Behind the Guardian’s analysis of 100 years of MPs’ language on immigration

The Guardian, in collaboration with University College London, analyzed a century of UK parliamentary language on immigration, revealing a recent shift towards more negative sentiment. The analysis, covering 1925 to 2025, used a custom-built machine learning model developed by The Guardian's Data Science and Data Projects teams. The model was trained on over 22,600 fragments of manually annotated parliamentary speeches, identified using immigration-related trigger terms. Large Language Models were used to expand the training dataset. This bespoke model analyzed almost 238,000 fragments related to immigration to determine the overall sentiment score for each year. The goal was to objectively measure and track changes in sentiment towards immigration expressed by MPs in the House of Commons over time.

Explore how the topics in this article connect to other news stories

Discover articles with similar content using semantic similarity analysis.