Dual Consistency-enhanced Semi-supervised Sentiment Analysis towards COVID-19 Tweets

In the context of COVID-19, numerous people present their opinions through social networks. It is thus highly desired to conduct sentiment analysis towards COVID-19 tweets to learn the public's attitudes, and facilitate the government to make proper guidelines for avoiding the social unrest. Although many efforts have studied the text-based sentiment classification from various domains (e.g., delivery and shopping reviews), it is hard to directly use these classifiers for the sentiment analysis towards COVID-19 tweets due to the domain gap. In fact, developing the sentiment classifier for COVID-19 tweets is mainly challenged by the limited annotated training dataset, as well as the diverse and informal expressions of user-generated posts. To address these challenges, we construct a large-scale COVID-19 dataset from Weibo and propose a dual COnsistency-enhanced semi-superVIseD network for Sentiment Anlaysis (COVID-SA). In particular, we first introduce a knowledge-based augmentation method to augment data and enhance the model's robustness. We then employ BERT as the text encoder backbone for both labeled data, unlabeled data, and augmented data. Moreover, we propose a dual consistency (i.e., label-oriented consistency and instance-oriented consistency) regularization to promote the model performance. Extensive experiments on our self-constructed dataset and three public datasets show the superiority of COVID-SA over state-of-the-art baselines on various applications. IEEE

Fake News Detection Using Transformer and Ensemble Learning Models

On social media, misinformation can spread quickly, posing serious problems. Understanding the content and sensitive nature of fake news and misinformation is critical to prevent the damage caused by them. To this end, the characteristics of information must first be discerned. In this paper, we propose a transformer-based hybrid ensemble model to detect misinformation on the Internet. First, false and true news on Covid-19 were analyzed, and various text classification tasks were performed to understand their content. The results were utilized in the proposed hybrid ensemble learning model. Our analysis revealed promising results, establishing the capability of the proposed system to detect misinformation on social media. The final model exhibited an excellent F1 score (0.98) and accuracy (0.97). The AUC (Area Under The Curve) score was also high at 0.98, and the ROC (Receiver Operating Characteristics) curve revealed that the true-positive rate of the data was close to one in this model. Thus, the proposed hybrid model was demonstrated to be successful in recognizing false information online. © 2022 IEEE.

Classification of COVID-19 Tweets

Classification is a vital work to human beings in day today life as it breaks down complex subjects. In the same way, text classification is very important to understand and realize the subject of the text. © 2021 Copyright for this paper by its authors.

Forecasting COVID-19 Vaccination Rates using Social Media Data

The COVID-19 pandemic has had a profound impact on the global community, and vaccination has been recognized as a crucial intervention. To gain insight into public perceptions of COVID-19 vaccines, survey studies and the analysis of social media platforms have been conducted. However, existing methods lack consideration of individual vaccination intentions or status and the relationship between public perceptions and actual vaccine uptake. To address these limitations, this study proposes a text classification approach to identify tweets indicating a user's intent or status on vaccination. A comparative analysis between the proportions of tweets from different categories and real-world vaccination data reveals notable alignment, suggesting that tweets may serve as a precursor to actual vaccination status. Further, regression analysis and time series forecasting were performed to explore the potential of tweet data, demonstrating the significance of incorporating tweet data in predicting future vaccination status. Finally, clustering was applied to the tweet sets with positive and negative labels to gain insights into underlying focuses of each stance. © 2023 ACM.

Ensemble Deep Learning Framework for Situational Aspects-Based Annotation and Classification of International Student's Tweets during COVID-19

As the COVID-19 pandemic swept the globe, social media plat-forms became an essential source of information and communication for many. International students, particularly, turned to Twitter to express their struggles and hardships during this difficult time. To better understand the sentiments and experiences of these international students, we developed the Situational Aspect-Based Annotation and Classification (SABAC) text mining framework. This framework uses a three-layer approach, combining baseline Deep Learning (DL) models with Machine Learning (ML) models as meta-classifiers to accurately predict the sentiments and aspects expressed in tweets from our collected Student-COVID-19 dataset. Using the pro-posed aspect2class annotation algorithm, we labeled bulk unlabeled tweets according to their contained aspect terms. However, we also recognized the challenges of reducing data's high dimensionality and sparsity to improve performance and annotation on unlabeled datasets. To address this issue, we proposed the Volatile Stopwords Filtering (VSF) technique to reduce sparsity and enhance classifier performance. The resulting Student-COVID Twitter dataset achieved a sophisticated accuracy of 93.21% when using the random forest as a meta-classifier. Through testing on three benchmark datasets, we found that the SABAC ensemble framework performed exceptionally well. Our findings showed that international students during the pandemic faced various issues, including stress, uncertainty, health concerns, financial stress, and difficulties with online classes and returning to school. By analyzing and summarizing these annotated tweets, decision-makers can better understand and address the real-time problems international students face during the ongoing pandemic.

