Analysing Personal Attacks in U.S. Presidential Debates
This work addresses the need for transparency in political discourse for journalists, analysts, and the public, but it is incremental as it applies existing methods to a new dataset.
The researchers tackled the problem of detecting personal attacks in U.S. presidential debates by manually annotating transcripts from 2016, 2020, and 2024 elections and using fine-tuned transformer models and general-purpose LLMs, showing that task-specific adaptation of these models can enhance understanding of political communication.
Personal attacks have become a notable feature of U.S. presidential debates and play an important role in shaping public perception during elections. Detecting such attacks can improve transparency in political discourse and provide insights for journalists, analysts and the public. Advances in deep learning and transformer-based models, particularly BERT and large language models (LLMs) have created new opportunities for automated detection of harmful language. Motivated by these developments, we present a framework for analysing personal attacks in U.S. presidential debates. Our work involves manual annotation of debate transcripts across the 2016, 2020 and 2024 election cycles, followed by statistical and language-model based analysis. We investigate the potential of fine-tuned transformer models alongside general-purpose LLMs to detect personal attacks in formal political speech. This study demonstrates how task-specific adaptation of modern language models can contribute to a deeper understanding of political communication.