A Comprehensive Survey of News Bias Detection Datasets
TL;DR
The quality, relevance, and annotation guidelines of the data have a key influence on the accuracy and effectiveness of models that could perform automatic bias detection.
Abstract
News is one of the leading sources of knowledge that provides insight into what kind of perception shapes public beliefs, affects decisions, and also facilitates democratic systems. Today, digital journalism, along with extensive use of news portals on the web and social media, has considerably increased the speed and frequency in terms of its production and circulation. Digitisation of info is continuous, as it can alter content that is not natural and that contains a particular agenda that would challenge how someone interprets this content and understand exactly why it had been generated. Auto News Bias Detection research has become a major discipline utilizing various machine learning and natural language processing techniques in order to automatically assess different categories of media bias. The performance of this model mainly depends on the quality of benchmark data, which enables objective annotation, diversity and relevance of training/testing data. So, the quality, relevance, and annotation guidelines of the data have a key influence on the accuracy and effectiveness of models that could perform automatic bias detection.