Detecting COVID-19 Vaccine Stance and Symptom Reporting from Tweets using Contextual Embeddings

Determining sentiments of the public with regard to COVID-19 vaccines is crucial for nations to efficiently carry out vaccination drives and spread awareness. Hence, it is a field requiring accurate analysis and captures the interest of many researchers. Microblogs from social media websites such as Twitter sometimes contain colloquial expressions or terminology difficult to interpret making the task a challenging one. In this paper, we propose a method for multi-label text classification for the track of”Information Retrieval from Microblogs during Disasters (IRMiDis)” presented by the”Forum of Information Retrieval Evaluation” in 2022, related to vaccine sentiment among the public and reporting of someone experiencing COVID-19 symptoms. The following methodologies have been utilised: (i) Word2Vec and (ii) BERT, which uses contextual embedding rather than the fixed embedding used by conventional natural language models. For Task 1, the overall F1 score and Accuracy are 0.503 and 0.529, respectively, placing us fourth among all the teams, while for Task 2, they are 0.740 and 0.790, placing us second among all the teams who submitted their work. Our code is openly accessible through GitHub. 1 © 2022 Copyright for this paper by its authors.

A real-world evaluation of the implementation of NLP technology in abstract screening of a systematic review.

The laborious and time-consuming nature of systematic review production hinders the dissemination of up-to-date evidence synthesis. Well-performing natural language processing (NLP) tools for systematic reviews have been developed, showing promise to improve efficiency. However, the feasibility and value of these technologies have not been comprehensively demonstrated in a real-world review. We developed an NLP-assisted abstract screening tool that provides text inclusion recommendations, keyword highlights, and visual context cues. We evaluated this tool in a living systematic review on SARS-CoV-2 seroprevalence, conducting a quality improvement assessment of screening with and without the tool. We evaluated changes to abstract screening speed, screening accuracy, characteristics of included texts, and user satisfaction. The tool improved efficiency, reducing screening time per abstract by 45.9% and decreasing inter-reviewer conflict rates. The tool conserved precision of article inclusion (positive predictive value; 0.92 with tool vs. 0.88 without) and recall (sensitivity; 0.90 vs. 0.81). The summary statistics of included studies were similar with and without the tool. Users were satisfied with the tool (mean satisfaction score of 4.2/5). We evaluated an abstract screening process where one human reviewer was replaced with the tool's votes, finding that this maintained recall (0.92 one-person, one-tool vs. 0.90 two tool-assisted humans) and precision (0.91 vs. 0.92) while reducing screening time by 70%. Implementing an NLP tool in this living systematic review improved efficiency, maintained accuracy, and was well-received by researchers, demonstrating the real-world effectiveness of NLP in expediting evidence synthesis.

Ensemble of deep learning language models to support the creation of living systematic reviews for the COVID-19 literature.

BACKGROUND: The COVID-19 pandemic has led to an unprecedented amount of scientific publications, growing at a pace never seen before. Multiple living systematic reviews have been developed to assist professionals with up-to-date and trustworthy health information, but it is increasingly challenging for systematic reviewers to keep up with the evidence in electronic databases. We aimed to investigate deep learning-based machine learning algorithms to classify COVID-19-related publications to help scale up the epidemiological curation process. METHODS: In this retrospective study, five different pre-trained deep learning-based language models were fine-tuned on a dataset of 6365 publications manually classified into two classes, three subclasses, and 22 sub-subclasses relevant for epidemiological triage purposes. In a k-fold cross-validation setting, each standalone model was assessed on a classification task and compared against an ensemble, which takes the standalone model predictions as input and uses different strategies to infer the optimal article class. A ranking task was also considered, in which the model outputs a ranked list of sub-subclasses associated with the article. RESULTS: The ensemble model significantly outperformed the standalone classifiers, achieving a F1-score of 89.2 at the class level of the classification task. The difference between the standalone and ensemble models increases at the sub-subclass level, where the ensemble reaches a micro F1-score of 70% against 67% for the best-performing standalone model. For the ranking task, the ensemble obtained the highest recall@3, with a performance of 89%. Using an unanimity voting rule, the ensemble can provide predictions with higher confidence on a subset of the data, achieving detection of original papers with a F1-score up to 97% on a subset of 80% of the collection instead of 93% on the whole dataset. CONCLUSION: This study shows the potential of using deep learning language models to perform triage of COVID-19 references efficiently and support epidemiological curation and review. The ensemble consistently and significantly outperforms any standalone model. Fine-tuning the voting strategy thresholds is an interesting alternative to annotate a subset with higher predictive confidence.

