Improving the robustness of machine reading comprehension model with hierarchical knowledge and auxiliary unanswerability prediction

This repository provides the official TensorFlow implementation of the research paper Improving the robustness of machine reading comprehension model with hierarchical knowledge and auxiliary unanswerability prediction (Accepted by [Knowledge-Based Systems 2020]).

Requirements

• tensorflow-gpu >= 1.5.0
• Glove
• Elmo

Usage

python config.py --mode prepro
python config.py --mode train
python config.py --mode test

Performance

	EM	F1
Dev 1.1	73.9	82.4
Test 1.1	75.4	83.4
AddSent	58.3	65.5
AddOneSent	65.2	72.5

How to cite

If you extend or use this work, please cite the relevant papers:

@article{DBLP:journals/kbs/WuX20,
  author       = {Zhijing Wu and
                  Hua Xu},
  title        = {Improving the robustness of machine reading comprehension model with
                  hierarchical knowledge and auxiliary unanswerability prediction},
  journal      = {Knowledge-Based Systems},
  volume       = {203},
  pages        = {106075},
  year         = {2020},
}
@inproceedings{DBLP:conf/aaai/WuX20,
  author       = {Zhijing Wu and
                  Hua Xu},
  title        = {A Multi-Task Learning Machine Reading Comprehension Model for Noisy
                  Document (Student Abstract)},
  booktitle    = {The Thirty-Fourth {AAAI} Conference on Artificial Intelligence 2020},
  pages        = {13963--13964},
  year         = {2020},
}