/Behind-the-Scenes

Code and data for CIKM-2021 paper《Behind the Scenes: An Exploration of Trigger Biases Problem in Few-Shot Event Classification》

Primary LanguagePython

Dataset

  • MAVEN
    • download
    • put the files train.jsonl, val.jsonl, test.jsonl under the file dataset/maven_origin_data.
  • FewEvent
    • already in dataset/fewevent_origin_data
  • wordvector
    • download glove.6B.300d.txt, and put it under the file tool_data

Environment

pip install -r requirements.txt

process data

bash bash/prepare.sh

run

bash bash/Glove.sh gpu_id

you will get the results:

  • IUS(normal): 5-way-5-shot test Test accuracy: 84.97
  • TUS(trigger_unifirom): 5-way-5-shot test Test accuracy: 60.74
  • COS(blurry_uniform): 5-way-5-shot test Test accuracy: 45.91

Citation

If you use our code and data in your work, please cite our paper:

@inproceedings{wang2021behind,
  title={Behind the Scenes: An Exploration of Trigger Biases Problem in Few-Shot Event Classification},
  author={Wang, Peiyi and Xun, Runxin and Liu, Tianyu and Dai, Damai and Chang, Baobao and Sui, Zhifang},
  booktitle={Proceedings of the 30th ACM International Conference on Information \& Knowledge Management},
  pages={1969--1978},
  year={2021},
  url={https://doi.org/10.1145/3459637.3482236}
 
}

Acknowledgements

Thanks

@inproceedings{wang2020MAVEN,
  title={{MAVEN}: A Massive General Domain Event Detection Dataset},
  author={Wang, Xiaozhi and Wang, Ziqi and Han, Xu and Jiang, Wangyi and Han, Rong and Liu, Zhiyuan and Li, Juanzi and Li, Peng and Lin, Yankai and Zhou,   Jie},
  booktitle={Proceedings of EMNLP 2020},
  year={2020}
}

for providing the Maven dataset. and

@inproceedings{deng2020meta,
  title={Meta-learning with dynamic-memory-based prototypical network for few-shot event detection},
  author={Deng, Shumin and Zhang, Ningyu and Kang, Jiaojian and Zhang, Yichi and Zhang, Wei and Chen, Huajun},
  booktitle={Proceedings of the 13th International Conference on Web Search and Data Mining},
  pages={151--159},
  year={2020}
}

for providing the FewEvent dataset.