Background: It is important to measure the public response to the COVID-19 pandemic. Twitter is an important data source for infodemiology studies involving public response monitoring. Objective: The objective of this study is to examine COVID-19–related discussions, concerns, and sentiments using tweets posted by Twitter users. Methods: We analyzed 4 million Twitter messages related to the COVID-19 pandemic using a list of 20 hashtags (eg, “coronavirus,” “COVID-19,” “quarantine”) from March 7 to April 21, 2020. We used a machine learning approach, Latent Dirichlet Allocation (LDA), to identify popular unigrams and bigrams, salient topics and themes, and sentiments in the collected tweets. Results: Popular unigrams included “virus,” “lockdown,” and “quarantine.” Popular bigrams included “COVID-19,” “stay home,” “corona virus,” “social distancing,” and “new cases.” We identified 13 discussion topics and categorized them into 5 different themes: (1) public health measures to slow the spread of COVID-19, (2) social stigma associated with COVID-19, (3) COVID-19 news, cases, and deaths, (4) COVID-19 in the United States, and (5) COVID-19 in the rest of the world. Across all identified topics, the dominant sentiments for the spread of COVID-19 were anticipation that measures can be taken, followed by mixed feelings of trust, anger, and fear related to different topics. The public tweets revealed a significant feeling of fear when people discussed new COVID-19 cases and deaths compared to other topics. Conclusions: This study showed that Twitter data and machine learning approaches can be leveraged for an infodemiology study, enabling research into evolving public discussions and sentiments during the COVID-19 pandemic. As the situation rapidly evolves, several topics are consistently dominant on Twitter, such as confirmed cases and death rates, preventive measures, health authorities and government policies, COVID-19 stigma, and negative psychological reactions (eg, fear). Real-time monitoring and assessment of Twitter discussions and concerns could provide useful data for public health emergency responses and planning. Pandemic-related fear, stigma, and mental health concerns are already evident and may continue to influence public trust when a second wave of COVID-19 occurs or there is a new surge of the current pandemic.

Many countries are taking strict quarantine policies to prevent the rapid spread of COVID-19 (Corona Virus Disease 2019) around the world, such as city lockdown. Cities in China and Italy were locked down in the early stage of the pandemic. The present study aims to examine and compare the impact of COVID-19 lockdown on individuals’ psychological states in China and Italy. We achieved the aim by (1) sampling Weibo users (geo-location = Wuhan, China) and Twitter users (geo-location = Lombardy, Italy); (2) fetching all the users’ published posts two weeks before and after the lockdown in each region (e.g., the lockdown date of Wuhan was 23 January 2020); (3) extracting the psycholinguistic features of these posts using the Simplified Chinese and Italian version of Language Inquiry and Word Count (LIWC) dictionary; and (4) conducting Wilcoxon tests to examine the changes in the psycholinguistic characteristics of the posts before and after the lockdown in Wuhan and Lombardy, respectively. Results showed that individuals focused more on “home”, and expressed a higher level of cognitive process after a lockdown in both Wuhan and Lombardy. Meanwhile, the level of stress decreased, and the attention to leisure increased in Lombardy after the lockdown. The attention to group, religion, and emotions became more prevalent in Wuhan after the lockdown. Findings provide decision-makers timely evidence on public reactions and the impacts on psychological states in the COVID-19 context, and have implications for evidence-based mental health interventions in two countries.

Background: Family violence (including intimate partner violence/domestic violence, child abuse, and elder abuse) is a hidden pandemic happening alongside COVID-19. The rates of family violence are rising fast, and women and children are disproportionately affected and vulnerable during this time. Objective: This study aims to provide a large-scale analysis of public discourse on family violence and the COVID-19 pandemic on Twitter. Methods: We analyzed over 1 million tweets related to family violence and COVID-19 from April 12 to July 16, 2020. We used the machine learning approach Latent Dirichlet Allocation and identified salient themes, topics, and representative tweets. Results: We extracted 9 themes from 1,015,874 tweets on family violence and the COVID-19 pandemic: (1) increased vulnerability: COVID-19 and family violence (eg, rising rates, increases in hotline calls, homicide); (2) types of family violence (eg, child abuse, domestic violence, sexual abuse); (3) forms of family violence (eg, physical aggression, coercive control); (4) risk factors linked to family violence (eg, alcohol abuse, financial constraints, guns, quarantine); (5) victims of family violence (eg, the LGBTQ [lesbian, gay, bisexual, transgender, and queer or questioning] community, women, women of color, children); (6) social services for family violence (eg, hotlines, social workers, confidential services, shelters, funding); (7) law enforcement response (eg, 911 calls, police arrest, protective orders, abuse reports); (8) social movements and awareness (eg, support victims, raise awareness); and (9) domestic violence–related news (eg, Tara Reade, Melissa DeRosa). Conclusions: This study overcomes limitations in the existing scholarship where data on the consequences of COVID-19 on family violence are lacking. We contribute to understanding family violence during the pandemic by providing surveillance via tweets. This is essential for identifying potentially useful policy programs that can offer targeted support for victims and survivors as we prepare for future outbreaks.

Recently, researches related to unsupervised disentanglement learning with deep generative models have gained substantial popularity. However, without introducing supervision, there is no guarantee that the factors of interest can be successfully recovered (Locatello et al. 2018). Motivated by a real-world problem, we propose a setting where the user introduces weak supervision by providing similarities between instances based on a factor to be disentangled. The similarity is provided as either a binary (yes/no) or real-valued label describing whether a pair of instances are similar or not. We propose a new method for weakly supervised disentanglement of latent variables within the framework of Variational Autoencoder. Experimental results demonstrate that utilizing weak supervision improves the performance of the disentanglement method substantially.

Electroencephalography (EEG) is a non-invasive technique that can be used in brain machine interface (BMI) systems to measure and record brain electrical activity. Deep learning techniques have proved superior to conventional methods in EEG-based intent decoding. However, some deep learning models have overly complex structures while ensuring the accuracy of eEg recognition, resulting in reduced training and recognition speed. In this study, we proposed a compact multi-head self-attention deep learning decoder that combined the convolutional neural network-based EEGNet decoder with the ProbSparse multi-head self-attention mechanism. Compared with traditional self-attention methods, this decoder ensures alignment dependent on both time complexity and memory usage of O ( L log L ) and it has been demonstrated to enhance the accuracy of EEG-based intent recognition. The test results on dataset 2a from BCI Competition IV showed that the EEGNet-MSD decoder performed approximately 8% better than the competition winning decoder FBCSP e NBPW, and achieved better results than the latest LSTM neural decoding method. In addition, a binary classification test was performed on the Physiobank EEG motor imagery dataset, and the results showed that the accuracy of EEGNet-MSD was approximately 4% higher than EEGNet, validating the stability of the EEGNet-MSD decoder. This study provides a new solution for enhancing the performance of EEG-based intent decoding in both accuracy and decoding speed.