How the world's collective attention is being paid to a pandemic: COVID-19 related n-gram time series for 24 languages on Twitter

What We Ask about When We Ask about Quarantine? Content and Sentiment Analysis on Online Help-Seeking Posts during COVID-19 on a Q&A Platform in China

The COVID-19 outbreak, a recent major public health emergency, was the first national health crisis since China entered the era of mobile social media. In this context, the public posted many quarantine-related posts for help on social media. Most previous studies of social media during the pandemic focused only on people's emotional needs, with less analysis of quarantine help-seeking content. Based on this situation, this study analyzed the relationship between the number of quarantine help-seeking posts and the number of new diagnoses at different time points in the pandemic using Zhihu, the most comprehensive topic discussion platform in China. It showed a positive correlation between the number of help-seeking posts and the pandemic's severity. Given the diversity of people's help-seeking content, this study used topic model analysis and sentiment analysis to explore the key content of people's quarantine help-seeking posts during the pandemic. In light of the framework of uses and gratifications, we found that people posted the most questions in relation to help with information related to pandemic information and quarantine information. Interestingly, the study also found that the content of people's quarantine posts during the pandemic was primarily negative in sentiment. This study can thus help the community understand the changes in people's perceptions, attitudes, and concerns through their reactions to emergencies and then formulate relevant countermeasures to address pandemic control and information regulation, which will have implications for future responses to public health emergencies. Moreover, in terms of psychological aspects, it will help implement future mental health intervention strategies and better address the public's psychological problems.

On network backbone extraction for modeling online collective behavior

Collective user behavior in social media applications often drives several important online and offline phenomena linked to the spread of opinions and information. Several studies have focused on the analysis of such phenomena using networks to model user interactions, represented by edges. However, only a fraction of edges contribute to the actual investigation. Even worse, the often large number of non-relevant edges may obfuscate the salient interactions, blurring the underlying structures and user communities that capture the collective behavior patterns driving the target phenomenon. To solve this issue, researchers have proposed several network backbone extraction techniques to obtain a reduced and representative version of the network that better explains the phenomenon of interest. Each technique has its specific assumptions and procedure to extract the backbone. However, the literature lacks a clear methodology to highlight such assumptions, discuss how they affect the choice of a method and offer validation strategies in scenarios where no ground truth exists. In this work, we fill this gap by proposing a principled methodology for comparing and selecting the most appropriate backbone extraction method given a phenomenon of interest. We characterize ten state-of-the-art techniques in terms of their assumptions, requirements, and other aspects that one must consider to apply them in practice. We present four steps to apply, evaluate and select the best method(s) to a given target phenomenon. We validate our approach using two case studies with different requirements: online discussions on Instagram and coordinated behavior in WhatsApp groups. We show that each method can produce very different backbones, underlying that the choice of an adequate method is of utmost importance to reveal valuable knowledge about the particular phenomenon under investigation.

Can people hear others' crying?: A computational analysis of help-seeking on Weibo during COVID-19 outbreak in China

Social media like Weibo has become an important platform for people to ask for help during COVID-19 pandemic. Using a complete dataset of help-seeking posts on Weibo during the COVID-19 outbreak in China (N=3,705,188), this study mapped their characteristics and analyzed their relationship with the epidemic development at the aggregate level, and examined the influential factors to determine whether and the extent the help-seeking crying could be heard at the individual level using computational methods for the first time. It finds that the number of help-seeking posts on Weibo has a Granger causality relationship with the number of confirmed COVID-19 cases with a time lag of eight days. This study then proposes a 3C framework to examine the direct influence of content, context, and connection on the responses (measured by retweets and comments) and assistance that help-seekers might receive as well as their indirect effects on assistance through the mediation of both retweets and comments. The differential influences of content (theme and negative sentiment), context (Super topic community, spatial location of posting, and the period of sending time), and connection (the number of followers, whether mentioning others, and verified status of authors and sharers) have been reported and discussed.

Sentiment Analysis of Users’ Reactions on Social Media During the Pandemic

During the outbreak of the COVID-19 pandemic, social networks became the preeminent medium for communication, social discussion, and entertainment. Social network users are regularly expressing their opinions about the impacts of the coronavirus pandemic. Therefore, social networks serve as a reliable source for studying the topics, emotions, and attitudes of users that have been discussed during the pandemic. In this paper, we investigate the reactions and attitudes of people towards topics raised on social media platforms. We collected data of two large-scale COVID-19 datasets from Twitter and Instagram for six and three months, respectively. This paper analyzes the reaction of social network users in terms of different aspects including sentiment analysis, topic detection, emotions, and the geo-temporal characteristics of our dataset. We show that the dominant sentiment reactions on social media are neutral, while the most discussed topics by social network users are about health issues. This paper examines the countries that attracted a higher number of posts and reactions from people, as well as the distribution of health-related topics discussed in the most mentioned countries. We shed light on the temporal shift of topics over countries. Our results show that posts from the top-mentioned countries influence and attract more reactions worldwide than posts from other parts of the world.

