FreqEmo: A Twitter Dataset Comprising Users' Tweeting Frequency and Emotions

Social media is now indispensable, used by all social, economic, age, and demographic groups. It provides extensive user-generated data, offering insights into users' mental states. This paper introduces a novel Twitter dataset comprising users' tweeting frequency and emotions shared. The dataset is based on Tweets collected from the USA and India during a 60-day period using adjectives representing emotions. Further, for frequently tweeting users, NRC emotion lexicon has been implemented which extracted sentiment and emotion scores of the users. The final dataset includes users' tweeting patterns along with the distribution of emotions and sentiments in their tweets. This dataset has been made publicly available and this project aims to help researchers develop models to assess users' emotional states based on social media activity without self-reporting, with applications in correlating usage patterns with well-being, clustering users by emotions and tweeting patterns, and analyzing temporal, spatial, and demographic trends. To provide baseline performance, k-means clustering is implemented in this study to cluster the users with similar usage patterns. Clusters are analyzed to determine the types of emotions and sentiments shared by users within each cluster. Based on the analysis, it is concluded that as Twitter usage increases within the clusters, the proportion of users dominated by negative sentiment also rises significantly, peaking at 41.67%. Additionally, the percentage of users whose predominant emotion is joy decreases with increased usage, reaching a low of 8.33% as opposed to the peak value of 69.44% in the group characterized by least usage, while the percentage of users dominated by anger and disgust grows to 17.24% and 4.17% as their usage intensifies.