CogSolver & CogSolver+

Source code for ICDM-2022 paper A Cognitive Solver with Autonomously Knowledge Learning for Reasoning Mathematical Answers and manuscipt Enhancing Mathematical Reasoning through Autonomously Learning Knowledge.

Dependencies

• python >= 3.6

• stanfordcorenlp

• torch == 1.8.1

Usage

Train and test model

python main.py

For running arguments, please refer to config.py.

The autonomouslly learned knowledge (as well as its strength) of CogSolver

• Math23K

knowledge\math23k\math23k_know_ww.txt: learned word-word relation knowledge

knowledge\math23k\math23k_know_wo.txt: learned word-operator relation knowledge

• MAWPS

knowledge\mawps\mawps_know_ww.txt: learned word-word relation knowledge

knowledge\mawps\mawps_know_wo.txt: learned word-operator relation knowledge

Citation

If you find this work useful, please cite our paper:

@inproceedings{liu2022cognitive,
  title={A cognitive solver with autonomously knowledge learning for reasoning mathematical answers},
  author={Liu, Jiayu and Huang, Zhenya and Lin, Xin and Liu, Qi and Ma, Jianhui and Chen, Enhong},
  booktitle={2022 IEEE International Conference on Data Mining (ICDM)},
  pages={269--278},
  year={2022},
  organization={IEEE}
}