Skip to content
Implementation of NAACL 2019 paper "Semantically-Aligned Equation Generation for Solving and Reasoning Math Word Problem"
Python
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
src
README.md

README.md

End-to-End Math Solver

The implementation of the NAACL 2019 paper Semantically-Aligned Equation Generation for Solving and Reasoning Math Word Problems.

Usage

  1. Download required files:
  2. Preprocess dataset:
cd src/
mkdir ../data/
python make_train_valid.py ~/storage/MWP/data/math23k_fix.json ~/storage/cc.zh.300.vec ../data/train_valid.pkl --valid_ratio 0 --char_based

Note that the warning is not unexpected, because some of the problems use operator out of +, -, *, /. The purpose of the flags are as follow:

  • --valid-ratio 0: Set the ratio the validation dataset should take. It should be set to 0 when running 5-fold cross validation.
  • --char_based: Specify if you want to tokenize the problem text into characters rather into words.
  1. Start 5-fold cross-validation by
mkdir ../models
python train.py ../data/train_valid.pkl ../models/model.pkl --five_fold
You can’t perform that action at this time.