- Paper link: https://www.csie.ntu.edu.tw/~yvchen/doc/SIGDIAL22_CUDA.pdf
- CUDA-augmented data: https://github.com/MiuLab/CUDA-DST/blob/main/cuda/cuda_data/sampleData.zip
The package general requirements are
- Python >= 3.7
- Pytorch >= 1.5 (installation instructions here)
- Transformers >= 3.0.2 (installation instructions here)
The package can be installed by running the following command. Run
sh setup.sh
It includes preprocessed MultiWOZ 2.1 and MultiWOZ 2.2 dataset.
Download, uncompress it, and place the
resulting multiwoz
folder under the root of the repository as ./multiwoz
.
cd multiwoz
unzip multiwoz.zip
mv multiwoz/* ./
rmdir multiwoz
If zip file not exist, downloads it from Github
cd MultiWOZ-coref
unzip MultiWOZ2_3.zip
python ./split_data.py
See ./cuda/README.md
See ./coco-dst/README.md
See ./trade-dst/README.md
See ./simpletod/README.md
See ./trippy-public/README.md
Please cite the following paper:
@inproceedings{lai2022controllable,
title={Controllable User Dialogue Act Augmentation for Dialogue State Tracking},
author={Lai, Chun-Mao and Hsu, Ming-Hao and Huang, Chao-Wei and Chen, Yun-Nung},
booktitle={Proceedings of the 23rd Annual Meeting of the Special Interest Group on Discourse and Dialogue},
year={2022}
}
This code includes other open source software components:
coco-dst,
trade-dst,
simpletod, and
trippy-public.
Each of these software components have their own license. Please see them under
./coco-dst
, ./trade-dst
, ./simpletod
, and ./trippy-public
folders.