Towards Quantifying Magnitude of Political Bias in News Articles Using a Novel Annotation Schema

Proceedings of Recent Advances in Natural Language Processing, pages 671–678 Sep 1–3, 2021. https://doi.org/10.26615/978-954-452-072-4_076 671 Towards Quantifying Magnitude of Political Bias in News Articles Using a Novel Annotation Schema Lalitha Kameswari, Radhika Mamidi Language Technologies Research Centre IIIT Hyderabad Telangana, India v.a.lalitha@research.iiit.ac.in radhika.mamidi@iiit.ac.in Abstract Media bias is a predominant phenomenon present in most forms of print and electronic media such as news articles, blogs, tweets, etc. Since media plays a pivotal role in shaping public opinion towards political happenings, both political parties and media houses often use such sources as outlets to propagate their own prejudices to the public. There has been some research on detecting political bias in news articles. However, none of it attempts to analyse the nature of bias or quantify the mag- nitude of the bias in a given text. This paper presents a political bias annotated corpus viz. PoBiCo-21, which is annotated us- ing a schema specifically designed with 10 la- bels to capture various techniques used to cre- ate political bias in news. We create a ranking of these techniques based on their contribution to bias. After validating the ranking, we pro- pose methods to use it to quantify the magni- tude of bias in political news articles. 1 Introduction An increase in Internet and social media access has made it easier for people to know the happenings from all parts of the world. This improved reach of the media has led to people and organisations utilising such platforms to increase their popularity. With this extensive usage of mass communication, there is an increased risk of misinformation, fake news, twisting of facts, and the spread of contro- versies. In such cases, it is the responsibility of the media houses to verify the authenticity of the reported information and ensure that it is complete, neutral, and not misleading. However, we often see instances where some news articles emphasize more on particular view- points selectively. In some cases, journalists and media houses either intentionally or subconsciously present biased information aligned with their own political ideologies. This can alter public opin- ion hugely and impact the decision of the general public. This phenomenon of political bias is very prevalent in democracies, especially during elec- tions, where influencing public opinion can have far-reaching consequences. Detection of political bias in the news is a com- plex task when it comes to a multi-party political system like India, where multiple parties operate at both state and national levels1. There has been limited work to date in this area, as discussed in Section 2. However, in a scenario where most of the political content contains bias in some form, it is not just sufficient to identify the existence of bias. Therefore, we propose two important contribu- tions through this paper - a fine-grained annotation schema specifically designed for analysing polit- ical bias and a new sentiment analysis based ap- proach to quantify the magnitude of political bias in an article. We also present PoBiCo-21, an an- notated corpus for political bias, …

Content truncated to 500 words