Predicting Politicians' Stance with Text

Politicians and media outlets produce tons of text every day. To get a fair view of news, we need to understand the political leanings of these texts. This study introduces a clever way to estimate political bias using simple statistics. Researchers looked at speeches and party manifestos to predict biases in politics and views. The results were promising—they could guess biases better than just guessing. When the model made mistakes, it was often due to politicians changing their policies. To make the findings clearer, researchers identified words and sentiments that correlated with political power and party orientation. For example, positive sentiments often went hand in hand with political influence. A fun web app was made to show how this model works on unclear texts like news articles.