Text classifier of sensasionalist headlines in spanish using BERT-Based models

Abstract

Information technologies play a crucial role in keeping society informed during global events like pandemics. However, sensational headlines can negatively impact public perception and trust in institutions. In this chapter, several BERT-based text classifiers were developed to classify sensational and non-sensational health-related headlines in Spanish. The models were fine-tuned on almost 2000 headlines from major Mexican newspapers, achieving up to 94% F1-Score and accuracy. This demonstrates the effectiveness of machine learning techniques in detecting sensationalism in news headlines.