Large language models can segment narrative events similarly to humans

Area of Science:

• Cognitive Science
• Computational Linguistics
• Artificial Intelligence

Background:

• Human event perception involves segmenting continuous experience into discrete occurrences.
• Quantifying event boundaries typically relies on aggregating multiple human annotations.
• Automated methods are needed to efficiently study event perception.

Approach:

• Utilized a large language model, GPT-3, to computationally derive event boundaries from narrative text.
• Compared GPT-3's event segmentation performance against human annotations.
• Assessed the correlation and proximity of GPT-3 derived boundaries to human consensus.

Key Points:

• GPT-3 successfully segmented continuous narrative text into distinct events.
• GPT-3's event annotations showed significant correlation with human annotations.
• GPT-3 identified event boundaries closer to the human consensus than individual annotators.

Conclusions:

• GPT-3 offers a feasible and automated solution for event boundary annotation.
• This demonstrates parallels between large language model predictions and human cognitive processes.
• Automated event annotation may advance the study of human event perception principles.