Heroes, Villains, and Victims, and GPT-3: Automated Extraction of Character Roles Without Training Data
This provides a method for automated narrative analysis in fields like media studies or political science, but it is incremental as it applies an existing model to a new task without novel algorithmic contributions.
The paper tackled the problem of extracting character roles from narrative texts without training data by using GPT-3 with zero-shot question-answering prompts, achieving successful identification of heroes, villains, and victims across diverse domains such as newspaper articles, movie plots, and political speeches.
This paper shows how to use large-scale pre-trained language models to extract character roles from narrative texts without training data. Queried with a zero-shot question-answering prompt, GPT-3 can identify the hero, villain, and victim in diverse domains: newspaper articles, movie plot summaries, and political speeches.