The article presents a comprehensive dataset, Navigating News Narratives: A Media Bias Analysis Dataset, aimed at detecting and analyzing bias in news narratives. The dataset encompasses a broad spectrum of biases and is a valuable tool in the field of media studies and artificial intelligence. The data includes dimensions such as race, gender, age, occupation, and climate change, acknowledging that biases are not monolithic but a confluence of various factors. The dataset was created using active learning, which involves training models on a progressively expanding dataset, combining manual labeling with semi-supervised learning and iterative human verification. The data is intended to help researchers, journalists, and policymakers understand and mitigate the influence of biased narratives in shaping public discourse.
Publication date: 30 Nov 2023
Project Page: https://huggingface.co/datasets/newsmediabias/news-bias-full-data, https://zenodo.org/records/10231028
Paper: https://arxiv.org/pdf/2312.00168
