基于RoBERTa和RCNN的互联网文字消防接警辅助系统的设计与研究  

Design and Research of Internet Text Fire Alarm Receiving Assistant System Based on RoBERTa and RCNN

作　　者：张昊东 董雷 ZHANG Haodong;DONG Lei(Wuhan Research Institute of Posts and Telecommunications,Wuhan 430000;Wuhan Ligong Guangke Co.,Ltd.,Wuhan 430000)

机构地区：[1]武汉邮电科学研究院,武汉430000 [2]武汉理工光科股份有限公司,武汉430000

出　　处：《计算机与数字工程》2023年第8期1720-1724,共5页Computer & Digital Engineering

摘　　要：对于来自互联网的多渠道消防报警信息,各地接警中心往往由于人力有限等原因,不能很好地使用。针对这个问题,论文提出了基于RoBERTa和RCNN的互联网文字消防接警辅助系统的设计。该系统设计中,论文通过后端逻辑部分的设计为外界服务需求提供了接口。算法模型上,论文利用了业界先进的RoBERTa模型对文本进行了特征抽取,而后对后端分类模型RCNN进行了适配及改进。经实验证明,后端逻辑设计方案切实可用,深度学习短文本分类方案较工业界常用方案及经典方案在准确率上进行了提升,并保证了不错的时间效率。在新时代、新形势下的人民消防工作下,该设计及系统具有相当的实战意义,有助于节约国家财务成本,有助于保障人民群众的生命健康权。With regard to the multi-channel fire alarm information from the Internet,the local police receiving centers are often unable to use them well due to limited manpower and other reasons.In response to this problem,the paper proposes the design of an Internet text fire-fighting alarm support system based on RoBERTa and RCNN.In the system design,the paper provides an interface for external service requirements through the design of the back-end logic part.On the algorithm model,the paper uses the industry's advanced RoBERTa model to extract features from the text,and then adapts and improves the back-end classification model RCNN.Experiments have proved that the back-end logic design scheme is practical and usable.The deep learning short text classification scheme has improved accuracy compared with the common schemes and classic schemes in the industry,and has ensured good time efficiency.Under the people's fire protection work in the new era and new situation,the design and system have considerable practical significance,help save the country's financial costs,and help protect the people's right to life and health.

关 键 词：BERT RCNN 消防接警 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]