Air pollution perception in ten countries during the COVID-19 pandemic

As largely documented in the literature, the stark restrictions enforced worldwide in 2020 to curb the COVID-19 pandemic also curtailed the production of air pollutants to some extent. This study investigates the perception of the air pollution as assessed by individuals located in ten countries: Australia, Brazil, China, Ghana, India, Iran, Italy, Norway, South Africa and the USA. The perceptions towards air quality were evaluated by employing an online survey administered in May 2020. Participants (N = 9394) in the ten countries expressed their opinions according to a Likert-scale response. A reduction in pollutant concentration was clearly perceived, albeit to a different extent, by all populations. The survey participants located in India and Italy perceived the largest drop in the air pollution concentration; conversely, the smallest variation was perceived among Chinese and Norwegian respondents. Among all the demographic indicators considered, only gender proved to be statistically significant.

Facilitators and Barriers of COVID-19 Vaccine Promotion on Social Media in the United States: A Systematic Review

Background and Objectives: Information regarding the COVID-19 pandemic has spread internationally through a variety of platforms, including social media. While efforts have been made to help reduce the spread of misinformation on social media, many platforms are still largely unregulated. The influence of social media use on vaccination promotion is not fully understood. This systematic review aims to identify facilitators and barriers associated with vaccine promotion through social media use. Materials and Methods: Reviewers analyzed 25 articles and identified common themes. Facilitators of vaccine promotion included an increase in the efforts of social media companies to reduce misinformation, the use of social media to spread information on public health and vaccine promotion, and the positive influence towards vaccinations of family and friends. Results and Conclusions: Identified barriers to vaccine promotion included the spread of misinformation, decreased vaccine acceptance among users of social media for COVID-19 related information due to polarization, and a lack of regulation on social media platforms. The results of this review provide insight for improving public health campaign promotion on social media and can help inform policy on social media regulation and misinformation prevention.

Augmenting Semantic Lexicons Using Word Embeddings and Transfer Learning

Sentiment-aware intelligent systems are essential to a wide array of applications. These systems are driven by language models which broadly fall into two paradigms: Lexicon-based and contextual. Although recent contextual models are increasingly dominant, we still see demand for lexicon-based models because of their interpretability and ease of use. For example, lexicon-based models allow researchers to readily determine which words and phrases contribute most to a change in measured sentiment. A challenge for any lexicon-based approach is that the lexicon needs to be routinely expanded with new words and expressions. Here, we propose two models for automatic lexicon expansion. Our first model establishes a baseline employing a simple and shallow neural network initialized with pre-trained word embeddings using a non-contextual approach. Our second model improves upon our baseline, featuring a deep Transformer-based network that brings to bear word definitions to estimate their lexical polarity. Our evaluation shows that both models are able to score new words with a similar accuracy to reviewers from Amazon Mechanical Turk, but at a fraction of the cost.

COVID-19 vaccine prioritization based on district classification in Yogyakarta Province, Indonesia

Due to limited availability, Indonesia's coronavirus disease 2019 (COVID-19) vaccination will be done in 4 stages until herd immunity has been reached. Yogyakarta, an education and tourist destination, needs to get a specific, spatial estimation of the exact need for COVID-19 vaccination without delay. This study sheds light on identifying which districts should be prioritized at each vaccination phase. Secondary data collected from provincial, and county-level statistical agencies were quantitatively calculated by the Z-Score method. The results indicate that the first phase of vaccination should prioritize Pengasih and Sentolo districts in Kulon Progo Regency, which have a large number of health workers; the districts of Depok, Banguntapan, Piyungan, Sewon, Wonosari, Gamping, Mlati and Ngaglik should be done in the second phase based on the fact that these districts have many public service officials as well as elderly people; Umbulharjo and Depok districts will be approached in the third phase since they have more vulnerable groups and facilities that may promote COVID- 19 transmission during their daily activities; while the fourth phase should focus on the districts of Banguntapan, Sewon, Kasihan, Gamping, Mlati, Depok, and Ngaglik due to the intensity of COVID-19 clusters discovered there. Overall, vaccination would be given the priority in the districts with the largest number of people in need, i.e., public service officers, elderly people and those likely to be exposed to the coronavirus causing COVID-19.

