摘要
在水利领域中,传统的命名实体识别方法存在有效性差、精度不高、无法解决一词多义和缺乏水利领域语料等问题。基于此,利用BERT语言模型对自建水利文本语料进行训练,并引入FreeLB对抗训练模型增强模型的泛化能力,最后通过条件随机场(CRF)来得到水利实体识别方法。实验结果表明,相对于其他模型,本文提出的FreeLB-BERT-CRF模型对水利领域文本实体识别效果更好。
In the field of water conservancy,the traditional named entity recognition methods are poor in effectiveness,low in accuracy,and unable to solve the problems of polysemous words.This paper uses the BERT language model to train the water conservancy text corpus,and introduces FreeLB confrontation training model to enhance the generalization ability of the model,and finally obtains the water conservancy entity category through the conditional random field(CRF).The experimental results show that,compared with other models,the FreeLB-BERT-CRF model proposed in this paper has a better effect on text entity recognition in the field of water conservancy.
作者
顾乾晖
徐力晨
涂振宇
黄逸翠
GU Qianhui;XU Lichen;TU Zhenyu;HUANG Yicui(School of Information Engineering,Nanchang Institute of Technology,Nanchang 330099,China)
出处
《南昌工程学院学报》
CAS
2022年第3期29-34,共6页
Journal of Nanchang Institute of Technology
基金
江西省水利厅科技项目(KT201639)
江西省科技厅重点研发项目(20151BBE50077)。
关键词
命名实体识别
BERT
CRF
对抗训练
水利信息化
named entity recognition
BERT
CRF
confrontation training
water conservancy informatization