Retweet Count Research Articles

An Innovative Way of Analyzing COVID Topics with LLM

In the aftermath of the COVID-19 pandemic, international landscapes have been profoundly reshaped, with shifts in political alliances, economic priorities, and socio-cultural norms. Such evolutions, reflected in the vast expanse of digital conversations, particularly on Twitter, necessitate advanced tools for analysis given their impact on policy and strategy. In this context, the presented study underscores the indispensability of Artificial Intelligence (AI) in discerning intricate patterns from voluminous and multifaceted Twitter data on COVID-19. Through an innovative methodology leveraging AI modalities such as language detection, sentiment analysis, topic analysis, Large Language Model (LLM), regression, clustering, this study distills textual features from 152,070 multilingual tweets across 58 languages, spanning 645 days from 15 July 2021 to 20 April 2023. Our analyses, automatically identify five pivotal COVID-19 discussion topics and expound on four critical factors—tweet language, retweet count, and positive and negative sentiments—that significantly influence these conversations. In essence, the paper's contributions lie in: 1) unveiling an AI-centric autonomous methodology for deep insights into COVID-19 discussions; 2) empirically validating this approach using a diverse, multilingual dataset that resulted in five key discussion areas; and 3) presenting 52 nuanced AI-generated observations that detail factors influencing these discussions. Comparative literature suggests that our approach offers unparalleled depth in AI-driven analytics related to COVID-19 discourse. In summary, this paper underline the pressing need to harness the power of AI-based Tweet analytics as an indispensable tool in formulating strategic decisions pertaining to disaster responses.

Quick-and-Wide Propagation of Disaster Tweets: Why It Matters and How to Measure It

As dynamic, non-routine disaster events appear and disappear in a short period, quickly and widely disseminating emergency warnings and alerts to the public is a critical communication goal for emergency management officers and citizen journalists. Twitter has become a prominent communication medium to achieve such a communication goal, attracting academic scholars, government agencies, and industry practitioners. However, most academic research and government reports have relied on either the retweet count or retweet time to understand factors affecting the propagation of disaster information. We argue that neither measure is sufficient to reflect the extent to which disaster tweets are quickly and widely propagated. Hence, we propose a new measure called the Propagation Index (PI) to better evaluate quick and wide propagation of disaster information. Using two Twitter datasets, we empirically examine the proposed measure in association with the Twitter features and compare its statistical results with those of the retweet count and retweet time. Our findings suggest that the PI more accurately captures how quickly and widely disaster tweets are propagated than the other measures. Therefore, this study contributes to academia, governments, and industry to improve their understanding of the propagation of disaster information on Twitter.

Negativity bias in the spread of voter fraud conspiracy theory tweets during the 2020 US election

During the 2020 US presidential election, conspiracy theories about large-scale voter fraud were widely circulated on social media platforms. Given their scale, persistence, and impact, it is critically important to understand the mechanisms that caused these theories to spread. The aim of this preregistered study was to investigate whether retweet frequencies among proponents of voter fraud conspiracy theories on Twitter during the 2020 US election are consistent with frequency bias and/or content bias. To do this, we conducted generative inference using an agent-based model of cultural transmission on Twitter and the VoterFraud2020 dataset. The results show that the observed retweet distribution is consistent with a strong content bias causing users to preferentially retweet tweets with negative emotional valence. Frequency information appears to be largely irrelevant to future retweet count. Follower count strongly predicts retweet count in a simpler linear model but does not appear to drive the overall retweet distribution after temporal dynamics are accounted for. Future studies could apply our methodology in a comparative framework to assess whether content bias for emotional valence in conspiracy theory messages differs from other forms of information on social media.

Seeking and Providing Social Support on Twitter for Trauma and Distress During the COVID-19 Pandemic: Content and Sentiment Analysis.

