The media framing dataset: Analyzing news narratives in Mexico and Colombia.

Abstract

This paper introduces "The Media Framing Dataset," a dataset developed through an in-depth examination of news articles from 140 local newspapers in Mexico and Colombia, covering events from May 2022 to August 2023. Our dataset captures a broad spectrum of topics, including politics, immigration, public opinion, and crime. The data collection involved a meticulous keyword-based search strategy designed to identify articles that illustrate various news-framing dimensions, such as Economics, Policy, Morality, and more. To construct this dataset, we employed a combination of manual and automated annotation techniques. Articles were categorized based on specific framing dimensions using a structured framework, developed in collaboration with experts in computational linguistics. The annotation process, conducted by trained annotators from Mexico's Delfin program, guarantees both precision and depth. "The Media Framing Dataset" serves as a valuable resource for NLP research with high potential for reuse. It is particularly suitable for analyzing cultural and linguistic nuances in media framing, assessing the impact of framing on public perception, and supporting the development of models that automatically detect framing techniques. Additionally, it provides a foundation for linguistic analysis and machine learning projects, enabling researchers and practitioners to explore media framing dynamics and develop innovative tools for media analysis.