开源软件名称（OpenSource Name）：

开源软件地址(OpenSource Url)：

开源编程语言(OpenSource Language)：

开源软件介绍(OpenSource Introduction)：

DialogueGCN

A pre-processing and training code for DialogueGCN on DailyDialogue and Mastodon dataset. Use Bert base to preprocess the sentences. Based on DialogueGCN

I have commited the changes to DialogueGCN repo. If you still have any questions, welcome pr and issues!

Paper Approach

See preprocess_dailydialog2.py and train_daily_feature.py

Preprocess uses glove.840B.300d.txt to preprocess the InputSequence.

Training Use CNN to extract features from 300-dimension vector to 100-dimension vector.

The difference between train_daily_feature2.py and train_daily_feature3.py is the metrics. We use the same settings in DialogRNN (see train_daily_feature3.py), which uses micro-f1 and masks the 'no-emotion' label. Finally we get a f1 score about 44.24 on test set (50 epochs and select the final epoch results)

Citation

If you use any source codes included in this repo in your work, please cite the following paper. The bibtex are listed below:

@inproceedings{ghosal2019dialoguegcn,
  title={DialogueGCN: A Graph Convolutional Neural Network for Emotion Recognition in Conversation},
  author={Ghosal, Deepanway and Majumder, Navonil and Poria, Soujanya and Chhaya, Niyati and Gelbukh, Alexander},
  booktitle={Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP)},
  pages={154--164},
  year={2019}
}
@inproceedings{qin2021co,
  title={Co-GAT: A Co-Interactive Graph Attention Network for Joint Dialog Act Recognition and Sentiment Classification},
  author={Qin, Libo and Li, Zhouyang and Che, Wanxiang and Ni, Minheng and Liu, Ting},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  volume={35},
  number={15},
  pages={13709--13717},
  year={2021}
}