RL-based-Graph2Seq-for-NQG

Code & data accompanying the ICLR 2020 paper "Reinforcement Learning Based Graph-to-Sequence Model for Natural Question Generation"

Get started

Prerequisites

This code is written in python 3. You will need to install a few python packages in order to run the code. We recommend you to use virtualenv to manage your python packages and environments. Please take the following steps to create a python virtual environment.

If you have not installed virtualenv, install it with pip install virtualenv.
Create a virtual environment with virtualenv venv.
Activate the virtual environment with source venv/bin/activate.
Install the package requirements with pip install -r requirements.txt.

Run the model

Download the preprocessed data from squad-split1 and squad-split2. And put the data under the root directory. So the file hierarchy will be like: data/squad-split1 and data/squad-split2.

Run the model
```
python main.py -config config/squad_split1/graph2seq_static_bert_finetune_word_70k_0.4_bs_60.yml
```
Note that you can specify the output path by modifying out_dir in a config file. If you want to finetune a pretrained model, you can specify the path to the pretrained model by modifying pretrained and you need to set out_dir to null. If you just want to load a pretrained model and evaluate it on a test set, you need to set both trainset and devset to null.

Finetune the model using RL

python main.py -config config/squad_split1/rl_graph2seq_static_bert_finetune_word_70k_0.4_bs_60.yml

Reference

If you found this code useful, please consider citing the following paper:

Yu Chen, Lingfei Wu and Mohammed J. Zaki. "Reinforcement Learning Based Graph-to-Sequence Model for Natural Question Generation." In Proceedings of the 8th International Conference on Learning Representations (ICLR 2020), Addis Ababa, Ethiopia, Apr 26-30, 2020.

@article{chen2019reinforcement,
  title={Reinforcement learning based graph-to-sequence model for natural question generation},
  author={Chen, Yu and Wu, Lingfei and Zaki, Mohammed J},
  journal={arXiv preprint arXiv:1908.04942},
  year={2019}
}