https://github.com/ha-lins/maml-dst

Model-agnostic meta-learning (MAML) for few-shot dialogue state tracking (DST) based on TRADE.
https://github.com/ha-lins/maml-dst

dialog-state-tracking maml meret meta-learning trade

Last synced: 29 days ago
JSON representation

Model-agnostic meta-learning (MAML) for few-shot dialogue state tracking (DST) based on TRADE.

• Host: GitHub
• URL: https://github.com/ha-lins/maml-dst
• Owner: ha-lins
• Created: 2019-12-16T02:47:50.000Z (about 5 years ago)
• Default Branch: master
• Last Pushed: 2022-12-08T05:24:57.000Z (about 2 years ago)
• Last Synced: 2023-10-20T20:08:25.045Z (about 1 year ago)
• Topics: dialog-state-tracking, maml, meret, meta-learning, trade
• Language: Python
• Homepage:
• Size: 725 KB
• Stars: 3
• Watchers: 3
• Forks: 2
• Open Issues: 3
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
## Model-Agnostic Meta-Learning (MAML) for Dialogue State Tracking

This repo implements the model-agnostic meta-learning (MAML) for dialogue state tracking (DST). It achieves better performance than [TRADE](https://github.com/jasonwu0731/trade-dst) in the few-shot setting. It also implements the multi-task learning (MTL-DST) as the baseline. This code has been written using PyTorch >= 1.0. 

This project has a similar idea to [MERET](https://www.aclweb.org/anthology/2020.acl-main.636.pdf). Our project is preliminarily named recallable meta-learning for DST (RM-DST), which can not only fast adapt to a low-resource new domain, but also employ a recallable mechanism to maintain the

high performance in old tasks. If you are interested in the work, you can contact me for the draft. The overview of RM-DST is as follows: 







Download the MultiWOZ dataset and the processed dst version.

```console

❱❱❱ python3 create_data.py

```

## Dependency

Check the packages needed or simply run the command

```console

❱❱❱ pip install -r requirements.txt

```

If you run into an error related to Cython, try to upgrade it first.

```console

❱❱❱ pip install --upgrade cython

```

## Unseen Domain DST

#### MAML-DST

Training

```console

❱❱❱ python3 myTrain_maml_DND.py -dec=TRADE -bsz=32 -dr=0.2 -lr=0.001 -le=1 -exceptd=${domain}

```

#### MTL-DST

Training

```console

❱❱❱ python3 myTrain_MTL.py -dec=TRADE -bsz=32 -dr=0.2 -lr=0.001 -le=1 -exceptd=${domain} -add_name=MTL

```

* -exceptd: except domain selection, choose one from {hotel, train, attraction, restaurant, taxi}.

#### Fine-tune

MAML-DST

```console

❱❱❱ python3 fine_tune_dnd.py -bsz=8 -dr=0.2 -lr=0.001 -path=${save_path_except_domain} -exceptd=${except_domain}

```

MTL-DST

```console

❱❱❱ python3 fine_tune.py -bsz=8 -dr=0.2 -lr=0.001 -path=${save_path_except_domain} -exceptd=${except_domain}

```

## Results







## Reference

1. [DAML](https://github.com/qbetterk/DAML): Domain Adaptive Dialog Generation via Meta Learning. [ACL 2019]

2. [TRADE](https://github.com/jasonwu0731/trade-dst): Transferable Multi-Domain State Generator for Task-Oriented Dialogue Systems. [ACL 2019]

3. Meta-Reinforced Multi-Domain State Generator for Dialogue Systems. [ACL 2020]