bazingagin / TWR-VAE

Improving Variational Autoencoder for Text Modelling withTimestep-Wise Regularisation

Improving Variational Autoencoder for Text Modelling withTimestep-Wise Regularisation

This repository includes the source code to reproduce the results presented in the paper Improving Variational Autoencoder for Text Modelling with Timestep-Wise Regularisation (COLING 2020)

Contents

1. Language modelling

2. Dialogue response generation

   2.1. Dialogue response generation on Switchboard

   2.2. Dialogue response generation on DailyDialog

1. Language modelling

To train the TWR-VAE on PTB/Yelp/Yahoo

cd lang_model/
python main.py -dt ptb/yelp/yahoo --z_type normal

To load trained model

python main.py -dt ptb/yelp/yahoo -l --model_dir path-to-the-trained-model/

To train the TWR-VAE-mean or TWR-VAE-sum on PTB/Yelp/Yahoo

python main.py -dt ptb/yelp/yahoo --z_type mean/sum

To train the TWR-VAE-LSTM-last25 or TWR-VAE-LSTM-last50 or TWR-VAE-LSTM-last75 on PTB/Yelp/Yahoo

python main.py -dt ptb/yelp/yahoo --z_type normal -par --partial_type last25/last50/last75

2. Dialogue response generation

Use pre-trained Word2vec: download Glove word embeddings glove.twitter.27B.200d.txt from https://nlp.stanford.edu/projects/glove/ and save it to the ./data folder. The default setting use 200 dimension word embedding trained on Twitter.

2.1 Dialogue response generation on Switchboard

To train TWR-VAE on Switchboard

cd dialogue_switchboard/
python train_swda.py

2.2 Dialogue response generation on Dailydialog

To train TWR-VAE on Dailydialog

cd dialogue_dd/
python train_dailydial.py

Acknowledgements

Thanks for the code published on Github repositories:

• https://github.com/guxd/DialogWAE
• https://github.com/fangleai/Implicit-LVM

@inproceedings{Li_TWRVAE_2020,
  title={Improving Variational Autoencoder for Text Modelling withTimestep-Wise Regularisation},
  author={Li, Ruizhe and Li, Xiao and Chen, Guanyi and Lin, Chenghua},
  booktitle={COLING},
  year={2020}
}