The COVID-19 pandemic can be recognized as a traumatic event that led to stressors, resulting in trauma or distress among the general population. Social support is vital in the management of these stressors, especially during a traumatic event, such as the COVID-19 pandemic. Because of the limited face-to-face interactions enforced by physical distancing regulations during the pandemic, people sought solace on social media platforms to connect with, and receive support from, one another. Hence, it is crucial to investigate the ways in which people seek and offer support on social media for mental health management. The research aimed to examine the types of social support (eg, emotional, informational, instrumental, and appraisal) sought and provided for trauma or distress on Twitter during the COVID-19 pandemic. In addition, this study aimed to gain insight into the difficulties and concerns of people during the pandemic by identifying the associations between terms representing the topics of interest related to trauma or distress and their corresponding sentiments. The study methods included content analysis to investigate the type of social support people sought for trauma or distress during the pandemic. Sentiment analysis was also performed to track the negative and positive sentiment tweets posted between January 1, 2020, and March 15, 2021. Association rule mining was used to uncover associations between terms and sentiments in tweets. In addition, the research used Kruskal-Wallis and Mann-Whitney U tests to determine whether the retweet count and like count varied based on the social support type. Most Twitter users who indicated trauma or distress sought emotional support. Regarding sentiment, Twitter users mostly posted negative sentiment tweets, particularly in January 2021. An intriguing observation was that wearing masks could trigger and exacerbate trauma or distress. The results revealed that people mostly sought and provided emotional support on Twitter regarding difficulties with wearing masks, mental health status, financial hardships, and treatment methods for trauma or distress. In addition, tweets regarding emotional support received the most endorsements from other users, highlighting the critical role of social support in fostering a sense of community and reducing the feelings of isolation during the pandemic. This study demonstrates the potential of social media as a platform to exchange social support during challenging times and to identify the specific concerns (eg, wearing masks and exacerbated symptoms) of individuals with self-reported trauma or distress. The findings provide insights into the types of support that were most beneficial for those struggling with trauma or distress during the pandemic and may inform policy makers and health organizations regarding better practices for pandemic response and special considerations for groups with a history of trauma or distress.

A Multimodal Ensemble Machine Learning Approach to COVID-19 Misinformation Detection in Twitter

The emergence of social media platforms has unquestionably altered the manner in which people ingest information, with tweets now functioning as the primary source for news and other types of content. However, the proliferation of false news on these platforms has become a major concern, as it poses a severe threat to both individuals and society as a whole. Consequently, it is crucial to develop efficient methods for detecting false news in tweets. This study presents a novel hybrid approach that integrates the textual content of tweets with auxiliary features to detect false news. Our approach uses a pre-trained transformer-based language model, COVID-twitter-BERT to encode the text content of tweets into a dense representation that captures their meaning. The auxiliary features, such as sentiment score, credibility score, engagement score, average retweet count, average favourite count, and average followers of followers, are fed into a stacking classifier-based model to predict the trustworthiness score of the tweet. By combining the predictions of both models, we demonstrate that our approach outperforms baseline methods, emphasising the significance of utilising both text content and auxiliary features for Twitter false news detection. Our research considerably advances the field of detecting false news by demonstrating the effectiveness of integrating transformer-based language models and machine learning models for this task. Our findings provide valuable insights for improving the detection of false news on social media.

Combating Misinformation by Sharing the Truth: a Study on the Spread of Fact-Checks on Social Media.

Misinformation on social media has become a horrendous problem in our society. Fact-checks on information often fall behind the diffusion of misinformation, which can lead to negative impacts on society. This research studies how different factors may affect the spread of fact-checks over the internet. We collected a dataset of fact-checks in a six-month period and analyzed how they spread on Twitter. The spread of fact-checks is measured by the total retweet count. The factors/variables include the truthfulness rating, topic of information, source credibility, etc. The research identifies truthfulness rating as a significant factor: conclusive fact-checks (either true or false) tend to be shared more than others. In addition, the source credibility, political leaning, and the sharing count also affect the spread of fact-checks. The findings of this research provide practical insights into accelerating the spread of the truth in the battle against misinformation online.

Role of twitter user profile features in retweet prediction for big data streams.

