Karen Zhou,Alexander A. Meitus,Milo Chase,Grace Wang,Anne Mykland,William Howell,Chenhao Tan
Abstract:Does Donald Trump speak differently from other presidents? If so, in what ways? Are these differences confined to any single medium of communication? To investigate these questions, this paper introduces a novel metric of uniqueness based on large language models, develops a new lexicon for divisive speech, and presents a framework for comparing the lexical features of political opponents. Applying these tools to a variety of corpora of presidential speeches, we find considerable evidence that Trump's speech patterns diverge from those of all major party nominees for the presidency in recent history. Some notable findings include Trump's employment of particularly divisive and antagonistic language targeting of his political opponents and his patterns of repetition for emphasis. Furthermore, Trump is significantly more distinctive than his fellow Republicans, whose uniqueness values are comparably closer to those of the Democrats. These differences hold across a variety of measurement strategies, arise on both the campaign trail and in official presidential addresses, and do not appear to be an artifact of secular time trends.
Computation and Language,Artificial Intelligence,Computers and Society,Social and Information Networks
What problem does this paper attempt to address?
The paper aims to explore the uniqueness of Donald Trump's speaking style compared to past U.S. presidents, particularly in terms of the divisiveness and targeted nature of his language. Researchers developed a novel uniqueness measurement method based on large language models, created a divisive language lexicon, and proposed a framework for comparing descriptions of political opponents. By analyzing different corpora such as presidential debates, State of the Union addresses, and campaign speeches, the study found that Trump's language patterns are the most distinctive among all major party presidential candidates. Specifically, Trump tends to use more divisive and confrontational language to target his political opponents and exhibits repetitive patterns when emphasizing. Moreover, Trump is more unique compared to his Republican colleagues, whose uniqueness scores are closer to those of Democrats. These differences are consistent across various measurement strategies, present both during the campaign and in formal presidential addresses, and are not a product of temporal trends.
Key findings of the paper include:
1. Trump used particularly divisive and confrontational language to target political opponents, making his speech patterns significantly different from other presidential candidates.
2. Trump also showed uniqueness in his repetitive patterns, tending to repeat certain words for emphasis.
3. In debates, State of the Union addresses, and campaign speeches, Trump's language patterns are more unique than those of other modern presidents, including his immediate predecessor and successor.
4. Trump's uniqueness score is higher compared to other members within the Republican Party, while the Republicans' scores are closer to those of the Democrats.
5. Trump's language patterns are not only reflected in the use of specific vocabulary but also in sentence structure and repetitive patterns.
The methods used in the study include large language models to quantify uniqueness, a divisive language lexicon to analyze language content, and a framework for mentioning opponents to assess the way political opponents are described. These methods together provide a comprehensive comparison of Trump's speech patterns, revealing specific features that may have been overlooked in previous research.