Storywrangler: A massive exploratorium for sociolinguistic, cultural, socioeconomic, and political timelines using Twitter

In real time, Twitter strongly imprints world events, popular culture, and the day-to-day, recording an ever-growing compendium of language change. Vitally, and absent from many standard corpora such as books and news archives, Twitter also encodes popularity and spreading through retweets. Here, we describe Storywrangler, an ongoing curation of over 100 billion tweets containing 1 trillion 1-grams from 2008 to 2021. For each day, we break tweets into 1-, 2-, and 3-grams across 100+ languages, generating frequencies for words, hashtags, handles, numerals, symbols, and emojis. We make the dataset available through an interactive time series viewer and as downloadable time series and daily distributions. Although Storywrangler leverages Twitter data, our method of tracking dynamic changes in n-grams can be extended to any temporally evolving corpus. Illustrating the instrument's potential, we present example use cases including social amplification, the sociotechnical dynamics of famous individuals, box office success, and social unrest.

Hurricanes and hashtags: Characterizing online collective attention for natural disasters

We study collective attention paid towards hurricanes through the lens of n-grams on Twitter, a social media platform with global reach. Using hurricane name mentions as a proxy for awareness, we find that the exogenous temporal dynamics are remarkably similar across storms, but that overall collective attention varies widely even among storms causing comparable deaths and damage. We construct 'hurricane attention maps' and observe that hurricanes causing deaths on (or economic damage to) the continental United States generate substantially more attention in English language tweets than those that do not. We find that a hurricane's Saffir-Simpson wind scale category assignment is strongly associated with the amount of attention it receives. Higher category storms receive higher proportional increases of attention per proportional increases in number of deaths or dollars of damage, than lower category storms. The most damaging and deadly storms of the 2010s, Hurricanes Harvey and Maria, generated the most attention and were remembered the longest, respectively. On average, a category 5 storm receives 4.6 times more attention than a category 1 storm causing the same number of deaths and economic damage.

Prediction of COVID-19 Waves Using Social Media and Google Search: A Case Study of the US and Canada

The ongoing COVID-19 pandemic has posed a severe threat to public health worldwide. In this study, we aimed to evaluate several digital data streams as early warning signals of COVID-19 outbreaks in Canada, the US and their provinces and states. Two types of terms including symptoms and preventive measures were used to filter Twitter and Google Trends data. We visualized and correlated the trends for each source of data against confirmed cases for all provinces and states. Subsequently, we attempted to find anomalies in indicator time-series to understand the lag between the warning signals and real-word outbreak waves. For Canada, we were able to detect a maximum of 83% of initial waves 1 week earlier using Google searches on symptoms. We divided states in the US into two categories: category I if they experienced an initial wave and category II if the states have not experienced the initial wave of the outbreak. For the first category, we found that tweets related to symptoms showed the best prediction performance by predicting 100% of first waves about 2-6 days earlier than other data streams. We were able to only detect up to 6% of second waves in category I. On the other hand, 78% of second waves in states of category II were predictable 1-2 weeks in advance. In addition, we discovered that the most important symptoms in providing early warnings are fever and cough in the US. As the COVID-19 pandemic continues to spread around the world, the work presented here is an initial effort for future COVID-19 outbreaks.

Multilingual topic modeling for tracking COVID-19 trends based on Facebook data analysis

Social data has shown important role in tracking, monitoring and risk management of disasters. Indeed, several works focused on the benefits of social data analysis for the healthcare practices and curing domain. Similarly, these data are exploited now for tracking the COVID-19 pandemic but the majority of works exploited Twitter as source. In this paper, we choose to exploit Facebook, rarely used, for tracking the evolution of COVID-19 related trends. In fact, a multilingual dataset covering 7 languages (English (EN), Arabic (AR), Spanish (ES), Italian (IT), German (DE), French (FR) and Japanese (JP)) is extracted from Facebook public posts. The proposal is an analytics process including a data gathering step, pre-processing, LDA-based topic modeling and presentation module using graph structure. Data analysing covers the duration spanned from January 1st, 2020 to May 15, 2020 divided on three periods in cumulative way: first period January-February, second period March-April and the last one to 15 May. The results showed that the extracted topics correspond to the chronological development of what has been circulated around the pandemic and the measures that have been taken according to the various languages under discussion representing several countries.

