When a crisis strikes—whether it is a hurricane, a pandemic, or a social movement—public reaction is rarely uniform. While traditional sentiment analysis often aggregates data to provide a "global" view of public opinion, a growing body of research suggests that this approach obscures the most critical insights. Instead, researchers are finding that the physical distance between a person and the "epicenter" of an event fundamentally reshapes both the topics they discuss and the emotions they express.
By integrating geographic and temporal dimensions into natural language processing (NLP), researchers can map the "emotional epidemiology" of a crisis, revealing how proximity dictates the transition from abstract sympathy to immediate survival concerns.
The Epicenter Effect on Topic and Tone
The most significant finding across this research area is that geographic proximity acts as a filter for public discourse. In Analysis of public opinion evolution of COVID-19 based on LDA-ARMA hybrid model, Muni Zhuang and colleagues demonstrated this during the early days of COVID-19. They found that individuals at the epicenter (Wuhan) prioritized immediate survival and infrastructure concerns, while those further away focused on institutional oversight and supportive actions.
This divergence is not just about what is said, but how it is felt. In Identifying and Tracking Sentiments and Topics from Social Media Texts during Natural Disasters, Min Yang and Heng Ji developed the LDST (Location-based Dynamic Sentiment-Topic) model. Their work showed that during Hurricane Sandy, concerns about power outages were localized to specific regions, while post-disaster relief became a broader, non-local topic. This suggests that "public opinion" is actually a collection of distinct regional trajectories that evolve at different speeds.
Mapping Emotional Epidemiology
Proximity also dictates the specific "blend" of emotions a population experiences. Jie Zhuang and Tai-Quan Peng explored this in Mixed and blended emotional reactions to 2014 Ebola outbreak, noting that fear and anger peak in areas with confirmed cases, while sadness—often driven by distant sympathy—is more prevalent in remote regions.
Interestingly, the magnitude of a physical event is often less important than its proximity to population centers. Evgenia Ilyinskaya and Hanne Krage Carlsen found that "small-scale" geohazards in Iceland caused massive shifts in public emotion simply because they occurred near urban centers, whereas larger events in remote areas went virtually unnoticed on social media (Brief communication: Small-scale geohazards cause significant and highly variable impacts on emotions).
Sociological Gradients and Recovery
Beyond immediate disasters, geographic sentiment analysis reveals deep-seated sociological divides. In Examining Rural and Urban Sentiment Difference in COVID-19–Related Topics on Twitter, Yongtai Liu and Zhiyu Wan used word embeddings to show that rural populations expressed significantly more negativity toward mandates, while urban users focused their frustration on political figures and misinformation.
This regional context also influences how quickly a community "recovers" emotionally. Grace M. Leffler demonstrates in A Tutorial on Collecting and Processing Longitudinal Social Media Data that the duration and geographic extent of an event (like the 2020 U.S. wildfires) create different "slopes" in sentiment return-to-baseline.
Why It Works: The Proximity-Empathy Link
The underlying mechanism for these shifts appears to be a combination of personal risk and community empathy. While proximity to a health threat increases fear, proximity to social issues can actually increase support. Hanjia Lyu and Jiebo Luo found that areas with higher frequencies of hate crimes showed a decrease in negative sentiment toward the #StopAsianHate movement, suggesting that local exposure to violence can foster community empathy (State-level Racially Motivated Hate Crimes Contrast Public Opinion on the #StopAsianHate and #StopAAPIHate Movement).
By treating social media not as a single global monolith but as a high-resolution map of local experiences, these researchers provide a framework for more effective public health messaging, disaster response, and social policy. As Mauro Coletto and Raffaele Perego argue, understanding these geographic gradients is vital for navigating the social tensions inherent in global crises (Sentiment-enhanced multidimensional analysis of online social networks).