Vaccine-related Tweets Research Articles

Tracing the dynamics of misinformation and vaccine stance in Finland amid COVID-19

ABSTRACT This study analyzed 1,683,700 vaccine-related tweets in Finnish using FinBERT language model, Botometer, and BERTopic, from December 2019 to October 2022. A strong correlation was identified between Negative Stance towards Vaccines (NSV) and misinformation, with an upward trend over time and a significant role of malicious bot accounts. Topic modeling revealed persistent themes of vaccine skepticism and diverse misinformation themes adapting to debunking efforts. Approximately a third of tweets contained misinformation, irrespective of source reliability, leading to increased NSV. The study observed a gap in active countering of misinformation by authorities, suggesting proactive involvement by public authorities, public education for effective misinformation management and social media literacy.

Identification, analysis and prediction of valid and false information related to vaccines from Romanian tweets.

The online misinformation might undermine the vaccination efforts. Therefore, given the fact that no study specifically analyzed online vaccine related content written in Romanian, the main objective of the study was to detect and evaluate tweets related to vaccines and written in Romanian language. 1,400 Romanian vaccine related tweets were manually classified in true, neutral and fake information and analyzed based on wordcloud representations, a correlation analysis between the three classes and specific tweet characteristics and the validation of several predictive machine learning algorithms. The tweets annotated as misinformation showed specific word patterns and were liked and reshared more often as compared to the true and neutral ones. The validation of the machine learning algorithms yielded enhanced results in terms of Area Under the Receiver Operating Characteristic Curve Score (0.744-0.843) when evaluating the Support Vector Classifier. The predictive model estimates in a well calibrated manner the probability that a specific Twitter post is true, neutral or fake. The current study offers important insights regarding vaccine related online content written in an Eastern European language. Future studies must aim at building an online platform for rapid identification of vaccine misinformation and raising awareness for the general population.

Vaccine Hesitancy Hotspots in Africa: An Insight From Geotagged Twitter Posts

Many social media users express concerns about vaccines and their side effects on Twitter. These concerns lead to a compromise of confidence which brings about vaccine hesitancy. In Africa, vaccine hesitancy is a major challenge faced by health policymakers in the fight against COVID-19. Given that most tweets are geotagged, clustering them according to their sentiments could help identify locations that may likely experience vaccine hesitancy for health policy and planning. In this study, we collected 70 000 geotagged vaccine-related tweets in nine African countries, from December 2020 to February 2022. The tweets were classified into three sentiment classes—positive, negative, and neutral. The quality of the classification outputs was achieved using Naíve Bayes (NB), logistic regression (LR), support vector machines (SVMs), decision tree (DT), and K-nearest neighbor (KNN) machine learning classifiers. The LR achieved the highest accuracy of 71% with an average area under the curve of 85%. The point-based location technique was used to calculate the hotspots based on the locations of the classified tweets. Locations with green, red, and gray backgrounds on the map signify a hotspot for positive, negative, and neutral sentiments. The outcome of this research shows that discussions on social media can be analyzed to identify hotspots during a disease outbreak, which could inform health policy in planning and management of vaccine hesitancy in Africa.

Extracting factors associated with vaccination from Twitter data and mapping to behavioral models

ABSTRACT Social media platform, particularly Twitter, is a rich data source that allows monitoring of public opinions and attitudes toward vaccines.Established behavioral models like the 5C psychological antecedents model and the Health Belief Model (HBM) provide a well-structured framework for analyzing shifts in vaccine-related behavior. This study examines if the extracted data from Twitter contains valuable insights regarding public attitudes toward vaccines and can be mapped to two behavioral models. This study focuses on the Arab population, and a search was carried out on Twitter using: ’ تلقيحي OR تطعيم OR تطعيمات OR لقاح OR لقاحات’ for two years from January 2020 to January 2022. Then, BERTopicmodeling was applied, and several topics were extracted. Finally, the topics were manually mapped to the factors of the 5C model and HBM. 1,068,466 unique users posted 3,368,258 vaccine-related tweets in Arabic. Topic modeling generated 25 topics, which were mapped to the 15 factors of the 5C model and HBM. Among the users, 32.87%were male, and 18.06% were female. A significant 55.77% of the users were from the MENA (Middle East and North Africa) region. Twitter users were more inclined to accept vaccines when they trusted vaccine safety and effectiveness, but vaccine hesitancy increased due to conspiracy theories and misinformation. The association of topics with these theoretical frameworks reveals the availability and diversity of Twitter data that can predict behavioral change toward vaccines. It allows the preparation of timely and effective interventions for vaccination programs compared to traditional methods.