To study the various factors influencing the process of information sharing on Twitter is a very active research area. This paper aims to explore the impact of numerical features extracted from user profiles in retweet prediction from the real-time raw feed of tweets. The originality of this work comes from the fact that the proposed model is based on simple numerical features with the least computational complexity, which is a scalable solution for big data analysis. This research work proposes three new features from the tweet author profile to capture the unique behavioral pattern of the user, namely “Author total activity”, “Author total activity per year”, and “Author tweets per year”. The features set is tested on a dataset of 100 million random tweets collected through Twitter API. The binary labels regression gave an accuracy of 0.98 for user-profile features and gave an accuracy of 0.99 when combined with tweet content features. The regression analysis to predict the retweet count gave an R-squared value of 0.98 with combined features. The multi-label classification gave an accuracy of 0.9 for combined features and 0.89 for user-profile features. The user profile features performed better than tweet content features and performed even better when combined. This model is suitable for near real-time analysis of live streaming data coming through Twitter API and provides a baseline pattern of user behavior based on numerical features available from user profiles only.

People lie, actions Don't! Modeling infodemic proliferation predictors among social media users

Social media is interactive, and interaction brings misinformation. With the growing amount of user-generated data, fake news on online platforms has become much more frequent since the arrival of social networks. Now and then, an event occurs and becomes the topic of discussion, generating and propagating false information. Existing literature studying fake news elaborates primarily on fake news classification models. Approaches exploring fake news characteristics to distinguish it from real news are minimal. Not much research has focused on statistical testing and generating new factor discoveries. This study assumes fifteen hypotheses to identify factors exhibiting a relationship with fake news. We perform the experiments on two real-world COVID-19 datasets using qualitative and quantitative testing methods. We determine the impact of conditional effects among sentiment, gender, and media usage. This study concludes that sentiment polarity and gender can significantly identify fake news. Dependence on the presence of visual media is, however, inconclusive. Additionally, Twitter-specific user engagement factors like followers count, friends count, favorite count, and retweet count significantly differ in fake and real news. Though, the contribution of status count is currently disputed. This study identifies practical factors to be conjunctly utilized in developing fake news detection algorithms.

Predicting the Audience Size of a Tweet

We study information dissemination in Twitter. We present an analysis of two important characteristics of so called retweet cascades: retweet count and show count, i.e., number of users that receive the tweet in their feed. We show that these two measures behave differently. We describe three models that are aimed to predict the audience size of a tweet: first one utilizes only the data available at the moment of the initial tweet, the second one utilizes the spread of the cascade up to some moment while the third one is an online prediction.

Twitter chirps for Syrian people: Sentiment analysis of tweets related to Syria Chemical Attack

PurposeThe sentiment analysis of tweets provides information about peoples’ attitudes and perceptions towards an event. The current study showcases the role of Twitter in a crisis by analyzing the nature of tweets and the sentiments expressed by the Twitter-sphere during and after the “Khan Shaykhun Syria Chemical Attack." MethodologyA total of 13,156 tweets posted in English on Twitter during the first 27 days of the attack were downloaded and considered for the study. The content analysis of the tweets was done manually, and accordingly, the sentiments of the tweets were highlighted through eight broader categories. Furthermore, to visualize the positive, negative, and neutral sentiments of the tweets, the Orange DataMining Software, a powerful toolkit for machine learning, data mining, and data visualization, was used. VOSviewer (a software tool used for creating maps based on network data and for visualizing and exploring the maps) was also used to visualize the word frequency of the tweets. FindingsTwitter is primarily used for situational awareness and acts as an emotional, social support system by sharing sentiments. 35.71% of the tweets are associated with "sharing news and information" , with just 2.12% "supporting the government". People mostly retweet the tweets that “criticize the government,” with an average retweet count of 15.84, followed by the ones “evincing emotions” (12.21). However, tweets that “raise questions” (3.32) and “provide suggestions” (2.51) fail to gain the attention of too many tweeter users, thus having less impact. People mostly like the tweets that “supportgovernment” and “evince emotions,” with such tweets on an average receiving 9.89 and 8.37 likes, respectively. Individuals post a large number of tweets (10,137; 77.05%), followed by news channels (1157; 8.79%) and organizations of varied nature (950; 7.22%). However, 912 (6.93%) tweets are posted by users of anonymous nature. Text and text with images form most tweets contributing to 8061 (61.27%) and 3137 (23.84%) of the total tweet count. However, none of the tweets contain video only, and just 3 (0.02%) tweets embed only images. Text-video and text-image formats are highly re-tweeted and liked. It is evident that 53.70% of the tweets (n = 7065) reflect negative sentiments, while 12.67% (n = 1667) emulate positive sentiments and, 33.63% (n = 4424) showcase a neutral perception about the attack. One can visualize the U.S.A. among the top tweeting countries with the highest percentage of positive sentiments, followed by Canada and Israel. Turkey outscores all the countries in terms of negative tweets, followed by Syria and U.K. However, in terms of neutral tweets, Germany ranks first, followed by Iran and Canada. The tweets pour mainly for the first few days, indicating the concern of users for the victims. Later on, a declining trend of tweets is witnessed. “idlib”, “Syria”, “Syria chemical attack,” and “Assad” are the leading words used more than a thousand times in the tweets. Research implicationsThe current study adds to the growing body of knowledge to the existing literature on Twitter and its use to narrowcast situational awareness during crisis episodes. One of the implications of the study is that the news agencies highly exploit the sharing side of Twitter during disasters by communicating real-time and unique information, create situational awareness, and connect to the digital audience. Twitter acts as an emotional outlet that facilitates the mining of condensed varied reactions towards an event to frame disaster response strategies and provides a sociological understanding of social media use during the crisis by the victims and viewers. OriginalityThe study represents the sentiments of the Twitter-sphere towards the “Syria Chemical Attack."

