CCL23-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 CCL23-Eval Task 1:::GuNER Based on Incremental Pretraining and Adversarial Learning)
Jianlong Li (剑龙李,), Youren Yu (于右任), Xueyang Liu (刘雪阳), Siwen Zhu (朱思文)
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%。”- Anthology ID:
- 2023.ccl-3.3
- Volume:
- Proceedings of the 22nd Chinese National Conference on Computational Linguistics (Volume 3: Evaluations)
- Month:
- August
- Year:
- 2023
- Address:
- Harbin, China
- Editors:
- Maosong Sun, Bing Qin, Xipeng Qiu, Jing Jiang, Xianpei Han
- Venue:
- CCL
- SIG:
- Publisher:
- Chinese Information Processing Society of China
- Note:
- Pages:
- 23–33
- Language:
- Chinese
- URL:
- https://aclanthology.org/2023.ccl-3.3
- DOI:
- Cite (ACL):
- Jianlong Li, Youren Yu, Xueyang Liu, and Siwen Zhu. 2023. CCL23-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 CCL23-Eval Task 1:::GuNER Based on Incremental Pretraining and Adversarial Learning). In Proceedings of the 22nd Chinese National Conference on Computational Linguistics (Volume 3: Evaluations), pages 23–33, Harbin, China. Chinese Information Processing Society of China.
- Cite (Informal):
- CCL23-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 CCL23-Eval Task 1:::GuNER Based on Incremental Pretraining and Adversarial Learning) (Li et al., CCL 2023)
- PDF:
- https://preview.aclanthology.org/nschneid-patch-3/2023.ccl-3.3.pdf