Mapping Emotions: A Sentiment Analysis of Social Media Discourses on Violence in Guayaquil

This study aims to analyze the sentiments expressed in social media posts (X and TikTok) concerning urban violence in five Ecuadorian provinces, namely, Esmeraldas, Guayas, Los Ríos, Manabí, Pichincha, and Santo Domingo, from January 2021 to January 2025. The methodology uses web scraping tools to gather relevant comments, and integrates the RoBERTuito model, -a transformer-based model- with a BiLSTM layer, to provide critical insights into the dynamics of sentiment progression over time. Our hybrid approach produces a sentiment thermometer expressed in a trace map containing the five events considered the most violent in Ecuador with their respective chronology. Two of the facts that always remained with negative feelings were “corrupt judges” who released prisoners, as well as “missing children case” implying the rejection of the actions of the military involved with the minors for committing forced disappearance.