Geospatial vaccine misinformation risk on social media: Online insights from an English/Spanish natural language processing (NLP) analysis of vaccine-related tweets

BackgroundMisinformation is known to affect norms, attitudes, and intentions to engage with healthy behaviors. Evidence strongly supports that Spanish speakers may be particularly affected by misinformation and its outcomes, yet current insights into the scope and scale of misinformation is primarily ethnocentric, with greater emphasis on English-language design. ObjectiveThis study applies Natural Language Processing (NLP) to analyze a corpus of English/Spanish tweets about vaccines, broadly defined, for misinformation indicators. MethodsWe analyzed NEnglish = 247,140 and NSpanish = 104,445 tweets using Latent Dirichlet Allocation (LDA) topic models with Coherence score calculation (model fit) with a Mallet adjustment (topic optimization). We used informal coding to name computer-identified topics and compare misinformation scope and scale between languages. ResultsThe LDA analysis yielded a 12-topic solution for English and a 14-topic solution for Spanish. Both corpora contained overlapping misinformation, including uncertainty of research guiding policy recommendations or standing in support of antivax movements. However, the Spanish data were positioned in a global context, where misinformation was directed at government equity and disparate vaccine distribution. ConclusionOur findings support that misinformation is a global issue. However, misinformation may vary depending on culture and language. As such, tailored strategies to combat misinformation in digital planes are strongly encouraged.

Identifying drivers of COVID-19 vaccine sentiments for effective vaccination policy

The COVID-19 pandemic has had far-reaching consequences globally, including a significant loss of lives, escalating unemployment rates, economic instability, deteriorating mental well-being, social conflicts, and even political discord. Vaccination, recognized as a pivotal measure in mitigating the adverse effects of COVID-19, has evoked a diverse range of sentiments worldwide. In particular, numerous users on social media platforms have expressed concerns regarding vaccine availability and potential side effects. Therefore, it is imperative for governmental authorities and senior health policy strategists to gain insights into the public's perspectives on vaccine mandates in order to effectively implement their vaccination initiatives. Despite the critical importance of comprehending the underlying factors influencing COVID-19 vaccine sentiment, the existing literature offers limited research studies on this subject matter. This paper presents an innovative methodology that harnesses Twitter data to extract sentiment pertaining to COVID-19 vaccination through the utilization of Artificial Intelligence techniques such as sentiment analysis, entity detection, linear regression, and logistic regression. The proposed methodology was applied and tested on live Twitter feeds containing COVID-19 vaccine-related tweets, spanning from February 14, 2021, to April 2, 2023. Notably, this approach successfully processed tweets in 45 languages originating from over 100 countries, enabling users to select from an extensive scenario space of approximately 3.55 × 10249 possible scenarios. By selecting specific scenarios, the proposed methodology effectively identified numerous determinants contributing to vaccine sentiment across iOS, Android, and Windows platforms. In comparison to previous studies documented in the existing literature, the presented solution emerges as the most robust in detecting the fundamental drivers of vaccine sentiment and demonstrates the vaccination sentiments over a substantially longer period exceeding 24 months.

Unveiling Vaccine Hesitancy on Twitter: Analyzing Trends and Reasons during the Emergence of COVID-19 Delta and Omicron Variants.

Given the high amount of information available on social media, the paper explores the degree of vaccine hesitancy expressed in English tweets posted worldwide during two different one-month periods of time following the announcement regarding the discovery of new and highly contagious variants of COVID-19-Delta and Omicron. A total of 5,305,802 COVID-19 vaccine-related tweets have been extracted and analyzed using a transformer-based language model in order to detect tweets expressing vaccine hesitancy. The reasons behind vaccine hesitancy have been analyzed using a Latent Dirichlet Allocation approach. A comparison in terms of number of tweets and discussion topics is provided between the considered periods with the purpose of observing the differences both in quantity of tweets and the discussed discussion topics. Based on the extracted data, an increase in the proportion of hesitant tweets has been observed, from 4.31% during the period in which the Delta variant occurred to 11.22% in the Omicron case, accompanied by a diminishing in the number of reasons for not taking the vaccine, which calls into question the efficiency of the vaccination information campaigns. Considering the proposed approach, proper real-time monitoring can be conducted to better observe the evolution of the hesitant tweets and the COVID-19 vaccine hesitation reasons, allowing the decision-makers to conduct more appropriate information campaigns that better address the COVID-19 vaccine hesitancy.

