Analyzing digital societal interactions and sentiment classification in Twitter (X) during critical events in Chile

Abstract

This study explores the influence of social media content on societal attitudes and actions during critical events, with a special focus on occurrences in Chile, such as the COVID-19 pandemic, the 2019 protests, and the wildfires in 2017 and 2023. By leveraging a novel tweet dataset, this study introduces new metrics for assessing sentiment, inclusivity, engagement, and impact, thereby providing a comprehensive framework for analyzing social media dynamics. The methodology employed enhances sentiment classification through the use of a Deep Random Vector Functional Link (D-RVFL) neural network, which demonstrates superior performance over traditional models such as Support Vector Machines (SVM), naive Bayes, and back propagation (BP) neural networks, achieving an overall average accuracy of 78.30% (0.17). This advancement is attributed to deep learning techniques with direct input–output connections that facilitate faster and more precise sentiment classification. This analysis differentiates the roles of influencers, press radio, and television handlers during crises, revealing how various social media actors affect information dissemination and audience engagement. By dissecting online behaviors and classifying sentiments using the RVFL network, this study sheds light on the effects of the digital landscape on societal attitudes and actions during emergencies. These findings underscore the importance of understanding the nuances of social media engagement to develop more effective crisis communication strategies.

Keywords

• Sentiment classification
• Twitter
• Engagement
• Inclusivity