• 首页
  • 期刊简介
  • 编委会
  • 投稿指南
  • 收录情况
  • 杂志订阅
  • 联系我们
引用本文:崔建青,仇测皓.基于辅助任务的BERT中文新闻文本分类研究[J].软件工程,2022,25(6):4-8.【点击复制】
【打印本页】   【下载PDF全文】   【查看/发表评论】  【下载PDF阅读器】  
←前一篇|后一篇→ 过刊浏览
分享到: 微信 更多
基于辅助任务的BERT中文新闻文本分类研究
崔建青,仇测皓
(太原工业学院计算机工程系,山西 太原 030008)
initgraph@163.com; 2694266017@qq.com
摘 要: 新闻文本分类是自然语言处理领域中一项重要任务,本文使用新闻标题进行文本分类。随着BERT预训练模型的崛起,BERT模型在多项NLP(Natural Language Processing)任务中都取得了较好的效果,并应用在新闻分类任务中。为了提高新闻分类这一主要任务的效果,本文引入辅助任务判断两个新闻是否是同类新闻,对BERT预训练模型在辅助任务和主要任务上进行微调。在THUCNews数据集上进行实验,实验结果表明,引入辅助任务的BERT新闻分类模型在效果上优于原BERT模型。
关键词: 新闻文本分类;BERT;辅助任务
中图分类号: TP391    文献标识码: A
基金项目: 太原工业学院省级大创项目(S202114101016).
Research on BERT Chinese News Text Classification based on Auxiliary Tasks
CUI Jianqing, QIU Cehao
(Department of Computer Engineering, Taiyuan Institute of Technology, Taiyuan 030008, China)
initgraph@163.com; 2694266017@qq.com
Abstract: News text classification is an important task in the field of natural language processing. This paper proposes to use news headlines for text classification. With the rise of BERT (Bidirectional Encoder Representation from Transformers) pre-training model, BERT model has achieved good results in many NLP (Natural Language Processing) tasks, and it is also applied to news classification tasks. In order to improve the effect of the main task of news classification, an auxiliary task is introduced to judge whether the two news are similar ones, and the BERT pre-training model is fine-tuned in the auxiliary task and the main task separately. Experiments are carried out on THUCNews data set. The experimental results show that BERT news classification model with auxiliary tasks is better than the original BERT model.
Keywords: news text classification; BERT; auxiliary task


版权所有:软件工程杂志社
地址:辽宁省沈阳市浑南区新秀街2号 邮政编码:110179
电话:0411-84767887 传真:0411-84835089 Email:semagazine@neusoft.edu.cn
备案号:辽ICP备17007376号-1
技术支持:北京勤云科技发展有限公司

用微信扫一扫

用微信扫一扫