Leveraging Twitter data to analyze the virality of Covid-19 tweets: a text mining approach

ABSTRACT As the novel coronavirus spreads across the world, work, pleasure, entertainment, social interactions, and meetings have shifted online. The conversations on social media have spiked, and given the uncertainties and new policies, COVID-19 remains the trending topic on all such platforms, including Twitter. This research explores the factors that affect COVID-19 content-sharing by Twitter users. The analysis was conducted using 57,000 plus tweets that mentioned COVID-19 and related keywords. The tweets were subjected to the Natural Language Processing (NLP) techniques like Topic modelling, Named Entity-Relationship, Emotion & Sentiment analysis, and Linguistic feature extraction. These methods generated features that could help explain the retweet count of the tweets. The results indicate that tweets with named entities (person, organisation, and location), expression of negative emotions (anger, disgust, fear, and sadness), reference to mental health, optimistic content, and greater length have higher chances of being shared (retweeted). On the other hand, tweets with more hashtags and user mentions are less likely to be shared.

Qualitative analysis of visual risk communication on twitter during the Covid-19 pandemic

BackgroundThe Covid-19 pandemic is characterized by uncertainty and constant change, forcing governments and health authorities to ramp up risk communication efforts. Consequently, visuality and social media platforms like Twitter have come to play a vital role in disseminating prevention messages widely. Yet to date, only little is known about what characterizes visual risk communication during the Covid-19 pandemic. To address this gap in the literature, this study’s objective was to determine how visual risk communication was used on Twitter to promote the World Health Organisations (WHO) recommended preventative behaviours and how this communication changed over time.MethodsWe sourced Twitter’s 500 most retweeted Covid-19 messages for each month from January–October 2020 using Crowdbreaks. For inclusion, tweets had to have visuals, be in English, come from verified accounts, and contain one of the keywords ‘covid19’, ‘coronavirus’, ‘corona’, or ‘covid’. Following a retrospective approach, we then performed a qualitative content analysis of the 616 tweets meeting inclusion criteria.ResultsOur results show communication dynamics changed over the course of the pandemic. At the start, most retweeted preventative messages came from the media and health and government institutions, but overall, personal accounts with many followers (51.3%) predominated, and their tweets had the highest spread (10.0%, i.e., retweet count divided by followers). Messages used mostly photographs and images were found to be rich with information. 78.1% of Tweets contained 1–2 preventative messages, whereby ‘stay home’ and ‘wear a mask’ frequented most. Although more tweets used health loss framing, health gain messages spread more.ConclusionOur findings can inform the didactics of future crisis communication. The results underscore the value of engaging individuals, particularly influencers, as advocates to spread health risk messages and promote solidarity. Further, our findings on the visual characteristic of the most retweeted tweets highlight factors that health and government organisations should consider when creating visual health messages for Twitter. However, that more tweets used the emotive medium of photographs often combined with health loss framing raises concerns about persuasive tactics. More research is needed to understand the implications of framing and its impact on public perceptions and behaviours.