Speech acts in the Dutch COVID-19 Press Conferences.

An open source corpus of all Dutch COVID-19 Press Conferences with sentences annotated on the basis of John Searle's Speech Act taxonomy was created. It contains all 58 press conferences held between March 6 2020 and April 20 2021 and has 9.441 manually annotated sentences. Speech acts were annotated in a consistent manner, with a Krippendorff's alpha of .71. The corpus is easy to use and rich in metadata, with lexical, syntactic, discourse (speaker, question or answer) features and information on the type of regulations being present. We analyse the press conferences in terms of speech act usage, giving insight into the use of speech acts over time, the relation of speech act usage to real world phenomena, the general structure of the press conferences and the division of roles between speakers. Relations were found between speech act usage and the type of press conference (i.e. easing, tightening or neutral) as well as the number of hospital admissions. Speech act classes showed preferred locations within the press conferences, indicating a general structure. Distinct roles between speakers were identified. We also investigate the use of our set of labelled sentences for training a speech act classifier and achieve a reasonable accuracy of .73 and a mean reciprocal rank of .74 with the state of the art transformer RoBERTa model. Supplementary Information: The online version of this article contains supplementary material available 10.1007/s10579-022-09602-7.

Automated analysis of postings in fourth grade online discussions to help teachers keep students on-track

In the recent, and ongoing, Covid-19 pandemic, remote or online K-12 schooling became the norm. Even if the pandemic tails off somewhat, remote K-12 schooling will likely remain more frequent than it was before the pandemic. A mainstay technique of online learning, at least at the college and graduate level, has been the online discussion. Since it does afford the potential for meaningful learner-learner and instructor-learner interaction, which are vital for distance learning, it is worth considering online discussions for K-12 remote schooling. One challenge with online learning in general, and online discussion in particular, is that it is labor intensive for teachers to moderate. Effective moderating of online discussions is vital for discussions to be nurturing, effective learning situations. Yet, moderating of online discussions is notoriously labor-intensive for teachers/instructors. Further, since younger learners are more likely to drift off topic, in general, but particularly in small group online discussions, automated early warning systems are helpful. The current study collected small group, "book club”, discussion data from fourth graders reading web-based eBooks in Slovenian primary schools, qualitatively coded the data and analyzed postings using computer-based natural language processing to predict when students went off-topic. One indicator that postings are on-topic is book relevance, i.e. that the posting is relevant to eBook content. The computer algorithm correctly predicted book relevance of postings 90 percent of the time, suggesting that automated computer algorithms could assist teachers with moderating online discussions, providing real-time notifications of problems in online discussions. Further, this study provided a proof-of-concept that small group online discussions, in web-based eBooks can be practical and educationally meaningful in fourth grade classes. [ FROM AUTHOR] Copyright of Interactive Learning Environments is the property of Routledge and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full . (Copyright applies to all s.)

Text Analysis of COVID-19 Tweets

During the COVID-19 pandemic many countries were forced to implement lockdowns to prevent further spread of the SARS-CoV-2, prohibiting people from face-to-face social interactions. This unprecedented circumstance led to an increase in traffic on social media platforms, one of the most popular of which is Twitter, with a diverse spectrum of users from around the world. This quality, along with the ability to use its API for research purposes, makes it a valuable resource for data collection and analysis. In this paper we aim to present the sentiments towards the COVID-19 pandemic and vaccines as it was imprinted through the users' tweets when the events were actually still in motion. For our research, we gathered the related data from Twitter and characterized the gathered tweets in two classes, positive and negative;using the BERT model, with an accuracy of 99%. Finally, we performed various time series analyses based on people's sentiment with reference to the pandemic period of 2021, the four major vaccine's companies as well as on the vaccine's technology.

Hate Speech in a Telegram Conspiracy Channel During the First Year of the COVID-19 Pandemic.

Research has explored how the COVID-19 pandemic triggered a wave of conspiratorial thinking and online hate speech, but little is empirically known about how different phases of the pandemic are associated with hate speech against adversaries identified by online conspiracy communities. This study addresses this gap by combining observational methods with exploratory automated text analysis of content from an Italian-themed conspiracy channel on Telegram during the first year of the pandemic. We found that, before the first lockdown in early 2020, the primary target of hate was China, which was blamed for a new bioweapon. Yet over the course of 2020 and particularly after the beginning of the second lockdown, the primary targets became journalists and healthcare workers, who were blamed for exaggerating the threat of COVID-19. This study advances our understanding of the association between hate speech and a complex and protracted event like the COVID-19 pandemic, and it suggests that country-specific responses to the virus (e.g., lockdowns and re-openings) are associated with online hate speech against different adversaries depending on the social and political context.

