摘要
[目的/意义]本文基于情绪唤醒度—效价理论,利用大语言模型构建网络问政文本情感数据集,并提出相应的情感分类方法。[方法/过程]首先,基于情绪唤醒度—效价理论构建8个情绪区间和56个情绪标签的情感体系。其次,利用大语言模型对网络问政文本进行细粒度情感标注,构建高质量的情感分类数据集。最后,在此基础上,本文提出了面向该类情感分类体系的情绪识别模型,并进行了系统评估。[结果/结论]实验结果验证了所构建数据集的质量和情感分类方法的有效性。研究为网络问政领域的细粒度情感分析提供了可复制的数据资源和技术方案。
[Purpose/Significance]This study proposes a method to construct an emotional classification dataset for online government-public interaction texts and develop corresponding sentiment classification approaches based on the arousal-valence theory of emotion using large language models.[Method/Process]First,the paper established an emotion classification system with 8 emotional intervals and 56 emotion labels based on the arousal-valence theory.Next,the study utilized large language models to perform fine-grained emotional annotation on government-public interaction texts,constructing a high-quality sentiment classification dataset.Finally based on this,the paper proposed and systematically evaluated an emotion recognition model for this type of sentiment classification system.[Result/Conclusion]Experimental results validate the quality of the constructed dataset and the effectiveness of the sentiment classification method.This research provides reproducible data resources and technical solutions for fine-grained sentiment analysis in the field of online government-public interaction.
作者
滕婕
贺荒兰
胡广伟
刘云
Teng Jie;He Huanglan;Hu Guangwei;Liu Yun(School of Information Management,Nanjing University,Nanjing 210023,China;Government Data Resources Institution of Nanjing University,Nanjing 210023,China;Jiangsu Provincial People s Hospital,Nanjing 210096,China)
出处
《现代情报》
北大核心
2025年第9期58-70,107,共14页
Journal of Modern Information
基金
2024年江苏省研究生科研创新项目“预训练语言模型与大模型协同:面向复杂文本情感精准分类研究”(项目编号:KYCX24_0101)
国家社会科学基金重大项目“大数据驱动的城乡社区服务体系精准化构建研究”(项目编号:20&ZD154)
2024年江苏省研究生科研创新项目“主动健康管理视角下老年人的活动风险感知模式研究”(项目编号:KYCX24_0100)。
关键词
大语言模型
网络问政
情感识别
细粒度情感分析
情绪唤醒度—效价理论
large language models
online government-public interaction
sentiment recognition
fine-grained sentiment analysis
emotion arousal-valence theory
