梯度法对词嵌入层添加干扰。实验结果表明,本文提出的方法能够有效挖掘潜藏在古籍文本中的实体信息,F1值为95.34{\%}。{''}",
    language = "Chinese",
}