Public Happiness in the Age of Social Media: A Thematic Review and Topic Modelling Analysis

Sentiment analysis has gained increased attention due to its ability to automatically extract opinions, feelings, and emotions from large volumes of data collected from social media platforms. Its significant capability in understanding people’s feelings and perceptions has resulted in its extensive usage in various domains, including marketing, finance, politics, and health. Over the years, scientists have come with a range of solutions, from complex data-driven and topic modelling techniques to lexicon-based and machine learning approaches. This paper offers an overview of sentiment analysis research in the age of social media, by identifying and analyzing the published manuscripts in the field. Thus, a selection of 5374 English-written articles, published between 2008-2025, were gathered Web of Science. With the R-tool assistance, namely Biblioshiny, multiple indicators were examined, and various illustrations were created. The top-10 most cited papers from the data collection set were summarized, and a thematic review was also conducted, including thematic maps and themes evolution, to better understand the field and identify the main research areas, as well as their key connections and progress over time. Lastly, topic modelling techniques, such as Latent Dirichlet Allocation (LDA) and BERTopic were applied to uncover the main topics in the dataset. Based on LDA, three main topics were delimited: social media content (mainly Twitter) and public opinion about various events, sentiment analysis during crises events, such as COVID-19 pandemic, and technical sentiment analysis methods. Similarly, BERTopic uncovered three topics related to public discussions about vaccines and stocks), multimodal emotion analysis and sarcasm detection.