<ul data-eligibleForWebStory="false"><li>Introduction of the largest and most comprehensive dataset of US presidential campaign television advertisements, along with machine-searchable transcripts and high-quality summaries for academic research.</li><li>Automation of the process through a large-scale parallelized, AI-based analysis pipeline to prepare, transcribe, and summarize over 9,700 presidential ads from the Julian P. Kanter Political Commercial Archive.</li><li>Human evaluations demonstrate that the AI-generated transcripts and summaries are of similar quality to manually created ones, proving the effectiveness of the methodology.</li><li>The dataset provides valuable insights, such as tracking the evolution of focal issue areas over seven decades of presidential elections, and the methodology can be applied to create high-quality summaries for other video datasets.</li></ul>

Using AI to Summarize US Presidential Campaign TV Advertisement Videos, 1952-2012

Discover more