Topic-Emotion Propagation Mechanism of Public Emergencies in Social Networks

Understanding Public Emotions: Spatiotemporal Dynamics in the Post-Pandemic Era Through Weibo Data

Prolonged exposure to public health crises in the post-pandemic era poses significant threats to global mental health. To address this, we developed a conceptual model to analyse the spatiotemporal distribution of public emotions, using Weibo data from the 2022 Beijing bar outbreak (9 June-18 August). The model integrates lexicon-based emotion analysis, spatial autocorrelation tests, and content analysis to provide a comprehensive understanding of emotional responses across stages and regions. The findings reveal a multi-peak emotional cycle spanning emergency, contagion, and resolution stages, with significant emotional clustering in emergency zones, surrounding areas, and regions visited by infected individuals. Through coding, we identified 24 main-categories and 90 sub-categories, distilled into nine core themes that illustrate the interplay between influencing factors, public emotions, and online behaviours. Positive public emotions (e.g., hopefulness, gratitude, optimism) were linked to pandemic improvements and policy implementation, driving behaviours such as supporting prevention measures and resisting misinformation. Negative emotions (e.g., anger, anxiety, sadness) stemmed from severe outbreaks, insufficient controls, and restrictions on freedoms, leading to criticism and calls for accountability. This study bridges big data analytics with behavioural science, offering critical insights into evolving public emotions and behaviours. By highlighting spatiotemporal patterns and emotional dynamics, it provides actionable guidance for governments and health organizations to design targeted interventions, foster resilience, and better manage future social crises with precision and empathy.

Investigating the Influencing Factors and Correlation Between Popularity and Emotion of Public Opinion During Disasters: Evidence from the "7.20" Rainstorm in China

Disasters not only directly cause casualties and property losses but also significantly impact public opinion. In order to identify the evolution characteristics and influencing factors of public opinion during disasters, this paper put forward an analytical framework for analyzing public opinion, which clarified the relationships among key characteristics of public opinion and emphasized the role of official agencies in the processes of information releasing and information feedback. Guided by this framework, this paper collected the public opinion on Sina Weibo during the extraordinary "7.20" rainstorm in Henan Province, China. By analyzing the changes in the discussion regarding both the popularity of and the emotion displayed in Sina Weibo comments over time, it was found that the evolution of public opinion was closely related to disaster development. Novel informational content or innovative forms of information contribute to enhancing the discussion of popularity, while the events or emotions expressed within the information elicit corresponding emotional responses from the public. As popularity increased, the prevalence of negative emotions was observed to diminish concurrently with a rise in the proportion of neutral emotions. Based on these results, some suggestions on the management of public opinion during disasters were put forward.

Exploring the impact of sentiment on multi-dimensional information dissemination using COVID-19 data in China

The outbreak of information epidemic in crisis events, with the channel effect of social media, has brought severe challenges to global public health. Combining information, users and environment, understanding how emotional information spreads on social media plays a vital role in public opinion governance and affective comfort, preventing mass incidents and stabilizing the network order. Therefore, from the perspective of the information ecology and elaboration likelihood model (ELM), this study conducted a comparative analysis based on two large-scale datasets related to COVID-19 to explore the influence mechanism of sentiment on the forwarding volume, spreading depth and network influence of information dissemination. Based on machine learning and social network methods, topics, sentiments, and network variables are extracted from large-scale text data, and the dissemination characteristics and evolution rules of online public opinions in crisis events are further analyzed. The results show that negative sentiment positively affects the volume, depth, and influence compared with positive sentiment. In addition, information characteristics such as richness, authority, and topic influence moderate the relationship between sentiment and information dissemination. Therefore, the research can build a more comprehensive connection between the emotional reaction of network users and information dissemination and analyze the internal characteristics and evolution trend of online public opinion. Then it can help sentiment management and information release strategy when emergencies occur.

Exploring users' content creation and information dissemination behavior in social media: The moderating effect of social presence

Users' personality traits reveal different social media behavior characteristics. In order to explore the intrinsic relationships between personality traits and social media behavior, this study analyzes the influence of users' personality traits on social media content creation and information dissemination behavior, as well as the moderating effect of social presence. We collect users' personality data via questionnaires, crawl social media behavior data of samples from social media sites, and then establish regression models to test the research hypotheses. The results show that extraversion has a positive impact on content creation and information dissemination behavior, conscientiousness has a negative impact on content creation behavior, openness and agreeableness have no significant impact on social media behavior, and social presence has significant moderating effects on the relationships between personality traits and social media behavior.

Multimodal Feature Fusion Method for Unbalanced Sample Data in Social Network Public Opinion

With the wide application of social media, public opinion analysis in social networks has been unable to be met through text alone because the existing public opinion information includes data information of various modalities, such as voice, text, and facial expressions. Therefore multi-modal emotion analysis is the current focus of public opinion analysis. In addition, multi-modal emotion recognition of speech is an important factor restricting the multi-modal emotion analysis. In this paper, the emotion feature retrieval method for speech is firstly explored and the processing method of sample disequilibrium data is then analyzed. By comparing and studying the different feature fusion methods of text and speech, respectively, the multi-modal feature fusion method for sample disequilibrium data is proposed to realize multi-modal emotion recognition. Experiments are performed using two publicly available datasets (IEMOCAP and MELD), which shows that processing multi-modality data through this method can obtain good fine-grained emotion recognition results, laying a foundation for subsequent social public opinion analysis.

Investigating the Public Sentiment in Major Public Emergencies Through the Complex Networks Method: A Case Study of COVID-19 Epidemic

The main purpose of this study is to investigate what topic indicators correlate with public sentiment during “coronavirus disease 2019 (COVID-19) epidemic” and which indicators control the complex networks of the topic indicators. We obtained 68,098 Weibo, categorized them into 11 topic indicators, and grouped these indicators into three dimensions. Then, we constructed the complex networks model of Weibo's topics and examined the key indicators affecting the public's sentiment during the major public emergency. The results showed that “positive emotion” is positively correlated with “recordings of epidemic” and “foreign comparisons,” while “negative emotion” is negatively correlated with “government image,” “recordings of epidemic,” and “asking for help online.” In addition, the two vertexes of “recordings of epidemic” and “foreign comparisons” are the most important “bridges” which connect the government and the public. The “recordings of epidemic” is the main connection “hub” between the government and the media. In other words, the “recordings of epidemic” is the central topic indicator that controls the entire topic network. In conclusion, the government should publish the advance of the events through official media on time and transparent way and create a platform where everyone can speak directly to the government for advice and assistance during a major public emergency in the future.