Application of natural language processing for the recognition of obesity-related topics in the discourses of Argentine Twitter users

Abstract The global burden of obesity has risen due to various factors, including sociocultural aspects. Social representations (SRs) of obesity could help to understand the problem. Nowadays, social networks activate new social interaction processes and enable the construction of SRs. Tweets can identify mind-sets as cultural reflections of the times. This study aimed to identify widely shared obesity topics on Twitter-Argentina using Natural Language Processing. First, 134,766 Spanish tweets about obesity were collected from August 2021 to July 2022. Next, a geolocation filter removed non-Argentinian messages, leaving 48,149 tweets. The body text was cleaned and prepared for analysis. K-means clustering model was applied and 21 clusters were identified after reaching theoretical saturation. Then, 10 tweets from each cluster were randomly selected to analyse word usage and identify cluster themes. Thus, the main themes were “Gender-related”, “Obesity and family”, “Hate speech and fatphobia”, “Body image, perceptions and feelings”, “Public health actions” and “Obesity as a health-disease process”. These aggregations allowed us to understand obesity as a sociocultural phenomenon, the Argentinian culture-specific discourses surrounding it, and their connections to health. We firmly believe that this valuable information will enlighten the planning of future social-political actions to address this health-disease process.