Skip to content

goodbai-nlp/Sem-Dialogue

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

21 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Sem-Dialogue

An implementation for paper "Semantic Representation for Dialogue Modeling". You may find our paper here.

Requirements

  • python 3.6
  • pytorch 1.6
  • Tesla V100 (32G)
  • Memory > 150G (for DialogRE)

We recommend to use conda to manage virtual environments:

conda create --name <env> --file requirements.txt

Data

The preprocessed data for DialogRE and DialogRG are avaliable at here and here, respectively.

Preprocessing

bash /path/to/code/preprocess.sh

Training

bash /path/to/code/run-dual(hier).sh

Evaluation

bash /path/to/code/eval.sh                   # for dialogue relation extraction
bash /path/to/code/decode.sh                 # for dialogue response generation

Pretrained Models

DialogRE

Data-v1

Setting dev-F1 dev-F1c test-F1 test-F1c checkpoint
Hier 68.3 62.9 68.4 62.3 model
Dual 68.6 62.6 68.0 61.5 model

Data-v2

Setting dev-F1 dev-F1c test-F1 test-F1c checkpoint
Hier 68.8 62.4 66.6 61.2 model
Dual 68.4 62.7 67.3 61.7 model

Todo

  • clean code

References

@inproceedings{bai-etal-2021-semantic,
    title = "Semantic Representation for Dialogue Modeling",
    author = "Bai, Xuefeng  and
      Chen, Yulong  and
      Song, Linfeng  and
      Zhang, Yue",
    booktitle = "Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers)",
    month = aug,
    year = "2021",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2021.acl-long.342",
    doi = "10.18653/v1/2021.acl-long.342",
    pages = "4430--4445"
}

About

Code for our paper "Semantic Representation for Dialogue Modeling" in ACL2021

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published