Credibility of vaccine-related content on Twitter during COVID-19 pandemic.

During national COVID-19 vaccine campaigns, people continuously engaged on Twitter to receive updates on the latest public health information, and to discuss and share their experiences. During this time, the spread of misinformation was widespread, which threatened the uptake of vaccines. It is therefore critical to understand the reasons behind vaccine misinformation and strategies to mitigate it. The current research aimed to understand the content of misleading tweets and the characteristics of their corresponding accounts. We performed a machine learning approach to identify misinformation in vaccine-related tweets, and calculated the demographic, engagement metrics and bot-like activities of corresponding accounts. We found critical periods where high amounts of misinformation coincided with important vaccine announcements, such as emergency approvals of vaccines. Moreover, we found Asian countries had a lower percentage of misinformation shared compared to Europe and North America. Our results showed accounts spreading misinformation had an overall 10% greater likelihood of bot activity and 15% more astroturf bot activity than accounts spreading accurate information. Furthermore, we found that accounts spreading misinformation had five times fewer followers and three times fewer verified badges than fact-sharing accounts. The findings of this study may help authorities to develop strategies to fight COVID-19 vaccine misinformation and improve vaccine uptake.

Lexicon-based sentiment analysis to detect opinions and attitude towards COVID-19 vaccines on Twitter in Italy

The paper proposes a methodology based on Natural Language Processing (NLP) and Sentiment Analysis (SA) to get insights into sentiments and opinions toward COVID-19 vaccination in Italy. The studied dataset consists of vaccine-related tweets published in Italy from January 2021 to February 2022. In the considered period, 353,217 tweets have been analyzed, obtained after filtering 1,602,940 tweets with the word “vaccin”. A main novelty of the approach is the categorization of opinion holders in four classes, Common users, Media, Medicine, Politics, obtained by applying NLP tools, enhanced with large-scale domain-specific lexicons, on the short bios published by users themselves. Feature-based sentiment analysis is enriched with an Italian sentiment lexicon containing polarized words, expressing semantic orientation, and intensive words which give cues to identify the tone of voice of each user category. The results of the analysis highlighted an overall negative sentiment along all the considered periods, especially for the Common users, and a different attitude of opinion holders towards specific important events, such as deaths after vaccination, occurring in some days of the examined 14 months.

Correction: Content and Dynamics of Websites Shared Over Vaccine-Related Tweets in COVID-19 Conversations: Computational Analysis

[This corrects the article DOI: 10.2196/29127.].

“As Reliable as a Kalashnikov Rifle”: How Sputnik News Promotes Russian Vaccine Technologies in the Turkish Twittersphere

Established in 2014, SputnikTR (a localized version of Sputnik News) is the most popular pro-Russian media outlet active in Turkey. The news content published by SputnikTR’s Twitter account currently attracts the highest engagement rates among the international public broadcasters active in Turkey. SputnikTR’s official Twitter account has more followers (1M) than Sputnik News English (326K). This article argues that SputnikTR’s Twitter account is used to promote Russian vaccine technologies in Turkey. We believe that it is also a conduit for the dissemination of pro-Russian as well as anti-Western narratives to the Turkish online public. Using a computational methodology, we collected 2,782 vaccine-related tweets posted by SputnikTR’s Twitter account between April 2019 and April 2021. We deployed framing as well as critical discourse analysis to study the contents of our dataset. Our findings suggest that SputnikTR uses (a) disinformation as well as misinformation in vaccine-related news and (b) unethical communication techniques to maximize engagement with content posted on Twitter. Our findings are significant insofar as they are the first documented instances of Russian propaganda efforts on Turkish Twitter. These efforts seem to be focused on promoting the Russian vaccine while encouraging public hesitancy toward Western vaccine technologies.

