摘要
[目的/意义]突发公共卫生事件中谣言的迅速传播可能会引发群体性的焦虑和恐慌,识别社交媒体中潜在的谣言传播者,研究及评估影响谣言传播者识别的重要特征,为舆情管控和网络治理提供策略。[方法/过程]提出一种突发公共卫生事件情景下多特征融合的潜在谣言传播者识别模型,首先基于BERT-BiLSTM模型提取微博的语义特征,然后与用户特征、微博特征以及情感特征进行融合,最后基于LightGBM算法构建用户分类模型,并利用SHAP值对模型进行分析。[结果/结论]研究结果表明,融合多特征的突发公共卫生事件谣言传播者识别模型在微博数据集上的准确率能够达到87.94%,说明该模型具有较好的识别效果,提出的4个维度的特征对谣言传播者识别均有贡献,其中文本语义特征对谣言传播者识别准确率的提升最高。
[Purpose/Significance]In public health emergencies,the rapid spread of rumors may cause mass anxiety and panic.This study aims to detect potential rumor spreaders in social media,explore and evaluate the important characteristics affecting rumor spreader identification,and provide strategies for public opinion control and network governance.[Method/Process]This study proposed a detection model for potential rumor spreaders based on multi-feature fusion in the context of public health emergencies.Firstly,the semantic features of Weibo were extracted by the BERT-BiLSTM model,and then fused with user features,Weibo features and emotion features.Finally,the user classification model was constructed based on LightGBM algorithm,and the model was explained by SHAP value.[Result/Conclusion]The experimental results show that the accuracy rate of the fusion multi-feature rumor spreader identification model for public health emergencies can reach 87.94%on the Weibo data set,indicating that the model has good detection effect.Moreover,the features of four dimensional proposed in this paper contribute to rumor spreader identification,and the text semantic features have the highest improvement in the accuracy of rumor spreader identification.
作者
曾子明
张瑜
李婷婷
Zeng Ziming;Zhang Yu;Li Tingting(School of Information Management,Wuhan University,Wuhan 430072)
出处
《图书情报工作》
CSSCI
北大核心
2022年第13期80-90,共11页
Library and Information Service
基金
国家社会科学基金项目"面向突发公共卫生事件的网络舆情时空演化与决策支持研究"(项目编号:21BTQ046)研究成果之一。
