Bayesian Surprise Predicts Human Event Segmentation in Story Listening

This repository contains data and code for the paper: https://psyarxiv.com/qd2ra/
The embeddings and outputs from GPT-2 are in the extract-embeddings-data.zip file.
The button press data is in the button-press-proportions.zip file. Button press measures computed using kernel density estimates are stored in files with the suffix "density".
The derived measures of disfluency are in the disfluency_measures.zip file. In this folder the files {story_name}_processed.csv contain the disfluency measures for each story.
The results from the analysis are in the results.zip and supplementary_results_kernel_density.zip files.
The code is available at https://github.com/manojneuro/bayesiansurprise