A Review on Twitter Data Sentiment Analysis Related to COVID-19

On March 11, 2020, Dr. Tedros Adhanom Ghebreyesus, Director-General of the WHO, pronounced the outbreak a pandemic. The term "pandemic” refers to a disease that spreads rapidly and engulfs an entire geographic region. Coronavirus is a brand-new viral disease named after the year it first appeared. There is a scarcity of academic research on the subject to help researchers. Social media content analysis can reveal a lot concerning the general temperament and mood of the human race. In the field of sentiment analysis, deep learning models have been widely used. Sentiment analysis is a set of techniques, tools, and methods for detecting and extracting information. People have been using social networking sites like Twitter to voice their opinions, report realities, and provide a point of view on what is happening in the world today. Folks have always used Twitter to share data about the COVID-19 pandemic. People randomly share data visualizations from news revealed by organizations and the government. The numerous studies surveyed are selected based on a similarity. Every paper which is supervised performs sentiment analysis of Twitter data. Various studies have made used a fusion of diverse word embedding's with either machine learning classifiers or deep learning classifiers. Albeit the interpretation of single classifiers is satisfactory, the studies those proposed hybrid models have shown outstanding performance. On top of that transformer based models demonstrated quality results. It is concluded that using hybrid classifiers on Twitter data for sentiment analysis can surpass the achievements of the single classifiers. © 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

A Hybrid Approach for Detection of Fake News in Sinhala Text

The impact of technology on people's lives has grown continuously. The consumption of online news is one of the important trends as the share of population with internet access grows rapidly over time. Global statistics have shown that the internet and social media usage has an increasing trend. Recent developments like the Covid 19 pandemic have amplified this trend even more. However, the credibility of online news is a very critical issue to consider since it directly impacts the society and the people's mindsets. Majority of users tend to instinctively believe what they encounter and come into conclusions based upon them. It is essential that the consumers have an understanding or prior knowledge regarding the news and its source before coming into conclusions. This research proposes a hybrid model to predict the accuracy of a particular news article in Sinhala text. The model combines the general news content based analysis techniques using machine learning/ deep learning classifiers with social network related features of the news source to make predictions. A scoring mechanism is utilized to provide an overall score to a given news item where two independent scores- Accuracy Score (by analyzing the news content) and Credibility Score (by a scoring mechanism on social network features of the news source) are combined. The hybrid model containing the Passive Aggressive Classifier has shown the highest accuracy of 88%. Also, the models containing deep neural netWorks has shown accuracy around 75-80%. These results highlight that the proposed method could efficiently serve as a Fake News Detection mechanism for news content in Sinhala Language. Also, since there's no publicly available dataset for Fake News detection in Sinhala, the datasets produced in this work could also be considered as a contribution from this research. © 2022 IEEE.

Genomic image representation of human coronavirus sequences for COVID-19 detection

Coronavirus (CoV) disease 2019 (COVID-19) is a severe pandemic affecting millions worldwide. Due to its rapid evolution, researchers have been working on developing diagnostic approaches to suppress its spread. This study presents an effective automated approach based on genomic image processing (GIP) techniques to rapidly detect COVID-19, among other human CoV diseases, with high acceptable accuracy. The GIP technique was applied as follows: first, genomic graphical mapping techniques were used to convert the genome sequences into genomic grayscale images. The frequency chaos game representation (FCGR) and single gray-level representation (SGLR) techniques were used in this investigation. Then, several statistical features were obtained from the images to train and test many classifiers, including the k-nearest neighbors (KNN). This study aimed to determine the efficacy of the FCGR (with different orders) and SGLR images for accurately detecting COVID-19, using a dataset containing both partial and complete genome sequences. The results recommended the fourth-order FCGR image as a proper genomic image for extracting statistical features and achieving accurate classification. Furthermore, the results showed that KNN achieved an overall accuracy of 99.39% in detecting COVID-19, among other human CoV diseases, with 99.48% precision, 99.31% sensitivity, 99.47% specificity, 0.99 F1-score, and 0.99 Matthew's correlation coefficient. © 2022 THE AUTHORS

Comparative Analysis of COVID 19 Detection from Cough Speech Using Machine Learning Classifiers

Corona Virus Disease-2019, or COVID-19, has been on the rise since its emergence, so its early detection is necessary to stop it from spreading rapidly. Speech detection is one of the best ways to detect it at an early stage as it exhibits variations in the nasopharyngeal cavity and can be performed ubiquitously. In this research, three standard databases are used for detection of COVID-19 from speech signal. The feature set includes the baseline perceptual features such as spectral centroid, spectral crest, spectral decrease, spectral entropy, spectral flatness, spectral flux, spectral kurtosis, spectral roll off point, spectral skewness, spectral slope, spectral spread, harmonic to noise ratio, and pitch. 05 ML based classification techniques have been employed using these features. It has been observed that Generalized Additive Model (GAM) classifier offers an average of 95% and a maximum of 97.55% accuracy for COVID-19 detection from cough signals. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.

Machine Learning Technique with Spider Monkey Optimization for COVID-19 Sentiment Analysis

The coronavirus (COVID-19) has unquestionably posed one of the most difficult problems in human history. As of May 5, 2021, the global death toll from the on going COVID-19 epidemic was at 150 million. Monitoring, managing, and finally eradicating the disease can be made easier by understanding the mood represented in people's social media remarks. It is the goal of this research to educate social media policy, such as how much reasonableness is required to curb misrepresentation on social media. This research also examines people's opinions on COVID-19 over social media sharing and interaction on Twitter. For our studies, we've generated a new large-scale sentiment data collection called COVIDSENTI that includes 90,000 tweets about the COVID-19 epidemic from February to March of the following year. We classified the tweets based on their emotional tone: good, negative, or neutral. In this work, ensemble machine learning classifier and uses the Spider Monkey optimization (SMO) for selecting the optimal features. The raw datasets are cleaned by using pre-processing techniques and important features are extracted by using feature extraction practises. We used multiple parameter measurements to assess the gathered tweets for sentiment categorization. © 2022 IEEE.

Evaluating The Preliminary Models to Identify Fake News on COVID-19 Tweets

The spread propagation of fake news about COVID-19 can make it distressing to handle the pandemic situation. Identifying the fake and real news on social media needs to be done as quickly as possible to prevent chaos in the community and hampering the handling of COVID-19. In this study, we conducted some experiments to get a model that works well for classifying information into fake or real news using tweet data. We implemented two different ways to represent data to train machine learning classifier models, syntactic-based using Bag-of-Words and TF-IDF, and semantic-based using Word2Vec and FastText. We evaluated each model produced by the training process using two types of testing data. The results show that The Linear Support Vector Machine model using TF-IDF obtained the best F1-Score value in both testing data. The model obtained F1-Score 92.21% in Testing Data 1 and 93.33% in Testing Data 2.