A BERT's Eye View: A Big Data Framework for Assessing Language Convergence and Accommodation

Abstract

The current paper details a novel quantitative framework leveraging recent advances in AI and Natural Language Processing to quantitatively assess language convergence and accommodation. This new framework is computationally cheap and straightforward to implement. The framework is then applied to a case study of immigration rhetoric in the lead up to the 2016 general election in the USA. Major results from the case study show that (1) Democrats and Republicans exhibited significant language convergence with members of their own parties, (2) President Barack Obama and Hillary Clinton converged with Senate Democrats’ immigration rhetoric, (3) Democrats accommodated the immigration rhetoric of both President Barack Obama and (candidate) Hillary Clinton, (4) contrary to initial hypotheses, Donald Trump's vitriolic immigration rhetoric did not show signs of language convergence with Republicans in the Senate, and (5) equally surprising, Senate Republicans showed significant non-accommodation to Donald Trump despite potential political costs for having done so.