Yudi Chen

In the post-truth era, effective governance of emergency public sentiment faces significant challenges due to the phenomenon of opinion deviation. Although online public opinion has been extensively investigated, the specific impact of Public Opinion Deviation (POD) on the evolution of group negative emotion remains inadequately understood. To address this gap, this study proposes an explicable framework that integrates community detection, text mining, and opinion dynamics. Opinion deviation communities are identified by applying the Louvain algorithm and TextRank to social media data, followed by a deviation analysis of community topics against core issues. Subsequently, a multi-stage quantification model is constructed to measure the severity of POD. During these processes, we develop a novel Opinion Dynamics F-J model (POD F-J model) and its intervention-oriented model (IPOD F-J model), which incorporate the quantified POD severity to simulate the evolution of group negative emotion. Our findings demonstrate an intrinsic correlation between the severity of opinion deviation and the intensity of group negative emotion at the information level, thereby confirming the necessity of targeted intervention. Simulation experiments indicate that different intervention strategies should be adopted depending on the situation. Moreover, the application of a greedy algorithm identifies the time points corresponding to the peak severity of deviation and its onset as the efficiency-oriented intervention timings. This study provides a data-driven framework for monitoring and mitigating emotional contagion in deviation communities, contributing to both the theory and practice of digital governance.

As a critical defense mechanism against COVID-19, the national public health emergency response system (PHERS) with high resilience enables effective identification, absorption, and resistance of epidemic crises. This resilience is essential for safeguarding public health and rapidly restoring social stability. However, existing studies primarily focus on single-aspect strategies in specific countries, lacking a systematic understanding of how resilience strategies influence PHERS resilience outcomes. Therefore, this study aims to establish evidence-based and configurational resilience strategies to improve the effectiveness of PHERS in responding to epidemic threats. This study proposes a theoretical framework to characterize resilience strategies and resilience outcomes for PHERS. The fuzzy-set qualitative comparative analysis (fsQCA) method is applied to analyze data from 40 countries during the COVID-19 crisis. The findings reveal three configuration paths to enhance robustness and three paths to enhance rapidity. These results emphasize the importance of the synergistic implementation of containment and closures, travel control, public personal protection, and early virus detection in improving PHERS resilience. This study provides a structured approach to understanding PHERS resilience by identifying key configuration paths that enhance robustness and rapidity. The results offer actionable insights for designing resilient PHERS to better respond to future epidemics.

As a method of business communication, emails play an essential role in establishing relationships and support systems among colleagues, partners, and sponsors to achieve business goals. Research has suggested that nonnative speakers of English and novice writers face challenges when composing professional email requests for business purposes. Previous studies have underscored the importance of phrase frames (p-frames), which are recurring multi-word sequences with a variable slot, to fulfill rhetorical functions in academic discourse. However, little research has explored how p-frames are used for written communications in business workplace contexts. This study investigated five-word p-frames used for rhetorical moves in business email requests. Using 1,148 authentic request emails extracted from the Avocado Research Email Collection corpus, we analyzed the distribution and linguistic characteristics of p-frames across five rhetorical moves. Results showed that p-frames were unevenly distributed across the rhetorical moves in business request emails. Two moves, making the inquiry and closing, showed the highest degree of formulaicity. P-frames were used in a variety of messages to soften demands and express politeness. This study has pedagogical implications for teaching English for business purposes to both L2 learners and novice writers.

Effective household and individual disaster preparedness can minimize physical harm and property damage during catastrophic events. To assess the risk and vulnerability of affected areas, it is crucial for relief agencies to understand the level of public preparedness. Traditionally, government agencies have employed nationwide random telephone surveys to gauge the public’s attitudes and actions towards disaster preparedness. However, these surveys may lack generalizability in certain affected locations due to low response rates or areas not covered by the survey. To address this issue and enhance the comprehensiveness of disaster preparedness assessments, we develop a framework that seamlessly integrates machine learning and simulation. Our approach leverages machine learning algorithms to establish relationships between public attitudes towards disaster preparedness and demographic characteristics. Using Monte Carlo simulation, we generate datasets that incorporate demographic information of the affected location based on government-provided demographic distribution data. The generated dataset is then input into the machine learning model to predict the disaster preparedness attitudes of the affected population. We demonstrate the effectiveness of our framework by applying it to Miami-Dade County, where it accurately predicts the level of disaster preparedness. With this innovative approach, relief agencies can have a clearer and more comprehensive understanding of public disaster preparedness.

Effective household and individual disaster preparedness can minimize physical harm and property damage during catastrophic events. To assess the risk and vulnerability of affected areas, it is crucial for relief agencies to understand the level of public preparedness. Traditionally, government agencies have employed nationwide random telephone surveys to gauge the public’s attitudes and actions towards disaster preparedness. However, these surveys may lack generalizability in certain affected locations due to low response rates or areas not covered by the survey. To address this issue and enhance the comprehensiveness of disaster preparedness assessments, we develop a framework that seamlessly integrates machine learning and simulation. Our approach leverages machine learning algorithms to establish relationships between public attitudes towards disaster preparedness and demographic characteristics. Using Monte Carlo simulation, we generate datasets that incorporate demographic information of the affected location based on government-provided demographic distribution data. The generated dataset is then input into the machine learning model to predict the disaster preparedness attitudes of the affected population. We demonstrate the effectiveness of our framework by applying it to Miami-Dade County, where it accurately predicts the level of disaster preparedness. With this innovative approach, relief agencies can have a clearer and more comprehensive understanding of public disaster preparedness.