Anti-vaccination attitude trends during the COVID-19 pandemic: A machine learning-based analysis of tweets.

Vaccine hesitancy has been ranked by the World Health Organization among the top 10 threats to global health. With a surge in misinformation and conspiracy theories against vaccination observed during the COVID-19 pandemic, attitudes toward vaccination may be worsening. This study investigates trends in anti-vaccination attitudes during the COVID-19 pandemic and within the United States, Canada, the United Kingdom, and Australia. Vaccine-related English tweets published between 1 January 2020 and 27 June 2021 were used. A deep learning model using a dynamic word embedding method, Bidirectional Encoder Representations from Transformers (BERTs), was developed to identify anti-vaccination tweets. The classifier achieved a micro F1 score of 0.92. Time series plots and country maps were used to examine vaccination attitudes globally and within countries. Among 9,352,509 tweets, 232,975 (2.49%) were identified as anti-vaccination tweets. The overall number of vaccine-related tweets increased sharply after the implementation of the first vaccination round since November 2020 (daily average of 6967 before vs. 31,757 tweets after 9/11/2020). The number of anti-vaccination tweets increased after conspiracy theories spread on social media. Percentages of anti-vaccination tweets were 3.45%, 2.74%, 2.46%, and 1.86% for the United States, the United Kingdom, Australia, and Canada, respectively. Strategies and information campaigns targeting vaccination misinformation may need to be specifically designed for regions with the highest anti-vaccination Twitter activity and when new vaccination campaigns are initiated.

Evolution of Public Opinion on COVID-19 Vaccination in Japan: Large-Scale Twitter Data Analysis.

Vaccines are promising tools to control the spread of COVID-19. An effective vaccination campaign requires government policies and community engagement, sharing experiences for social support, and voicing concerns about vaccine safety and efficiency. The increasing use of online social platforms allows us to trace large-scale communication and infer public opinion in real time. This study aimed to identify the main themes in COVID-19 vaccine-related discussions on Twitter in Japan and track how the popularity of the tweeted themes evolved during the vaccination campaign. Furthermore, we aimed to understand the impact of critical social events on the popularity of the themes. We collected more than 100 million vaccine-related tweets written in Japanese and posted by 8 million users (approximately 6.4% of the Japanese population) from January 1 to October 31, 2021. We used Latent Dirichlet Allocation to perform automated topic modeling of tweet text during the vaccination campaign. In addition, we performed an interrupted time series regression analysis to evaluate the impact of 4 critical social events on public opinion. We identified 15 topics grouped into the following 4 themes: (1) personal issue, (2) breaking news, (3) politics, and (4) conspiracy and humor. The evolution of the popularity of themes revealed a shift in public opinion, with initial sharing of attention over personal issues (individual aspect), collecting information from news (knowledge acquisition), and government criticism to focusing on personal issues. Our analysis showed that the Tokyo Olympic Games affected public opinion more than other critical events but not the course of vaccination. Public opinion about politics was significantly affected by various social events, positively shifting attention in the early stages of the vaccination campaign and negatively shifting attention later. This study showed a striking shift in public interest in Japan, with users splitting their attention over various themes early in the vaccination campaign and then focusing only on personal issues, as trust in vaccines and policies increased. An interrupted time series regression analysis showed that the vaccination rollout to the general population (under 65 years) increased the popularity of tweets about practical advice and personal vaccination experience, and the Tokyo Olympic Games disrupted public opinion but not the course of the vaccination campaign. The methodology developed here allowed us to monitor the evolution of public opinion and evaluate the impact of social events on public opinion, using large-scale Twitter data.

Futile or fertile? The effect of persuasive strategies on citizen engagement in COVID-19 vaccine-related tweets across six national health departments

National health departments across the globe have utilized persuasive strategies to promote COVID-19 vaccines through Twitter. However, the effectiveness of those strategies is unclear. This study thereby examined how national health departments deployed persuasive strategies to promote citizen engagement in COVID-19 vaccine-related tweets in six countries, including the UK, the US, Germany, Japan, South Korea, and India. Guided by the heuristic-systematic model and the health belief model, we found that national health departments differed significantly in the use of systematic-heuristic cues and health belief constructs in COVID-19 vaccine-related tweets. Generally, the provision of scientific information and appeals to anecdotes and fear positively, while appeals to bandwagon negatively, predicted citizen engagement. Messages about overcoming barriers and promoting vaccine benefits and self-efficacy positively affected engagement. Emphases of COVID-19 threats and cues to vaccinate demonstrated negative impacts. Importantly, health departments across countries often used futile or detrimental strategies in tweets. A locally adapted evidence-based approach for COVID-19 vaccination persuasion was discussed.

