A Machine Learning Approach for COVID-19 Tweet Classification

Vaccine-related information is awash on social media platforms like Twitter and Facebook. One party supports vaccination, while the other opposes vaccination and promotes misconceptions and misleading information about the risks of vaccination. The analysis of social media posts can give significant information into public opinion on vaccines, which can help government authorities in decision-making.This paper describes the dataset used in the shared task, and compares the performance of different classification that are provax, antivax and last neutral for identifying effective tweets related to Covid vaccines.We experimented with a classification-based approach. Our experiment shows that SVM classification performs well in order to effiective post.We're going to do this because vaccination is an important step for Covid19 so people can easily fix the news about the vaccine and grab their own slot and symptom detection is also playing a important part to arrest the spread of disease. © 2022 Copyright for this paper by its authors.

COVID-19 Fake News Detector

COVID-19 pandemic has been impacting people's everyday life for more than two years. With the fast spreading of online communication and social media platforms, the number of fake news related to COVID-19 is in a rapid growth and propagates misleading information to the public. To tackle this challenge and stop the spreading of fake news, this project proposes to build an online software detector specifically for COVID-19 news to classify whether the news is trustworthy. Specifically, as it is difficult to train a generic model for all domains, a base model is developed and fine-tuned to adapt the specific domain context. In addition, a data collection mechanism is developed to get latest COVID-19 news data and to keep the model fresh. We then conducted performance comparisons among different models using traditional machine learning techniques, ensemble machine learning, and the state-of-the-art deep learning mechanism. The most effective model is deployed to our online website for COVID-19 related fake news detection. © 2023 IEEE.

Detect misinformation of COVID-19 using deep learning: A comparative study based on word embedding

Since its emergence in December 2019, there have been numerous news of COVID-19 pandemic shared on social media, which contain information from both reliable and unreliable medical sources. News and misleading information spread quickly on social media, which can lead to anxiety, unwanted exposure to medical remedies, etc. Rapid detection of fake news can reduce their spread. In this paper, we aim to create an intelligent system to detect misleading information about COVID-19 using deep learning techniques based on LSTM and BLSTM architectures. Data used to construct the DL models are text type and need to be transformed to numbers. We test, in this paper the efficiency of three vectorization techniques: Bag of words, Word2Vec and Bert. The experimental study showed that the best performance was given by LSTM model with BERT by achieving an accuracy of 91% of the test set. © 2023 IEEE.

Learning from Machines? Social Bots Influence on COVID-19 Vaccination-Related Discussions: 2021 in Review

The World Health Organization defines vaccine hesitancy as a delay in acceptance or refusal of vaccination despite the availability of vaccination services. Vaccine hesitancy contributes to lower rates of vaccination in a population and delayed vaccine coverage. A large number of COVID-19 vaccines have been administered worldwide against COVID-19. Due to concerns people have about COVID-19 vaccine adverse events, a significant proportion of people exhibit hesitancy towards the vaccines. These are often prompted by information and misinformation spread through social media conversation, which is not driven exclusively by genuine human-run accounts. Social bots have been shown to be very active during the pandemic participating in discussions about vaccines, including the spread of conflicting and misleading information. Using a novel ensemble technique, we sought to identify and describe the involvement of social bots in COVID-19 vaccination-related discussions on Twitter and how this could have influenced sentiments and hesitancies about COVID-19 vaccines. We included tweets from January to December 2021 to present a whole year's analysis in relation to the vaccines. Unique usernames from these posts were passed to Botometer and Tweetbotornot, programs that review Twitter accounts, to detect a broad range of social bots using a scoring system. A domain-oriented transfer learning technique is applied by finetuning the CT-BERT V2 model to detect the influence of social bots on COVID-19 vaccine sentiments. We computed the ratio of sentiment transmission from bots-to-human, human-to-human, human-to-bots, and bots-to-bots. BERTopic was used to extract the topics of discussion to identify the amplified or transferred hesitancies. Social bots' participation in online discussions noticeably influenced human sentiments and hesitancies about COVID-19 vaccination. A major portion of sentiments transferred from bot to human during the period of study appeared to amplify or transfer hesitancies regarding COVID-19 vaccination. © 2023 ACM.

Health-related misinformation sharing on social media in Thailand: A case study during the Covid-19 pandemic

Misinformation affects people because it can convince them to believe in how to respond to uncertain situations. During the COVID-19 pandemic, a number of misinformation or fake news were distributed on social media, in Thailand. This research aimed to study attributes and causes of Health-Related Misinformation Sharing in Thailand on social media during the COVID-19 pandemic. Dataset used in this study was collected from the Anti-Fake News Center, the Thai government fact-checking website certified by the International Fact-Checking Network (IFCN). In-depth interviews based on qualitative research technique were also conducted to identify the causes of the transmission of false health news on Thai social media by applying the rumors transmission concept during times of crisis and the theory of Uses and Gratifications. The findings showed five main themes of fake news: conspiracy theories, pseudoscience, fake advertisements, inaccurate information, and misleading information. These elements may establish a conceptual framework for finding the root cause of misinformation spreads during crises. Factors that affect how psychological information was presented and shared are anxiety, insecurity, and uncertainty during the crisis. However, belief is not the only justification for sharing this information because some social media users have shared unverified and no evidence information for personal purposes. The Uses and Gratifications theories are found relevant. This study is intended to broaden the reach of disseminating misleading information as much as possible to lessen the effect of detrimental health fake news on Internet news consumers. © 2022 International Consortium for Electronic Business. All rights reserved.

