Responsible Streamlining of Radio News Production using Large Language Models

In collaboration with Algemeen Nederlands Persbureau (ANP), the largest news agency in the Netherlands, we are exploring how Large Language Models (LLMs) can be responsibly used to automate the conversion of news articles into radio scripts.

Our research examines whether an LLM like ChatGPT can maintain ANP's stringent standards of truth and quality while reducing the time and effort currently devoted to manually converting written news articles to text meant for radio recordings. We've established a close working relationship with ANP and its employees, which has been crucial in understanding the specific requirements, challenges, and expectations associated with their work.

The research project commenced with a six-week immersion period at ANP's premises, culminating in a research proposal designed to meet ANP's operational needs and align with the academic goals of the RAAIT programme. This proposal now forms the basis of a Master's thesis project, with the research being carried out on-site at ANP by a Master student connected to our programme.

This research is guided by user feedback and aims to use LLM responsibly, focusing on maintaining news content integrity and quality while seeking ways to enhance operational efficiency.

In summary, this project represents a synergistic partnership between academia and industry, aiming to enhance radio news production efficiency using LLMs while upholding standards of responsible AI use and journalistic integrity.

Researchers

Fabian Kok, MSc., researcher; Dr. Marieke Peeters, senior researcher; Christian Vesters, Master’s student; Dr. Daan Kolkman programme manager

Institute

Responsible Applied Artificial InTelligence (RAAIT) programme, Hogeschool Utrecht

Partners

ANP