**ConvLab-2** is an open-source toolkit that enables researchers to build task-oriented dialogue systems with state-of-the-art models, perform an end-to-end evaluation, and diagnose the weakness of systems. As the successor of [ConvLab](https://github.com/ConvLab/ConvLab), ConvLab-2 inherits ConvLab's framework but integrates more powerful dialogue models and supports more datasets. Besides, we have developed an analysis tool and an interactive tool to assist researchers in diagnosing dialogue systems. [[paper]](https://arxiv.org/abs/2002.04793)
**ConvLab-3** is a flexible dialog system platform based on a **unified data format** for task-oriented dialog (TOD) datasets. The unified format serves as the adapter between TOD datasets and models: datasets are first transformed to the unified format and then loaded by models. In this way, the cost of adapting $M$ models to $N$ datasets is reduced from $M\times N$ to $M+N$. While retaining all features of [ConvLab-2](https://github.com/thu-coai/ConvLab-2), ConvLab-3 greatly enlarges supported datasets and models thanks to the unified format, and enhances the utility of reinforcement learning (RL) toolkit for dialog policy module. For typical usage, see our [paper](). Datasets and Trained models are also available on [Hugging Face Hub](https://huggingface.co/ConvLab).
-[Installation](#installation)
-[Tutorials](#tutorials)
-[Documents](#documents)
-[Unified Datasets](#Unified-Datasets)
-[Models](#models)
-[Supported Datasets](#Supported-Datasets)
-[End-to-end Performance on MultiWOZ](#End-to-end-Performance-on-MultiWOZ)
-[Module Performance on MultiWOZ](#Module-Performance-on-MultiWOZ)
-[Issues](#issues)
-[Contributions](#contributions)
-[Code Structure]($Code-Structure)
-[Contributing](#contributing)
-[Team](#Team)
-[Citing](#citing)
-[License](#license)
## Updates
2021.9.13:
-**2022.11.30**: ConvLab-3 release.
- Add [MultiWOZ 2.3](https://github.com/lexmen318/MultiWOZ-coref) dataset in `data` dir. The dataset adds co-reference annotations in addition to corrections of dialogue acts and dialogue states. [[paper]](https://arxiv.org/abs/2010.05594)
## Installation
2021.6.18:
You can install ConvLab-3 in the following ways according to your need. Higher versions of `torch` and `transformers` may also work.
- Add [LAUG](https://github.com/thu-coai/LAUG), an open-source toolkit for Language understanding AUGmentation. It is an automatic method to approximate the natural perturbations to existing data. Augmented data could be used to conduct black-box robustness testing or enhancing training. [[paper]](https://arxiv.org/abs/2012.15262)
- Add [SC-GPT](https://github.com/pengbaolin/SC-GPT) for NLG. [[paper]](https://arxiv.org/abs/2002.12328)
### Git clone and pip install in development mode (Recommend)
## Installation
For the latest and most configurable version, we recommend installing ConvLab-3 in development mode.
-[Getting Started](https://github.com/thu-coai/ConvLab-2/blob/master/tutorials/Getting_Started.ipynb)(Have a try on [Colab](https://colab.research.google.com/github/thu-coai/ConvLab-2/blob/master/tutorials/Getting_Started.ipynb)!)
-[Add New Model](https://github.com/thu-coai/ConvLab-2/blob/master/tutorials/Add_New_Model.md)
For more details about these models, You can refer to `README.md` under `convlab/$module/$model/$dataset` dir such as `convlab/nlu/jointBERT/multiwoz/README.md`.
- We add user dialogue act (*inform*, *request*, *bye*, *greet*, *thank*), remove 5 sessions that have incomplete dialogue act annotation and place it under `data/multiwoz` dir.
- Train/val/test size: 8434/999/1000. Split as original data.
- We offers a rule-based user simulator and a complete set of models for building a pipeline system on the CrossWOZ dataset. We correct few state annotation and place it under `data/crosswoz` dir.
- Train/val/test size: 5012/500/500. Split as original data.
*Notice*: The results are for commits before [`bdc9dba`](https://github.com/thu-coai/ConvLab-2/commit/bdc9dba72c957d97788e533f9458ed03a4b0137b)(inclusive). We will update the results after improving user policy.
We perform end-to-end evaluation (1000 dialogues) on MultiWOZ using the user simulator below (a full example on `tests/test_end2end.py`) :
To use ConvLab-3 as an off-the-shelf tool, you can install via:
### Policy
*Notice*: The results are for commits before [`bdc9dba`](https://github.com/thu-coai/ConvLab-2/commit/bdc9dba72c957d97788e533f9458ed03a4b0137b)(inclusive). We will update the results after improving user policy.
By running `convlab/policy/evalutate.py --model_name $model`
| | Task Success Rate |
| --------- | ----------------- |
| MLE | 0.56 |
| PG | 0.54 |
| PPO | 0.89 |
| GDPL | 0.58 |
### NLG
By running `convlab/nlg/evaluate.py MultiWOZ $model sys`
| | corpus BLEU-4 |
| -------- | ------------- |
| Template | 0.3309 |
| SCLSTM | 0.4884 |
## Translation-train SUMBT for cross-lingual DST
### Train
With Convlab-2, you can train SUMBT on a machine-translated dataset like this:
Execute `evaluate.py` (under `convlab/dst/`) with following command:
We also provide [Dockerfile](https://github.com/ConvLab/ConvLab-3/blob/master/Dockerfile) for building docker. Basically it uses the `requirement.txt` and then installs ConvLab-3 in development mode.
evaluation of our pre-trained models are: (joint acc.)