Anti-Vaccine Discourse on Social Media: An Exploratory Audit of Negative Tweets about Vaccines and Their Posters.

As the anti-vaccination movement is spreading around the world, this paper addresses the ever more urgent need for health professionals, communicators and policy-makers to grasp the nature of vaccine mis/disinformation on social media. A one-by-one coding of 4511 vaccine-related tweets posted from the UK in 2019 resulted in 334 anti-vaccine tweets. Our analysis shows that (a) anti-vaccine tweeters are quite active and widely networked users on their own; (b) anti-vaccine messages tend to focus on the "harmful" nature of vaccination, based mostly on personal experience, values and beliefs rather than hard facts; (c) anonymity does not make a difference to the types of posted anti-vaccine content, but does so in terms of the volume of such content. Communication initiatives against anti-vaccination should (a) work closely with technological platforms to tackle anonymous anti-vaccine tweets; (b) focus efforts on mis/disinformation in three major arears (in order of importance): the medical nature of vaccines, the belief that vaccination is a tool of manipulation and control for money and power, and the "freedom of health choice" discourse against mandatory vaccination; and (c) go beyond common factual measures-such as detecting, labelling or removing fake news-to address emotions induced by personal memories, values and beliefs.

Twitter’da COVID-19 aşılarına karşı kamu duyarlılığının çoğunluk oylama sınıflandırıcısı temelli makine öğrenmesi ile duygu analizi

Dünyada milyarlarca kullanıcısı bulunan sosyal medya platformlarının yükselişiyle birlikte bilginin yayılması her zamankinden daha kolay hale gelmiştir. COVID-19 pandemisi aşılar da dâhil olmak üzere birçok konunun tartışılmasında sosyal medya kullanımını artırmıştır. Bu çalışmanın amacı, Türkiye’de, özellikle sosyal medya kullanıcılarının COVID-19 aşılarına ilişkin tutumunu ve endişelerini daha iyi anlamak adına Twitter üzerinde elde edilen aşıyla ilgili tweetlerin makine öğrenmesi ile kamu duyarlılığını analiz etmektir. Bu amaç doğrultusunda çalışma altı farklı sınıflandırma görevinde kullanılan makine öğrenmesi algoritması karşılaştırılarak en yüksek doğruluk oranına sahip Destek Vektör Makinesi, XGBoost ve Rastgele Orman ile bir kolektif öğrenme yöntemi olan çoğunluk oylama yöntemi geliştirilmiştir. Çoğunluk oylama yöntemlerinde birisi olan Yumuşak Oylama yöntemi hem Sert Oylama yaklaşımdan hem de bireysel diğer altı makine öğrenmesi yaklaşımlarından daha yüksek başarı oranı ile %90,5 başarı oranına ulaşmıştır. En yüksek doğruluk oranına sahip olan Yumuşak Oylama yöntemi ile Twitter’dan elde edilen 153 güne ait 412.588 adet günlük tweet analiz edilerek sonuçlar raporlanmıştır. Çalışmanın bulguları son derece çarpıcı olup, diğer ülkeler üzerine yapılan çalışmalardan da farklılık göstermektedir. Bu çalışma bildiğimiz kadarıyla Türkiye’de COVID-19 aşılarına yönelik duygu analizi gerçekleştiren ilk çalışma olmakla birlikte sosyal medya üzerinden duygu analizi yaklaşımıyla COVID-19 aşılarına ilişkin duyarlılığı izlemek için değerli ve kolayca uygulanan bir araç olduğunu göstermektedir.

Emotions and Incivility in Vaccine Mandate Discourse: Natural Language Processing Insights.

