Resumen
Since the onset of the COVID-19 crisis, scholarly investigations and policy formulation have harnessed the potent capabilities of artificial intelligence (AI)-driven social media analytics. Evidence-driven policymaking has been facilitated through the proficient application of AI and natural language processing (NLP) methodologies to analyse the vast landscape of social media discussions. However, recent research works have failed to demonstrate a methodology to discern the underlying factors influencing COVID-19-related discussion topics. In this scholarly endeavour, an innovative AI- and NLP-based framework is deployed, incorporating translation, sentiment analysis, topic analysis, logistic regression, and clustering techniques to meticulously identify and elucidate the factors that are relevant to any discussion topics within the social media corpus. This pioneering methodology is rigorously tested and evaluated using a dataset comprising 152,070 COVID-19-related tweets, collected between 15th July 2021 and 20th April 2023, encompassing discourse in 58 distinct languages. The AI-driven regression analysis revealed 37 distinct observations, with 20 of them demonstrating a higher level of significance. In parallel, clustering analysis identified 15 observations, including nine of substantial relevance. These 52 AI-facilitated observations collectively unveil and delineate the factors that are intricately linked to five core discussion topics that are prevalent in the realm of COVID-19 discourse on Twitter. To the best of our knowledge, this research constitutes the inaugural effort in autonomously identifying factors associated with COVID-19 discussion topics, marking a pioneering application of AI algorithms in this domain. The implementation of this method holds the potential to significantly enhance the practice of evidence-based policymaking pertaining to matters concerning COVID-19.