Codes for the AAAI 2020 paper SG-Net: Syntax-Guided Machine Reading Comprehension
Update:
The text->dependency script to obtain the syntax mask: convert_span.py
*working in progress
pip install -r requirements.txt
You can run the model with the script run_squad.sh and run_race.sh.
We upload the processed example data in data folder which is annotated by our dependency labeler for quick practice.
The labeler model is the HPSG-Neural-Parser. The implementation for this work will be publicly available soon.
Please kindly cite this paper in your publications if it helps your research:
@inproceedings{zhang2019sgnet,
title={{SG-Net}: Syntax-Guided Machine Reading Comprehension},
author={Zhang, Zhuosheng and Wu, Yuwei and Zhou, Junru and Duan, Sufeng and Zhao, Hai and Wang, Rui},
booktitle={Proceedings of the Thirty-Fourth AAAI Conference on Artificial Intelligence},
year={2020}
}
:)