BackgroundDespite vaccine availability, vaccine hesitancy has inhibited public health officials’ efforts to mitigate the COVID-19 pandemic in the United States. Although some US elected officials have responded by issuing vaccine mandates, others have amplified vaccine hesitancy by broadcasting messages that minimize vaccine efficacy. The politically polarized nature of COVID-19 information on social media has given rise to incivility, wherein health attitudes often hinge more on political ideology than science.ObjectiveTo the best of our knowledge, incivility has not been studied in the context of discourse regarding COVID-19 vaccines and mandates. Specifically, there is little focus on the psychological processes that elicit uncivil vaccine discourse and behaviors. Thus, we investigated 3 psychological processes theorized to predict discourse incivility—namely, anxiety, anger, and sadness.MethodsWe used 2 different natural language processing approaches: (1) the Linguistic Inquiry and Word Count computational tool and (2) the Google Perspective application programming interface (API) to analyze a data set of 8014 tweets containing terms related to COVID-19 vaccine mandates from September 14, 2021, to October 1, 2021. To collect the tweets, we used the Twitter API Tweet Downloader Tool (version 2). Subsequently, we filtered through a data set of 375,000 vaccine-related tweets using keywords to extract tweets explicitly focused on vaccine mandates. We relied on the Linguistic Inquiry and Word Count computational tool to measure the valence of linguistic anger, sadness, and anxiety in the tweets. To measure dimensions of post incivility, we used the Google Perspective API.ResultsThis study resolved discrepant operationalizations of incivility by introducing incivility as a multifaceted construct and explored the distinct emotional processes underlying 5 dimensions of discourse incivility. The findings revealed that 3 types of emotions—anxiety, anger, and sadness—were uniquely associated with dimensions of incivility (eg, toxicity, severe toxicity, insult, profanity, threat, and identity attacks). Specifically, the results showed that anger was significantly positively associated with all dimensions of incivility (all P<.001), whereas sadness was significantly positively related to threat (P=.04). Conversely, anxiety was significantly negatively associated with identity attack (P=.03) and profanity (P=.02).ConclusionsThe results suggest that our multidimensional approach to incivility is a promising alternative to understanding and intervening in the psychological processes underlying uncivil vaccine discourse. Understanding specific emotions that can increase or decrease incivility such as anxiety, anger, and sadness can enable researchers and public health professionals to develop effective interventions against uncivil vaccine discourse. Given the need for real-time monitoring and automated responses to the spread of health information and misinformation on the web, social media platforms can harness the Google Perspective API to offer users immediate, automated feedback when it detects that a comment is uncivil.

Automatically detecting and understanding the perception of COVID-19 vaccination: a middle east case study.

IntroductionThe development of COVID-19 vaccines has been a great relief in many countries that have been affected by the pandemic. As a result, many governments have made significant efforts to purchase and administer vaccines to their populations. However, accommodating such vaccines is typically confronted with people’s reluctance and fear. Like any other important event, COVID-19 vaccines have attracted people’s discussions on social media and impacted their opinions about vaccination.ObjectiveThe goal of this study is twofold: First, it conducts a sentiment analysis around COVID-19 vaccines by automatically analyzing Arabic users’ tweets. This analysis has been spread over time to better capture the changes in vaccine perceptions. This will provide us with some insights into the most popular and accepted vaccine(s) in the Arab countries, as well as the reasons behind people’s reluctance to take the vaccine. Second, it develops models to detect any vaccine-related tweets, to help with gathering all information related to people’s perception of the virus, and potentially detecting vaccine-related tweets that are not necessarily tagged with the virus’s main hashtags.MethodsArabic Tweets were collected by the authors, starting from January 1st, 2021, until April 20th, 2021. We deployed various Natural Language Processing (NLP) to distill our selected tweets. The curated dataset included in the analysis consisted of 1,098,376 unique tweets. To achieve the first goal, we designed state-of-the-art sentiment analysis techniques to extract knowledge related to the degree of acceptance of all existing vaccines and what are the main obstacles preventing the wide audience from accepting them. To achieve the second goal, we tackle the detection of vaccine-related tweets as a binary classification problem, where various Machine Learning (ML) models were designed to identify such tweets regardless of whether they use the vaccine hashtags or not.ResultsGenerally, we found that the highest positive sentiments were registered for Pfizer-BioNTech, followed by Sinopharm-BIBP and Oxford-AstraZeneca. In addition, we found that 38% of the overall tweets showed negative sentiment, and only 12% had a positive sentiment. It is important to note that the majority of the sentiments vary between neutral and negative, showing the lack of conviction of the importance of vaccination among the large majority of tweeters. This paper extracts the top concerns raised by the tweets and advocates for taking them into account when advertising for the vaccination. Regarding the identification of vaccine-related tweets, the Logistic Regression model scored the highest accuracy of 0.82. Our findings are concluded with implications for public health authorities and the scholarly community to take into account to improve the vaccine’s acceptance.

