Abstract
Evictions are common in South Africa and have caught the attention of traditional mainstream media. With the rise of social media platforms mainly to provide or communicate newsworthy information, mainstream media has been suspected of bias in some instances. Research across disciplines has focused on addressing this bias by developing and applying information systems approaches, primarily focusing on developing frameworks that bring sentiment analysis of social media platforms into the fold when dealing with traditional media bias. Our study used and extended an existing sentiment analysis framework to analyse the discrepancies between the reporting of evictions in conventional media (using newspaper articles) and social media (using Twitter data). Statistical machine-learning methods were used to predict the sentiments of the tweets and newspaper articles. The comparison between articles and tweets shows that articles written by mainstream media are impartial, whereas tweets are not. Our study demonstrates how sentiment analysis frameworks can be applied to automatically analyse the discrepancies in reporting news between social media platforms and traditional media, irrespective of the language used. In this regard, an objective of this study was to analyse and select by way of a literature review the best sentiment analysis framework for revealing media bias in the reporting of evictions. The research problem and knowledge gap that was addressed involved defining the specific computational requirements for a reliable and accurate sentiment analysis framework that can reveal the degree of bias in the reporting of evictions.