Intelligent Automation in Journalism: Are newsrooms ready to let machines write our news?

Abstract

The advancements in readily available structured data, artificial intelligence and automation tools have led to newsrooms exploring the possibility of creating news articles automatically. Sports, finance, politics and weather are all fields of journalism where an abundance of available data makes it possible to automate the process from beginning to end, using natural language generation (NLG) techniques to convert the data to human-readable articles. The difference between models based on templating and models based on machine learning algorithms is explained. The thesis is based on literature and media reports and is complemented by expert interviews from Finnish public broadcaster Yle and news agency STT. The aim of the study is to describe the techniques used to automate news production, explore how the production of news is generally automated, and to further build on these findings by interviews, and make visible some ethical questions related to machine-generated news.