COVID-19: Detecting Government Pandemic Measures and Public Concerns from Twitter Arabic Data Using Distributed Machine Learning

Today's societies are connected to a level that has never been seen before. The COVID-19 pandemic has exposed the vulnerabilities of such an unprecedently connected world. As of 19 November 2020, over 56 million people have been infected with nearly 1.35 million deaths, and the numbers are growing. The state-of-the-art social media analytics for COVID-19-related studies to understand the various phenomena happening in our environment are limited and require many more studies. This paper proposes a software tool comprising a collection of unsupervised Latent Dirichlet Allocation (LDA) machine learning and other methods for the analysis of Twitter data in Arabic with the aim to detect government pandemic measures and public concerns during the COVID-19 pandemic. The tool is described in detail, including its architecture, five software components, and algorithms. Using the tool, we collect a dataset comprising 14 million tweets from the Kingdom of Saudi Arabia (KSA) for the period 1 February 2020 to 1 June 2020. We detect 15 government pandemic measures and public concerns and six macro-concerns (economic sustainability, social sustainability, etc.), and formulate their information-structural, temporal, and spatio-temporal relationships. For example, we are able to detect the timewise progression of events from the public discussions on COVID-19 cases in mid-March to the first curfew on 22 March, financial loan incentives on 22 March, the increased quarantine discussions during March-April, the discussions on the reduced mobility levels from 24 March onwards, the blood donation shortfall late March onwards, the government's 9 billion SAR (Saudi Riyal) salary incentives on 3 April, lifting the ban on five daily prayers in mosques on 26 May, and finally the return to normal government measures on 29 May 2020. These findings show the effectiveness of the Twitter media in detecting important events, government measures, public concerns, and other information in both time and space with no earlier knowledge about them.

Attention dynamics on the Chinese social media Sina Weibo during the COVID-19 pandemic

Understanding attention dynamics on social media during pandemics could help governments minimize the effects. We focus on how COVID-19 has influenced the attention dynamics on the biggest Chinese microblogging website Sina Weibo during the first four months of the pandemic. We study the real-time Hot Search List (HSL), which provides the ranking of the most popular 50 hashtags based on the amount of Sina Weibo searches. We show how the specific events, measures and developments during the epidemic affected the emergence of different kinds of hashtags and the ranking on the HSL. A significant increase of COVID-19 related hashtags started to occur on HSL around January 20, 2020, when the transmission of the disease between humans was announced. Then very rapidly a situation was reached where COVID-related hashtags occupied 30-70% of the HSL, however, with changing content. We give an analysis of how the hashtag topics changed during the investigated time span and conclude that there are three periods separated by February 12 and March 12. In period 1, we see strong topical correlations and clustering of hashtags; in period 2, the correlations are weakened, without clustering pattern; in period 3, we see a potential of clustering while not as strong as in period 1. We further explore the dynamics of HSL by measuring the ranking dynamics and the lifetimes of hashtags on the list. This way we can obtain information about the decay of attention, which is important for decisions about the temporal placement of governmental measures to achieve permanent awareness. Furthermore, our observations indicate abnormally higher rank diversity in the top 15 ranks on HSL due to the COVID-19 related hashtags, revealing the possibility of algorithmic intervention from the platform provider.

SUPPLEMENTARY INFORMATION

The online version contains supplementary material available at 10.1140/epjds/s13688-021-00263-0.

"Show this thread": policing, disruption and mobilisation through Twitter. An analysis of UK law enforcement tweeting practices during the Covid-19 pandemic

Crisis and disruption are often unpredictable and can create opportunities for crime. During such times, policing may also need to meet additional challenges to handle the disruption. The use of social media by officials can be essential for crisis mitigation and crime reduction. In this paper, we study the use of Twitter for crime mitigation and reduction by UK police (and associated) agencies in the early stages of the Covid-19 pandemic. Our findings suggest that whilst most of the tweets from our sample concerned issues that were not specifically about crime, especially during the first stages of the pandemic, there was a significant increase in tweets about fraud, cybercrime and domestic abuse. There was also an increase in retweeting activity as opposed to the creation of original messages. Moreover, in terms of the impact of tweets, as measured by the rate at which they are retweeted, followers were more likely to 'spread the word' when the tweet was content-rich (discussed a crime specific matter and contained media), and account holders were themselves more active on Twitter. Considering the changing world we live in, criminal opportunity is likely to evolve. To help mitigate this, policy makers and researchers should consider more systematic approaches to developing social media communication strategies for the purpose of crime mitigation and reduction during disruption and change more generally. We suggest a framework for so doing.

Political Polarization Drives Online Conversations About COVID-19 in the United States

Since the outbreak in China in late 2019, the novel coronavirus (COVID-19) has spread around the world and has come to dominate online conversations. By linking 2.3 million Twitter users to locations within the United States, we study in aggregate how political characteristics of the locations affect the evolution of online discussions about COVID-19. We show that COVID-19 chatter in the US is largely shaped by political polarization. Partisanship correlates with sentiment toward government measures and the tendency to share health and prevention messaging. Cross-ideological interactions are modulated by user segregation and polarized network structure. We also observe a correlation between user engagement with topics related to public health and the varying impact of the disease outbreak in different US states. These findings may help inform policies both online and offline. Decision-makers may calibrate their use of online platforms to measure the effectiveness of public health campaigns, and to monitor the reception of national and state-level policies, by tracking in real-time discussions in a highly polarized social media ecosystem.