# create image
docker build -t convlab .
| type | CrossWOZ-en | MultiWOZ-zh |
| ----- | ----------- | ----------- |
| val | 12.4% | 48.5% |
| test | 12.4% | 46.0% |
| human_val | 10.6% | 47.4% |
# run container
docker run -dit convlab
`human_val` option will make the model evaluate on the validation set translated by human.
# open bash in container
docker exec-it CONTAINER_ID bash
```
Note: You may want to download pre-traiend BERT models and translation-train SUMBT models provided by us.
## Tutorials
Without modifying any code, you could:
-[Getting Started](https://github.com/thu-coai/ConvLab-2/blob/master/tutorials/Getting_Started.ipynb)(Have a try on [Colab](https://colab.research.google.com/github/thu-coai/ConvLab-2/blob/master/tutorials/Getting_Started.ipynb)!)
-[Introduction to unified data format](https://github.com/ConvLab/ConvLab-3/tree/master/data/unified_datasets)
-[Utility functions for unified datasets](https://github.com/ConvLab/ConvLab-3/blob/master/convlab/util/unified_datasets_util.py)
-[How to add a new dataset](https://github.com/thu-coai/ConvLab-2/blob/master/tutorials/Add_New_Model.md)
- How to add a new model
-[How to use RL toolkit](https://github.com/thu-coai/ConvLab-2/blob/master/tutorials/Train_RL_Policies)
Unified datasets are available under `data/unified_datasets` directory as well as [Hugging Face Hub](https://huggingface.co/ConvLab). We will continue adding more datasets listed in [this issue](https://github.com/ConvLab/ConvLab-3/issues/11). If you want to add a listed/custom dataset to ConvLab-3, you can create an issue for discussion and then create pull-request. We will list you as the [contributors](#Team) and highly appreciate your contributions!
- download pre-trained BERT models from:
## Models
-[bert-base-uncased](https://huggingface.co/bert-base-uncased) for CrossWOZ-en
-[chinese-bert-wwm-ext](https://huggingface.co/hfl/chinese-bert-wwm-ext) for MultiWOZ-zh
We list newly integrated models in ConvLab-3 that support unified data format and obtain strong performance. You can follow the link for more details about these models. Other models can be used in the same way as in ConvLab-2.
Trained models are available on [Hugging Face Hub](https://huggingface.co/ConvLab).
-[trained on CrossWOZ-en](https://convlab.blob.core.windows.net/convlab-2/crosswoz_en-pytorch_model.bin.zip)
-[trained on MultiWOZ-zh](https://convlab.blob.core.windows.net/convlab-2/multiwoz_zh-pytorch_model.bin.zip)
- Say the data set is CrossWOZ (English), (after extraction) just save the pre-trained model under `./convlab/dst/sumbt/crosswoz_en/pre-trained` and name it with `pytorch_model.bin`.
## Code structure
## Issues
You are welcome to create an issue if you want to request a feature, report a bug or ask a general question.
## Contributions
## Contributing
We welcome contributions from community.
We welcome contributions from community. Please see issues to find what we need.
- If you want to make a big change, we recommend first creating an issue with your design.
- Small contributions can be directly made by a pull request.
- If you like make contributions to our library, see issues to find what we need.
- If you want to add a new dataset, model, or other feature, please describe the dataset/model/feature in an issue before creating pull-request.
- Small change like fixing a bug can be directly made by a pull-request.
## Team
**ConvLab-3** is maintained and developed by Tsinghua University Conversational AI group (THU-coai), the [Dialogue Systems and Machine Learning Group](https://www.cs.hhu.de/en/research-groups/dialog-systems-and-machine-learning.html) at Heinrich Heine University, Düsseldorf, Germany and Microsoft Research (MSR).
**ConvLab-3** is maintained and developed by [Tsinghua University Conversational AI](http://coai.cs.tsinghua.edu.cn/) group (THU-COAI), the [Dialogue Systems and Machine Learning Group](https://www.cs.hhu.de/en/research-groups/dialog-systems-and-machine-learning.html) at Heinrich Heine University, Düsseldorf, Germany and Microsoft Research (MSR).
We would like to thank:
We would like to thank all contributors of ConvLab:
If you use ConvLab-2 in your research, please cite:
If you use ConvLab-3 in your research, please cite:
```
@inproceedings{zhu2020convlab,
title={ConvLab-2: An Open-Source Toolkit for Building, Evaluating, and Diagnosing Dialogue Systems},
author={Qi Zhu and Zheng Zhang and Yan Fang and Xiang Li and Ryuichi Takanobu and Jinchao Li and Baolin Peng and Jianfeng Gao and Xiaoyan Zhu and Minlie Huang},
year={2020},
booktitle={Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics},
}
@inproceedings{liu2021robustness,
title={Robustness Testing of Language Understanding in Task-Oriented Dialog},
author={Liu, Jiexi and Takanobu, Ryuichi and Wen, Jiaxin and Wan, Dazhen and Li, Hongguang and Nie, Weiran and Li, Cheng and Peng, Wei and Huang, Minlie},
year={2021},
booktitle={Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics},
