基于辟谣微博的互动及热门评论情感倾向的辟谣效果研究——以新冠疫情相关辟谣微博为例

Research on Rumor-refutation Effectiveness Based on the Interactions and Popular Comments'Emotional Tendencies of the Rumor-refuting Microblogs:Taking Rumor-refuting Microblogs Related with COVID-2019 as an Example

[目的/意义]增强社交媒体平台上的辟谣效果可以为辟谣提供新的思路,加快辟谣过程,减轻谣言的危害。[方法/过程]提出了辟谣效果指数(REI)。搜集2020年1月1日至2020年4月20日由“微博辟谣”转发的辟谣信息,应用自然语言处理(NLP)方法,以提取辟谣微博的文本特征,分析辟谣微博文本、热门评论的情感倾向。采用人工标注法验证了REI对辟谣效果评价的有效性,并探索REI与辟谣微博的内容特征与背景特征之间的关系,建立了四种回归模型,基于综合拟合表现最优的XGBoost回归模型,利用SHapley Additive exPlanations(SHAP)对回归结果进行可视化和分析。[结果/结论]基于分析结果,为不同辟谣情景(如辟谣者的影响力、话题热度等)提出了如何组织辟谣信息以增强辟谣效果的决策建议,REI也可应用于其他社交媒体平台。

[Purpose/Significance]Enhancing social media rumor refutation effectiveness can shed light on the rumor refutation,speed up the rumor-refuting process,and lessen the harm of rumors.[Method/Process]This paper first proposes a rumor-refutation effectiveness index(REI).We collected the rumor-refuting microblogs forwarded by"Weibo Refutes Rumor"from January 1,2020 to April 20,2020.To extract the text characteristics and analyze the sentiment of these rumor-refuting microblogs and popular comments,Natural Language Processing(NLP)approaches were applied.Human annotations were applied to verify the effectiveness of the REI.To explore the relationship between the REI and the content and contextual factors of the rumor-refuting microblogs,four regression models were established.Based on XGBoostRegressor with the best overall fitting performance,SHapley Additive exPlanations(SHAP)was used to visualize and analyze the regression results.[Result/Conclusion]Decision making suggestions on how to organize rumor-refuting messages under different situations such as rumor-refuting microblog author’s influence and heat of topics were proposed.The REI can also be applied in other social media platforms.