Fake News Detection System: An implementation of BERT and Boosting Algorithm

On social media, false information can proliferate quickly and cause big issues. To minimize the harm caused by false information, it is essential to comprehend its sensitive nature and content. To achieve this, it is necessary to first identify the characteristics of information. To identify false information on the internet, we suggest an ensemble model based on transformers in this paper. First, various text classification tasks were carried out to understand the content of false and true news on Covid-19. The proposed hybrid ensemble learning model used the results. The results of our analysis were encouraging, demonstrating that the suggested system can identify false information on social media. All the classification tasks were validated and shows outstanding results. The final model showed excellent accuracy (0.99) and F1 score (0.99). The Receiver Operating Characteristics (ROC) curve showed that the true-positive rate of the data in this model was close to one, and the AUC (Area Under The Curve) score was also very high at 0.99. Thus, it was shown that the suggested model was effective at identifying false information online. © 2023, EasyChair. All rights reserved.

Assessing the Impact of Contextual Information in Hate Speech Detection

Social networks and other digital media deal with huge amounts of user-generated contents where hate speech has become a problematic more and more relevant. A great effort has been made to develop automatic tools for its analysis and moderation, at least in its most threatening forms, such as in violent acts against people and groups protected by law. One limitation of current approaches to automatic hate speech detection is the lack of context. The spotlight on isolated messages, without considering any type of conversational context or even the topic being discussed, severely restricts the available information to determine whether a post on a social network should be tagged as hateful or not. In this work, we assess the impact of adding contextual information to the hate speech detection task. We specifically study a subdomain of Twitter data consisting of replies to digital newspapers posts, which provides a natural environment for contextualized hate speech detection. We built a new corpus in Spanish (Rioplatense variant) focused on hate speech associated to the COVID-19 pandemic, annotated using guidelines carefully designed by our interdisciplinary team. Our classification experiments using state-of-the-art transformer-based machine learning techniques show evidence that adding contextual information improves the performance of hate speech detection for two proposed tasks: binary and multi-label prediction, increasing their Macro F1 by 4.2 and 5.5 points, respectively. These results highlight the importance of using contextual information in hate speech detection. Our code, models, and corpus has been made available for further research. © 2013 IEEE.

Covid-19 Dataset Analysis: A Systematic Review

Due to their need to be connected to the rest of the world, people started to use social networks extensively to share their feelings and be informed, especially during the Covid-19 pandemic and its lockdown. The tremendous growth of content in social media increased the frequency of researchers' work on natural language understanding, text classification, and information retrieval. Unfortunately, not all languages have benefited equally from this interest. Arabic is an example of such languages. The main reason behind this gap is the limited number of datasets that addressed Covid-19-related topics. To this aim, we performed the first-of-its-kind systematic review that covered, to the best of our knowledge, the most Arabic Covid-19 datasets freely available or access granted upon request. This paper presents these 15 datasets alongside their features and the type of analysis conducted. The general concern of the authors is to direct researchers to reliable and freely available datasets that advance the progress of Arabic Covid-19-related studies. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

Detection of Depression-Related Tweets in Mexico Using Crosslingual Schemes and Knowledge Distillation.

Mental health problems are one of the various ills that afflict the world's population. Early diagnosis and medical care are public health problems addressed from various perspectives. Among the mental illnesses that most afflict the population is depression; its early diagnosis is vitally important, as it can trigger more severe illnesses, such as suicidal ideation. Due to the lack of homogeneity in current diagnostic tools, the community has focused on using AI tools for opportune diagnosis. Unfortunately, there is a lack of data that allows the use of IA tools for the Spanish language. Our work has a cross-lingual scheme to address this issue, allowing us to identify Spanish and English texts. The experiments demonstrated the methodology's effectiveness with an F1-score of 0.95. With this methodology, we propose a method to solve a classification problem for depression tweets (or short texts) by reusing English language databases with insufficient data to generate a classification model, such as in the Spanish language. We also validated the information obtained with public data to analyze the behavior of depression in Mexico during the COVID-19 pandemic. Our results show that the use of these methodologies can serve as support, not only in the diagnosis of depression, but also in the construction of different language databases that allow the creation of more efficient diagnostic tools.