Evolution of Intent and Social Influence Networks and Their Significance in Detecting COVID-19 Disinformation Actors on Social Media

Online disinformation actors are those individuals or bots who disseminate false or misleading information over social media, with the intent to sway public opinion in the information domain towards harmful social outcomes. Quantification of the degree to which users post or respond intentionally versus under social influence, remains a challenge, as individuals or organizations operating the profile are foreshadowed by their online persona. However, social influence has been shown to be measurable in the paradigm of information theory. In this paper, we introduce an information theoretic measure to quantify social media user intent, and then investigate the corroboration of intent with evolution of the social network and detection of disinformation actors related to COVID-19 discussions on Twitter. Our measurement of user intent utilizes an existing time series analysis technique for estimation of social influence using transfer entropy among the considered users. We have analyzed 4.7 million tweets originating from several countries of interest, during a 5 month period when the arrival of the first dose of COVID vaccinations were announced. Our key findings include evidence that: (i) a significant correspondence between intent and social influence;(ii) ranking over users by intent and social influence is unstable over time with evidence of shifts in the hierarchical structure;and (iii) both user intent and social influence are important when distinguishing disinformation actors from non-disinformation actors. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.

A Comparative Study on COVID-19 Fake News Detection Using Different Transformer Based Models

The rapid advancement of social networks and the convenience of internet availability have accelerated the rampant spread of false news and rumors on social media sites. Amid the COVID-19 epidemic, this misleading information has aggravated the situation by putting people's mental and physical lives in danger. To limit the spread of such inaccuracies, identifying the fake news from online platforms could be the first and foremost step. In this research, the authors have conducted a comparative analysis by implementing five transformer-based models such as BERT, BERT without LSTM, ALBERT, RoBERTa, and a Hybrid of BERT & ALBERT in order to detect the fraudulent news of COVID-19 from the internet. COVID-19 Fake News Dataset has been used for training and testing the models. Among all these models, the RoBERTa model has performed better than other models by obtaining an F1 score of 0.98 in both real and fake classes. © 2022 IEEE.

Datasets and Approaches of COVID-19 Misinformation Detection: A Survey

Social media has become a primary source of news, providing a fertile environment for spreading misinformation. Since the outbreak of the COVID-19 pandemic, misleading information related to COVID-19 has been spreading rapidly and widely on social media. Several conspiracy theories have emerged regarding the origin of the COVID-19, potential treatments, and vaccines posing a real threat to the public health of people. Fake news that promotes vaccine hesitancy might jeopardize achieving the levels of vaccination needed to reach herd immunity and end the pandemic. The need for automatic tools that detect COVID-19 related misinformation has encouraged researchers to propose several Machine learning (ML) and Deep Learning (DL). Many datasets have been released since the start of the pandemic, aiming to assess the performance of misinformation detection methods. This survey reviews the datasets that have been released to analyze the related to COVID-19 in general and COVID-19 misinformation detection in particular released in Arabic, English, and other languages. We also provide an overview of the different methods used to detect COVID-19 fake news. In this paper, the terms 'misleading information', 'misinformation', and 'fake news' are used interchangeably. © 2022 IEEE.

Bidirectional Encoder Representations from Transformers for the COVID-19 vaccine stance classification

Vaccine-related information is awash on social media platforms like Twitter and Facebook. One party supports vaccination, while the other opposes vaccination and promotes misconceptions and misleading information about the risks of vaccination. The analysis of social media posts can give significant information into public opinion on vaccines, which can help government authorities in decision-making. In this work, an ensemble-based BERT model has been proposed for the classification of COVID-19 vaccine-related tweets into AntiVax, ProVax, and neural sentiment classes. The proposed model performed significantly well with a micro F1-score of 0.532 and an accuracy of 0.532 and achieved the second rank in the shared competition. © 2021 Copyright for this paper by its authors.

