【作者】 张梦楠； 朱干成； 张红； 王鹏；

【Author】 Mengnan Zhang;Gancheng Zhu;Hong Zhang;Peng Wang;School of Psychology,Shandong Normal University;Zhejiang University,Center for Psychological Sciences;Department of Education,East China Normal University;

【机构】 山东师范大学心理学院； 浙江大学心理科学研究中心； 华东师范大学教育学部；

【摘要】 疫情对人们的生活、工作和学习方式产生了深远影响,显著影响了全球公众的心理健康,导致负面情绪增加。世界范围内实施了各种控制措施,包括限制社会活动的封控政策。虽然很多研究指出未来的公共卫生政策需要解决封控对公众情绪的潜在负面影响,但这方面的研究仍然存在很大空缺。现有研究主要通过问卷调查和访谈来识别消极情绪的类型及其潜在的影响因素。然而,自我报告方法存在被试效应、生态效度较低等局限性,阻碍了对潜在负面影响的深入探索。此外,很少有研究关注封控期间的情绪分析,且现有研究主要利用机器学习来提取信息,与基于词典或基于机器学习的模型相比,深度学习方法在情感分析中表现出更高的准确率和更好的预测性能。为了弥补现有研究的不足,进一步理解消极情绪及其影响机制,本研究利用情感分析和文本挖掘技术,利用社交媒体数据考察不同时期消极情绪的变化。2022年3月15日至2022年5月15日期间,共收集了58606条原始微博,其中包含关键词"上海""封控"。这些微博经过了严格的爬虫、收集和处理过程。随后,采用基于Transformer的情感分析技术对微博文本中表达的6种情感(快乐、惊讶、悲伤、恐惧、愤怒、厌恶)进行识别和分类。此外,为了更深入地了解话题的演变发展,研究使用动态主题模型(Dynamic Topic Modeling,DTM)来分析收集到的微博中与负面情绪有关的变化趋势、潜在影响因素和应对策略。结果表明,在封控期间,尽管负性情绪占主导地位,但总体情绪趋向积极发展,愤怒、厌恶和悲伤的变化显著。这些情绪的变化和调整可以通过马斯洛的需求层次理论来解释。首先,满足公众在封控期应对负性情绪时的生理需求和安全需求。第二,关注公众的归属和爱的需求。尤其是带着厌恶和悲伤的情绪,很多人觉得得不到他人的尊重。第三,可以通过强调追求尊重需要,甚至是自我实现来调动积极情绪。本研究通过创造性地结合定量和定性方法,并利用大数据分析,提供了对情绪的洞察,为通过社交媒体动态检测和干预公众心理健康提供了一种新的方法。从纵向角度理解情绪变化有助于缓解重大突发公共卫生事件或风险对个体的负面影响。这种心理学和计算机科学的跨学科研究为公众心理健康的改善提供了有价值的应用,并为未来该领域的研究提供了参考。更多还原

【Abstract】 The COVID-19 pandemic has had a profound impact on people’s way of life,work,and study,and has significantly affected global public mental health,leading to an increase in negative emotions.Various control measures,including lockdown policies restricting social activities,have been implemented in response to the infectious nature of the virus.However,these measures have inadvertently caused a surge in negative emotions,leading to emotional overeating,emotional dysregulation among children,and burnout among local social workers.While studies have acknowledged the need for future public health policies to address the unexpected negative impact of lockdowns on public sentiment,this aspect remains underexplored.Current research primarily relies on questionnaires and interviews to identify types of negative emotions and their potential influencing factors.However,these self-reported methods have limitations,such as researcher-driven content,low ecological validity,and biased self-reporting,hindering in-depth exploration of potential negative effects.Moreover,few studies have focused on analyzing emotions during the lockdown period,with existing research mainly utilizing machine learning analysis to extract information.Deep learning methods have demonstrated higher accuracy and better prediction performance in sentiment analysis compared to lexicon-based or machine learning-based models.To bridge the gap in existing research and further understand negative emotions and their effects,this study leverages sentiment analysis and text mining technology to examine changes in negative emotions during different periods using social media data.In present study,a total of 58,606 original microblogs were collected,which contain the keywords "Shanghai" and "lockdown" from the period between March 15,2022,and May 15,2022.These microblogs were subjected to a rigorous scraping,gathering,and filtering process.Subsequently,a Transformer-based sentiment analysis technique was employed to discern and categorize six emotions(happiness,surprise,sadness,fear,anger,and disgust) expressed in the microblog posts,thereby adopting a longitudinal perspective.Moreover,to gain deeper insights into the evolving discourse,Dynamic Topic Modeling(DTM) was utilized to explore and analyze the changing trends,potential influencing factors,and coping strategies pertaining to negative emotions within the collected microblogs.The results indicate that during the lockdown period,overall emotions tended to become more positive,though negative emotions still dominated,with significant variations observed in anger,disgust,and sadness.These changes and adjustments in emotions are interpreted through Maslow’s hierarchy of needs theory.The first is to meet the physiological needs and safety needs of the public when coping with negative emotions during the lockdown period.In this study,each kind of emotional public mentioned the most,is the lack of living supplies,epidemic prevention work is not in place easy to increase the risk of infection,and medical treatment difficulties.Second,pay attention to the public’s need for belonging & love.Especially with disgust and sadness emotion,many people feel not respected by them.Third,we can mobilize positive emotions by emphasizing the pursuit of self-esteem needs,even self-realization.By creatively combining quantitative and qualitative methods and utilizing big data analysis,this study provides insights into emotions,offering a novel approach for dynamically detecting and intervening in public mental health through social media.Understanding emotional changes from a longitudinal perspective can help mitigate the negative impact of major public health emergencies or risks on individuals.This interdisciplinary research in psychology and computer science offers valuable applications for the betterment of public mental health and provides a reference for future studies in this field.更多还原