Methodology

For the collection of data we used NEXIS – a large computer database storing the text of many newspapers and magazines, accessible via the licence provided by the University of Birmingham. Our data collection included news articles from major UK national news outlets from major nation wide British broadsheet and tabloid newspapers. containing key words (“smart meter” / “electric vehicle” / “heat pumps”), published since 2015 (Paris Agreement) and more specifically from 1 January 2015 until 30 June 2024. The filters we applied are: Timeline (1 January 2015 – 30 June 2024); Source Location (Europe – United Kingdom of GB and NI); Source Language (English); Source Type (Newspapers); Geography (Europe – Northern Europe – United Kingdom). 

We extracted and retrieved sentences which contain the key words ‘smart meter(s)’, ‘electric vehicle(s)’ or heat pump(s) to assemble our databases. We used probability of particular topics based on topic probabilities identified using topic modelling (BERTopic). We found higher probability among right and left-leaning sources compared to the overall data – indicates some left-right narrative divide. 

We ran a language model (BERT) on an equal sample of sentences (left & right) containing our key words to obtain clusters of similar sentences. Sentences were assigned to one of 10 clusters. Clusters do not exclusively contain only right or left leaning sentences; instead, sources report differently on similar themes. Google Gemini plug-in was used to summarise themes for each cluster. Right- and left-wing samples were taken from each cluster to identify narrative variations on themes. BERT language model is an open-source machine learning framework for natural language processing. BERT is designed to help computers understand the meaning of ambiguous language in text by using surrounding text to establish context.