基于评价主体识别的网络舆情辅助研判方法实践
Practice of network public opinion assistance analysis method based on evaluation subject identification
宋雪 1李进 1程姣1
作者信息
- 1. 国家计算机网络应急技术处理协调中心山东分中心,山东济南 250002
- 折叠
摘要
[目的/意义]评价主体识别是舆情知识图谱构建和用户画像生成的关键技术,通过评价主体的识别能够提升网络空间访问控制和权限管理能力,以增强对网络安全突发事件的应对能力,为舆情研判、辅助决策、态势分析和信息安全维护,提供有力的技术支持和保障.[方法/过程]为了实现对评价主体的准确、高效识别,促进对网络空间安全的监管.首先,通过BERT-wwm-ext模型将评价文本进行编码后,获取评价文本的向量化表征;然后,通过BiLSTM模型进行特征学习;最终,针对标签之间的依赖关系,通过CRF模型进行评分获得分数最高的标签序列.[结果/结论]经过对比实验,BERT-wwm-ext-BiLSTM-CRF模型在准确率、召回率、F1评价指标上均有显著提升,能够实现对评价文本中主体的精准化识别,为维护网络空间秩序提供辅助手段.
Abstract
[Purpose/Significance]The identification of evaluation subjects is a key technology for constructing public opinion knowledge graphs and generating user profiles.Through the identification of evaluation subjects,it can enhance network space access control and permission management capabilities,enhance the ability to respond to network security emergencies,provide strong technical support and guarantee for public opinion analysis,decision-making assistance,situation analysis,and information security maintenance,etc.[Method/Process]To achieve accurate and efficient identification of evaluation subjects and promote supervision of cyberspace security,this paper proposes an evaluation subject recognition method based on BERT-wwm-ext-BiLSTM-CRF.Firstly,the evaluation text is encoded using the BERT-wwm-ext model to obtain the vectorized representation of the evaluation text.Then,the BiLSTM model is used for feature learning.At the same time,the CRF model is used to evaluate the highest score label sequence for the dependency relationship between labels.[Results/Conclusion]Through comparative experiments,the BERT-wwm-ext-BiLSTM-CRF model used in this paper can achieve better recognition performance in accuracy,recall,and F1 evaluation indicators.It can realize the accurate identification of the subject in the evaluation text,and provide auxiliary means for maintaining the order of cyberspace.
关键词
评价主体识别/辅助研判/特征提取/舆情研判/信息安全Key words
identification of evaluation subjects/auxiliary analysis and judgment/feature extraction/public opinion analysis and judgment/information safety引用本文复制引用
出版年
2024
NETL
NSTL
万方数据