Sentiment Analysis, Tweet Analysis and Visualization on Big Data Using Apache Spark and Hadoop

Big Data lets companies generate new growth opportunities and whole new types of industries that can continually change and mushroom data from the market that can merge and evaluate. In today’s world, Apache Spark and Hadoop are the most popular and effective tools used to conduct big data analysis. This paper aims to perform two types of analysis-Sentiment Analysis of tweets and Tweet Analysis. For conducting sentiment analysis, we used a dataset created by Stanford University which contains 4 fields – ID of the tweet, Sentiment of the tweet, Sentiment Source and the Tweet itself. Two machine learning algorithms performance are compared these are Logistic Regression and Random Forest Classifier and observe which algorithm provides better accuracy and why. A data visualization tool -Tableau is used to view the sentiments of tweets by filtering the text of tweets by words and hashtags. For tweet analysis, we use Python and a library called Tweepy to download tweets from the Twitter API and perform analysis using another data visualization tool called Power BI. We conduct analysis on the basis of three parameters – user followers count, retweet count and favourite count. Finally, we draw a comparison on the performance and user-friendly nature of both data visualization tools – Power BI and Tableau.

Ranking influencers of social networks by semantic kernels and sentiment information

Inspired by the importance of social media, a Social Network Opinion Leaders (SNOL) system has been proposed in this paper. The purpose of this system is to identify topic-based opinion leaders of social media. In order to accomplish this goal, several steps have been taken, such as data collection, data processing, data analysis, data classification, ranking of topic-based opinion leaders, and evaluation. The SNOL system has two main parts. In the first part, collected tweets are classified by semantic kernels for topic-based analysis. In the second part, leadership scores are given to each user in the network according to topic modeling and user modeling results. Leadership scores are then calculated with the formula generated and opinion leaders are determined for each category. Experiments are performed on data gathered from Twitter including 17,234,924 tweets from 38,727 users. The evaluation of opinion leader detection is a difficult job since there is no standard method for identifying opinion leaders. Therefore, the evaluation of the results of this study has been done using two different methods, retweet count and spread score, to prove that the suggested methodology outperforms the PageRank algorithm. The results have also been evaluated considering the user-topic sentiment correlation of the retrieved lists. Furthermore, SNOL has been compared against some opinion leader detection methods previously presented in the literature. The experimental results show that SNOL generates remarkably higher performance than the PageRank algorithm and other existing algorithms in the literature for nearly all topics and all selected top N opinion leaders.

“I Definitely Did Not Report It When I Was Raped . . . #WeBelieveChristine #MeToo”: A Content Analysis of Disclosures of Sexual Assault on Twitter

The “Me Too” movement, founded by activist Tarana Burke, began in 2006, before Twitter became the viral platform for political activism and news media that it is known for today. The reemergence of the Me Too movement on Twitter in 2017 sparked a widespread focus on the societal issue of sexual misconduct. This study examines sexual assault disclosures as an aspect of such misconduct through the context of the Me Too movement on Twitter. Through the use of content analysis, online disclosures of sexual assault ( N = 1,459) are examined for variations of sexual explicitness and attainment of social functions per the functional theory of self-disclosure. Specifically, this study explores associations between Twitter network structure and (1) levels of sexual explicitness and (2) retweet count. Through manual coding and statistical analysis, the study finds associations between sexual explicitness of online disclosures and retweet count as well as associations between network structure and sexual explicitness of disclosures. The study shifts the focus of disclosure discourse from dyadic communication to the contemporary context of networked social media. Implications for theory and practice are discussed, which include, but are not limited to, the exploration of associations between disclosing and catharsis, disclosing that may be characterized as neutral or conflicted, and disclosing sexual assault without being sexually explicit.

