Social Network Analysis of COVID-19 Sentiments: 10 Metropolitan Cities in Italy

Abstract

The pandemic spread rapidly across Italy, putting the region’s health system on the brink of collapse, and generating concern regarding the government’s capacity to respond to the needs of patients considering isolation measures. This study developed a sentiment analysis using millions of Twitter data during the first wave of the COVID-19 pandemic in 10 metropolitan cities in Italy’s (1) north: Milan, Venice, Turin, Bologna; (2) central: Florence, Rome; (3) south: Naples, Bari; and (4) islands: Palermo, Cagliari. Questions addressed are as follows: (1) How did tweet-related sentiments change over the course of the COVID-19 pandemic, and (2) How did sentiments change when lagged with policy shifts and/or specific events? Findings show an assortment of differences and connections across Twitter sentiments (fear, anger, and joy) based on policy measures and geographies during the COVID-19 pandemic. Results can be used by policy makers to quantify the satisfactory level of positive/negative acceptance of decision makers and identify important topics related to COVID-19 policy measures, which can be useful for imposing geographically varying lockdowns and protective measures using historical data.

Analysis of Pharmaceutical Companies’ Social Media Activity during the COVID-19 Pandemic and Its Impact on the Public

“…The results of this work indicate a stark presence of fear in the public's response in the COVID-19 communication channel, with an upward trend. Recent studies support that negative sentiments are dominant on social media in the COVID-19 context [48,57] and detect fear and even panic as the prevailing emotion [28,58,59]. Although our findings do not provide grounds for similar conclusions, the steadily increasing presence of fear was detected in users' responses.…”

“…Sentiment analysis was employed to detect the fluctuations of the public's attitude from positive to negative and vice versa using wordbased techniques [20][21][22][23][24]. Other studies conducted a more fine-grained analysis combining sentiment with emotion analysis in order to investigate the dominant emotions and their correlation to external events and various periods [25][26][27][28]. Topic modeling, mainly using the latent Dirichlet allocation algorithm [29], was performed to identify the most debated themes of discussion [23,24,30,31].…”

Analyzing Spanish-Language Public Sentiment in the Context of a Pandemic and Social Unrest: The Panama Case

“…Some of these approaches, and their analogs in languages other than English, have been successfully applied to scenarios related to the scope of our research: the effects of the pandemic and massive social events on the general public sentiment of a determined territory. In an effort to measure the emotions expressed by residents of 10 Italian cities during the early months of the pandemic, Fernandez et al performed a lexicon-based sentiment analysis of more than 4 million coronavirus-related tweets [ 36 ], with results pointing at strong emotional responses towards health policies in the country. A similar analysis was also carried out by Kydros et al in Greece during the acute phase of the pandemic [ 37 ].…”