Predicting Confusion from Eye Tracking Data with Recurrent Neural Network-HAI 2019

Code for models and experiments presented in Predicting Confusion from Eye-Tracking Data with Recurrent Neural Network. This paper was presented at the Humanizing AI workshop of IJCAI 2019 on August 12, 2019 in Macao, China, where it won Best Paper Award.

This code is being made available to allow others to verify our procedure so that it can be used in other in related applications of eye-tracking data.

Please contact me if you notice any errors, have any suggestions, or just want to chat about classifying eye-tracker data using deep learning approaches.

Note: The dataset required to run these experiments has not yet been made publicly available.

Organization of this repository:

• The experiments reported in the paper are grouped into their own Jupyter Notebook by data condition.
• The main training function is in train.py with its helper functions. This function is common to all notebooks.
• Other utility functions are grouped together in utils.py. These functions are common to all notebooks, unless defined locally.