Understanding the vaccine stance of Italian tweets and addressing language changes through the COVID-19 pandemic: Development and validation of a machine learning model.

Social media is increasingly being used to express opinions and attitudes toward vaccines. The vaccine stance of social media posts can be classified in almost real-time using machine learning. We describe the use of a Transformer-based machine learning model for analyzing vaccine stance of Italian tweets, and demonstrate the need to address changes over time in vaccine-related language, through periodic model retraining. Vaccine-related tweets were collected through a platform developed for the European Joint Action on Vaccination. Two datasets were collected, the first between November 2019 and June 2020, the second from April to September 2021. The tweets were manually categorized by three independent annotators. After cleaning, the total dataset consisted of 1,736 tweets with 3 categories (promotional, neutral, and discouraging). The manually classified tweets were used to train and test various machine learning models. The model that classified the data most similarly to humans was XLM-Roberta-large, a multilingual version of the Transformer-based model RoBERTa. The model hyper-parameters were tuned and then the model ran five times. The fine-tuned model with the best F-score over the validation dataset was selected. Running the selected fine-tuned model on just the first test dataset resulted in an accuracy of 72.8% (F-score 0.713). Using this model on the second test dataset resulted in a 10% drop in accuracy to 62.1% (F-score 0.617), indicating that the model recognized a difference in language between the datasets. On the combined test datasets the accuracy was 70.1% (F-score 0.689). Retraining the model using data from the first and second datasets increased the accuracy over the second test dataset to 71.3% (F-score 0.713), a 9% improvement from when using just the first dataset for training. The accuracy over the first test dataset remained the same at 72.8% (F-score 0.721). The accuracy over the combined test datasets was then 72.4% (F-score 0.720), a 2% improvement. Through fine-tuning a machine-learning model on task-specific data, the accuracy achieved in categorizing tweets was close to that expected by a single human annotator. Regular training of machine-learning models with recent data is advisable to maximize accuracy.

Vaccine discourse during the onset of the COVID-19 pandemic: Topical structure and source patterns informing efforts to combat vaccine hesitancy.

Understanding public discourse about a COVID-19 vaccine in the early phase of the COVID-19 pandemic may provide key insights concerning vaccine hesitancy. However, few studies have investigated the communicative patterns in which Twitter users participate discursively in vaccine discussions. This study aims to investigate 1) the major topics that emerged from public conversation on Twitter concerning vaccines for COVID-19, 2) the topics that were emphasized in tweets with either positive or negative sentiment toward a COVID-19 vaccine, and 3) the type of online accounts in which tweets with either positive or negative sentiment were more likely to circulate. We randomly extracted a total of 349,979 COVID-19 vaccine-related tweets from the initial period of the pandemic. Out of 64,216 unique tweets, a total of 23,133 (36.03%) tweets were classified as positive and 14,051 (21.88%) as negative toward a COVID-19 vaccine. We conducted Structural Topic Modeling and Network Analysis to reveal the distinct topical structure and connection patterns that characterize positive and negative discourse toward a COVID-19 vaccine. Our STM analysis revealed the most prominent topic emerged on Twitter of a COVID-19 vaccine was "other infectious diseases", followed by "vaccine safety concerns", and "conspiracy theory." While the positive discourse demonstrated a broad range of topics such as "vaccine development", "vaccine effectiveness", and "safety test", negative discourse was more narrowly focused on topics such as "conspiracy theory" and "safety concerns." Beyond topical differences, positive discourse was more likely to interact with verified sources such as scientists/medical sources and the media/journalists, whereas negative discourse tended to interact with politicians and online influencers. Positive and negative discourse was not only structured around distinct topics but also circulated within different networks. Public health communicators need to address specific topics of public concern in varying information hubs based on audience segmentation, potentially increasing COVID-19 vaccine uptake.