Using Supervised Learning Models for Creating a New Fake News Analysis and Classification of a COVID-19 Dataset: A case study on Covid-19 in Iran

Today, the growth of the coronavirus as a pandemic and its global expansion is a significant concern in our society and the international community. However, in recent years, many individuals have shifted their major source of news and information to social networks. Consequently, the widespread dissemination of false and misleading information on social media is significant for most politicians. Our effort is not only against COVID-19 but against an 'infodemic' as well. To address this, on COVID-19, we have collected and released a labeled dataset of 7,000 social media postings Persian data, and articles of authentic and false news. Covid 19 fake news has been detected in other languages such as Arabic, English, Chinese, and Hindi. We execute a multi-label task (actual vs. fictitious) on the labeled dataset and compare it to six machine learning baselines: Logistic Regression, Support Vector Machine, Decision Tree, Naive Bayes, K-Nearest Neighbors, and Random Forest. On the test set, the support vector machine gives us the best results, with an 89 percent accuracy rate. © 2022 IEEE.

A Deep Learning Model for Classification of COVID-19 Related Fake News

The health crisis caused by COVID-19 throws the whole world into the biggest emergency of the century. Moreover, the pandemic has become awful because of the spread of inadequate and fake news or information among common people. Fake news, gossip and misleading information are on the rise due to the popularity of web-based information sources among people, such as social media, news feeds, online blogs and e-news articles. Monitoring and identifying such fake stories is a prerequisite to cease unwanted panic in this pandemic. But carrying out this task manually is challenging and labour intensive. Computer-assisted pattern recognition can now be used to replace human contact thanks to developments in machine learning, deep learning models and natural language processing. This is also essential for accurately distinguishing between true and false information automatically. A hybrid deep learning classification model has been proposed here to identify and classify the fake news and misleading information on the ‘COVID-19 Fake News Dataset’ (taken from Mendeley) which is a collection of news or web article related to COVID-19. The proposed classification model has achieved an accuracy of 75.34% and outperforms the existing LSTM and BiLSTM techniques. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

A Survey on Information Distortion of COVID19: Applying Machine Learning and Natural Language Processing

The online infodemic created a lot of misinformation about covid-19. It is uncontrollable to stop the spreading of misleading information. It has reached a peak where people cannot differentiate fake news from the real one. The rapid spreading of covid19 fake news created a havoc among people. Here in this study, we will be comparing and studying all the ML techniques of AI which can predict the fake news from the real one. And also, NLP is used for understanding the take on text sentiments. By collecting and analyzing all the data from social media i.e., Twitter, Facebook, WhatsApp, Digital news we will start mining for the hoaxes. Here we will be able to see which ML techniques of AI like SVM, random forest, decision tree, logistic regression and some more, can give more precise results, and also to what extend an NLP can predict a sentiment from the given piece of text. Altogether this article explores the potential by demonstrating how algorithms try to understand human sentiments. This provides a new perception of throughout pandemic, how people in general interacts with misinformation and information found on the internet. Out of all, SVM brought out an accuracy of 98%. © 2022 Bharati Vidyapeeth, New Delhi.

Bridging the Health Disparity of African Americans through Conversational Agents

African Americans have faced health disparities in terms of access to health care and treatment of illnesses. The novel coronavirus disease 2019 pandemic exacerbates those disparities caused by limited access to medical care and healthy lifestyles, vulnerability to misleading information, and mistrust of the medical profession, all of which disproportionately affect the African American population in terms of infection and mortality. Conversational agents (CAs) are a technological intervention with the potential to narrow the disparities because they make health care more accessible, are effective in disseminating health information among a population with low health literacy, and can increase users' trust in health information. However, designing CAs for this population presents challenges with regard to embodying the African American culture into CAs and addressing privacy and security concerns. This commentary discusses some advantages and challenges of using CAs to help African Americans protect themselves against coronavirus disease 2019, and calls for more research in this area. © 2020 ACM.

Graphcut as a Segmentation Method of Covid-19 X-Ray Image for Diagnose Purpose

Medical images are vital for disease detection. The misleading information during the detection will lead to the worst part of diagnosing. Corona Virus or COVID-19 shocked the whole world with the new viral epidemics with a lower respiratory tract febrile illness causes pulmonary syndrome. Chest X-Ray and Chest Computed Tomography Scans (CT Scan) are the imaging tests that can identify the infection. As the COVID-19 virus is dissimilar to bacterial or viral pneumonia consolidation, X-ray analysis is chosen as a discriminative element that helps in assisting in the timely identification of COVID-19 infections. However, there are limitations in detecting the virus on the X-Ray image with raw eyes only. Several types of image processing are used to enhance the capability to detect the disease. Image segmentation is an image processing method that focuses on the abnormalities that appear on the medical image. Graphcut is one of the potential methods that can enhance to produce an understandable and more precise image for analyzing the process that can precisely diagnose the disease. We proposed the Graphcut with the combination of several techniques such as Dilate mask with Disk, Region-based Active Contour, Edge-based Active Contour, and Fill Holes. The experimental results show that the segmented region is the right part of training in the next phase. In conclusion, the enhancement of the Graphcut for the X-ray image helps the affected part be seen clearly for the diagnose purpose. © 2021 IEEE.