Exploring Casual COVID-19 Data Visualizations on Twitter: Topics and Challenges

Social networking sites such as Twitter have been a popular choice for people to express their opinions, report real-life events, and provide a perspective on what is happening around the world. In the outbreak of the COVID-19 pandemic, people have used Twitter to spontaneously share data visualizations from news outlets and government agencies and to post casual data visualizations that they individually crafted. We conducted a Twitter crawl of 5409 visualizations (from the period between 14 April 2020 and 9 May 2020) to capture what people are posting. Our study explores what people are posting, what they retweet the most, and the challenges that may arise when interpreting COVID-19 data visualization on Twitter. Our findings show that multiple factors, such as the source of the data, who created the chart (individual vs. organization), the type of visualization, and the variables on the chart influence the retweet count of the original post. We identify and discuss five challenges that arise when interpreting these casual data visualizations, and discuss recommendations that should be considered by Twitter users while designing COVID-19 data visualizations to facilitate data interpretation and to avoid the spread of misconceptions and confusion.

Understanding the uncertainty of disaster tweets and its effect on retweeting: The perspectives of uncertainty reduction theory and information entropy

AbstractThe rapid and wide dissemination of up‐to‐date, localized information is a central issue during disasters. Being attributed to the original 140‐character length, Twitter provides its users with quick‐posting and easy‐forwarding features that facilitate the timely dissemination of warnings and alerts. However, a concern arises with respect to the terseness of tweets that restricts the amount of information conveyed in a tweet and thus increases a tweetʼs uncertainty. We tackle such concerns by proposing entropy as a measure for a tweetʼs uncertainty. Based on the perspectives of Uncertainty Reduction Theory (URT), we theorize that the more uncertain information of a disaster tweet, the higher the entropy, which will lead to a lower retweet count. By leveraging the statistical and predictive analyses, we provide evidence supporting that entropy validly and reliably assesses the uncertainty of a tweet. This study contributes to improving our understanding of information propagation on Twitter during disasters. Academically, we offer a new variable of entropy to measure a tweetʼs uncertainty, an important factor influencing disaster tweetsʼ retweeting. Entropy plays a critical role to better comprehend URLs and emoticons as a means to convey information. Practically, this research suggests a set of guidelines for effectively crafting disaster messages on Twitter.

Activeness of Syrian refugee crisis: an analysis of tweets

In this paper, we propose and apply a method to analyze the activeness of an event based on related tweets. The method characterizes and measures activeness of an event by a set of indicators. The indicators proposed in this paper are original tweet count, retweet count, follower count, positive sentiment, negative sentiment, daily change in users count, and sparseness of user community. We present procedures to compute the last two indicators. All indicators collectively are used to determine the activeness of an event. This approach is used to analyze the Syrian-refugee-crisis-related tweets. Its generality is demonstrated by applying it to analyze “immigration”-related tweets.

A Hierarchical Model of Nonhomogeneous Poisson Processes for Twitter Retweets

We present a hierarchical model of nonhomogeneous Poisson processes (NHPP) for information diffusion on online social media, in particular Twitter retweets. The retweets of each original tweet are modelled by a NHPP, for which the intensity function is a product of time-decaying components and another component that depends on the follower count of the original tweet author. The latter allows us to explain or predict the ultimate retweet count by a network centrality-related covariate. The inference algorithm enables the Bayes factor to be computed, to facilitate model selection. Finally, the model is applied to the retweet datasets of two hashtags. Supplementary materials for this article, including a standardized description of the materials available for reproducing the work, are available as an online supplement

Text mining and determinants of sentiments: Twitter social media usage by traditional media houses in Uganda

Unstructured data generated from sources such as the social media and traditional text documents are increasing and form a larger proportion of unanalysed data especially in the developing countries. In this study, we analysed data received from the major print and non-print media houses in Uganda through the Twitter platform to generate non-trivial knowledge by using text mining analytics. We also explored the determinants of derived sentiments in Twitter messaging. The results show that sentiments generated from tweets derived from the main print media houses (Daily Monitor and New Vision) were positively correlated, so were the sentiments from the non-print media (NBS TV and NTV) for the study period. Most of the sentiments on topics of security, politics and economics were found to be negative, while those on sports were positive. Furthermore, the tweet sentiment statistical logistic model revealed that negative sentiments were determined by the retweet status, retweet count and source of the tweets. Moreover, the positive sentiments were determined by the topic of discussion, type of media house and other sources of tweets (p<0.05). Therefore, we recommend further extensions on the predictive statistical models to classify sentiments from social media based on the concept of big data analytics.