New techniques from the field of Natural Language Processing (NLP) allow researchers to review and understand text at scale. These techniques have the potential to provide valuable insights into what people are reading online and could have applications across a great many different areas of regulatory work.
In this Economics Discussion Paper, we use NLP to classify the topic of each news article read by a sample of people who had their internet use tracked over one month in 2021. This allows us to measure the diversity of topics within an individual’s online news diet. We use this to study the relationship between how people access news articles and the diversity of news topics that they see.
We find that user reliance on online intermediaries (in particular social media and search engines) is associated with lower topic diversity; an insight which fed into Ofcom work on understanding the impact of social media on online news. We also find people that get a larger proportion of their online news from a Public Service Broadcaster (PSB) have a higher diversity of topics in their news diet and that people that make little or no use of PSBs online have a lower diversity of news topics. This latter finding is relevant to our ongoing review of public service media.
