KET

The PyTorch code for paper: Knowledge-Enriched Transformer for Emotion Detection in Textual Conversations (PDF)

The model is largely based on The Annotated Transformer

Steps

• Download data: download data to respective foler in ./data/: EC, DD, MELD, EmoryNLP, and IEMOCAP.
• Install Magnitude Medium GloVe for pretrained word embedding.
• Preprocess data: run preprocess.py to process csv or pkl (IEMOCAP) files into pkl data.
• Download ConceptNet and NRC_VAD.
• Preprocess ConceptNet and NRC_VAD: run preprocess_conceptnet.py and preprocess_NRC_VAD.py.
• Model training: run train.py.
• Model evaluation: run train.py with test_mode set.

Citing

If you find this repo or paper useful, please cite

@inproceedings{zhong2019knowledge,
    title = "Knowledge-Enriched Transformer for Emotion Detection in Textual Conversations",
    author = "Zhong, Peixiang and Wang, Di and Miao, Chunyan",
    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)",
    year = "2019",
    publisher = "Association for Computational Linguistics",
    pages = "165--176"
}