This repository contains the code used for the paper "On The Role of Context in Reading Time Prediction", accepted to EMNLP 2024 as a main conference short paper.
We use version 1.2 of the MECO dataset, stored in joint_data_trimmed.rda. However, an off-by-one issue was detected in the trial, sentence and word (interest area) ID
scheme for some of the tokens, which would lead to incorrect data when averaging across subjects. This is corrected in preprocessing.R. The corrected data is then merged with the surprisal estimates from mGPT, which are taken from this project and given in mgtp_lc/. This script writes the csv files in merged_data/.
Our analysis uses the L1 (native speakers) data. For the sake of convenience for researchers who might be interested in using corrected L2 (second language learners) data, we provide code that performs the same correction in MECO_L2_fix.R.
The analyses discussed in the paper can be found in analysis.Rmd.
Please cite our work as:
@inproceedings{opedal2024role,
title = {On the Role of Context in Reading Time Prediction},
author = {Opedal, Andreas and Chodroff, Eleanor and Cotterell, Ryan and Wilcox, Ethan},
booktitle = {Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing},
month = nov,
year = {2024},
publisher = {Association for Computational Linguistics},
address = {Miami, Florida, USA},
url = {https://arxiv.org/abs/2409.08160},
}