NLP Text Classification for COVID-19 Automatic Detection from Radiology Report in Indonesian Language

Radiology is used as an important assessment for patients with pulmonary disease. The radiology images are usually accompanied by a written report from a radiologist to be passed to the other referring physicians. These radiology reports are written in a natural language where they can have different systematic structures based on the language used. In our study, the radiology reports were collected from an Indonesian hospital and written in Bahasa Indonesia. We performed an automatic text classification to differentiate the information written in the radiology reports into two classes, COVID-19 and non COVID-19. To find the best model, we evaluated several embedding techniques available for Bahasa and five Machine Learning (ML) models, namely (1) XGBoost, (2) fastText, (3) LSTM, (4) Bi-LSTM and (5) IndoBERT. The result shows that IndoBERT outperformed the others with an accuracy of 98%. In terms of training speed, the shallow neural network architecture implemented with the fastText library can train the model in under one second and still result in a reasonably good accuracy of 86%. © 2022 IEEE.

Social media analytics with applications in disaster management and COVID-19 events

Social media such as Twitter offers a tremendous amount of data throughout an event or a disastrous situation. Leveraging social media data during a disaster is beneficial for effective and efficient disaster management. Information extraction, trend identification, and determining public reactions might help in the future disaster or even avert such an event. However, during a disaster situation, a robust system is required that can be deployed faster and process relevant information with satisfactory performance in real-time. This work outlines the research contributions toward developing such an effective system for disaster management, where it is paramount to develop automated machine-enabled methods that can provide appropriate tags or labels for further analysis for timely situation-awareness. In that direction, this work proposes machine learning models to identify the people who are seeking assistance using social media during a disaster and further demonstrates a prototype application that can collect and process Twitter data in real-time, identify the stranded people, and create rescue scheduling. In addition, to understand the people's reactions to different trending topics, this work proposes a unique auxiliary feature-based deep learning model with adversarial sample generation for emotion detection using tweets related to COVID-19. This work also presents a custom Q&A-based RoBERTa model for extracting related phrases for emotions. Finally, with the aim of polarization detection, this research work proposes a deep learning pipeline for political ideology detection leveraging the tweet texts and the expressed emotions in the text. This work also studies and conducts the historical emotion and polarization analysis of the COVID-19 pandemic in the USA and several individual states using tweeter data. (PsycInfo Database Record (c) 2023 APA, all rights reserved)

Covid-19 Dataset Analysis: A Systematic Review

Due to their need to be connected to the rest of the world, people started to use social networks extensively to share their feelings and be informed, especially during the Covid-19 pandemic and its lockdown. The tremendous growth of content in social media increased the frequency of researchers' work on natural language understanding, text classification, and information retrieval. Unfortunately, not all languages have benefited equally from this interest. Arabic is an example of such languages. The main reason behind this gap is the limited number of datasets that addressed Covid-19-related topics. To this aim, we performed the first-of-its-kind systematic review that covered, to the best of our knowledge, the most Arabic Covid-19 datasets freely available or access granted upon request. This paper presents these 15 datasets alongside their features and the type of analysis conducted. The general concern of the authors is to direct researchers to reliable and freely available datasets that advance the progress of Arabic Covid-19-related studies. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

SENTIMENTAL ANALYSIS OF COVID-19 TWITTER DATA USING DEEP LEARNING AND MACHINE LEARNING MODELS

The novel coronavirus disease (COVID-19) is an ongoing pandemic with large global attention. However, spreading fake news on social media sites like Twitter is creating unnecessary anxiety and panic among people towards this disease. In this paper, we applied machine learning (ML) techniques to predict the sentiment of the people using social media such as Twitter during the COVID-19 peak in April 2021. The data contains tweets collected on the dates between 16 April 2021 and 26 April 2021 where the text of the tweets has been labelled by training the models with an already labelled dataset of corona virus tweets as positive, negative, and neutral. Sentiment analysis was conducted by a deep learning model known as Bidirectional Encoder Representations from Transformers (BERT) and various ML models for text analysis and performance which were then compared among each other. ML models used were Naïve Bayes, Logistic Regression, Random Forest, Support Vector Machines, Stochastic Gradient Descent and Extreme Gradient Boosting. Accuracy for every sentiment was separately calculated. The classification accuracies of all the ML models produced were 66.4%, 77.7%, 74.5%, 74.7%, 78.6%, and 75.5%, respectively and BERT model produced 84.2 %. Each sentiment-classified model has accuracy around or above 75%, which is a quite significant value in text mining algorithms. We could infer that most people tweeting are taking positive and neutral approaches. © 2023, Universidad Politecnica Salesiana. All rights reserved.