/InstructDS

EMNLP 2023: Instructive Dialogue Summarization with Query Aggregations

Primary LanguagePythonMIT LicenseMIT

InstructDS: Instructive Dialogue Summarization with Query Aggregations

made-with-python arxiv ACL

[EMNLP 2023 Slides], [Paper], [Training and Evaluation Code], [Poster], [Dataset]

Dialogue summarization framework described in the paper Instructive Dialogue Summarization with Query Aggregations (EMNLP 2023). This part is for demo. The training and evaluation part can be found from above Training and Evaluation Code link.

[Demo Code], [Demo Slides],

Framework

Requirements

python 3.10

pip install -r requirements.txt

Main Contributions

  1. Data

    1. SAMSum
    2. SAMSum_QDS (Ours)
    3. DialogSum (Ours with name replacement)
    4. DialogSum_QDS (Ours)
    5. TODSum
    6. TODSum_QDS (Ours)
    7. DREAM

  2. Traned Model

    1. Our model is trained from Flan-T5-XL.
    2. The model is uploaded and accessible from InstructDS.

  3. Demo of Instruvtive Summarization

    bash demo.sh
# A100 GPU with 40G memory: Pass
# A5000 GPU with 24G memory: Pass

  4. Demo Page (You can run locally.)

Fore more information, please refer to Slides for Demo, Paper, and Poster.

Citation

@inproceedings{wang-etal-2023-instructive,
    title = "Instructive Dialogue Summarization with Query Aggregations",
    author = "Wang, Bin  and
      Liu, Zhengyuan  and
      Chen, Nancy",
    editor = "Bouamor, Houda  and
      Pino, Juan  and
      Bali, Kalika",
    booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing",
    month = dec,
    year = "2023",
    address = "Singapore",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2023.emnlp-main.474",
    pages = "7630--7653",
}

@misc{wang2023instructive,
      title={Instructive Dialogue Summarization with Query Aggregations}, 
      author={Bin Wang and Zhengyuan Liu and Nancy F. Chen},
      year={2023},
      eprint={2310.10981},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}