Graph-to-Tree Learning for Solving Math Word Problems

PyTorch implementation of Graph based Math Word Problem solver described in our ACL 2020 paper Graph-to-Tree Learning for Solving Math Word Problems. In this work, we propose a solution for Math Word Problem Solving via graph neural network.

Steps to run the experiments

Requirements

• Python 3.6
• >= PyTorch 1.0.0

For more details, please refer to requiremnt file.

Training

[MATH23K]

first get into the math23k directory:

• cd math23k

training-test setting :

• python run_seq2tree_graph.py

cross-validation setting :

• python cross_valid_graph2tree.py

[MAWPS]

cross-validation setting :

• cd mawps
• python cross_valid_mawps.py

Contact

• As I have graduated from my master school, my school email address will be invalid soon. If you have further questions about our work, you can refer to my new email address zhangjipeng20@outlook.com.

Reference

@article{zhang2020graph2tree,
  title={Graph-to-Tree Learning for Solving Math Word Problems},
  author={Zhang, Jipeng and Wang, Lei and Lee, Roy Ka-Wei and Bin, Yi and Shao, Jie and Lim, Ee-Peng},
  journal={ACL 2020},
  year={2020}
}