README.md

This code is for paper "HCL-TAT: A Hybrid Contrastive Learning Method for Few-shot Event Detection with Task-adaptive Threshold" accepted to Findings of EMNLP 2022.

Requirements

python=3.7
pytorch=1.10.2
cuda=10.2
transformers=2.8.0

Usage

The FewEvent dataset is under "data" dir. First, you should modify line 165-170 in main.py, line 32 in framework.py and line 226 in dataloader.py, by replacing the paths to yours. Then, you can train and evaluate the model by running the two following scripts.

1. Training

bash run.sh

You can modify the parameters in run.sh to train the model in different settings.

The checkpoint will be saved under the "checkpoint" dir. And if you specify the "embedding_visualization_path" and "result_dir", the visualization and metric results will be saved as well.

For example, if you want to train our HCL-TAT in 5-way-5-shot setting, the content in run.sh should be like:

for ((i=1; i<=5; i+=1))
do
  python main.py\
   --model=proto_dot\
   --trainN=5\
   --evalN=5\
   --K=5\
   --Q=1\
   --O=0\
   --distance_metric="proto"\
   --contrastive="Normal"\
   --temperature_alpha=0.5\
   --temperature_beta=0.1\
   --alpha=0.5\
   --beta=0.5\
   --threshold="mean"\
   --result_dir="result/proto_hcl";
done

The value of O and P should keep 0 and 1 respectively.

After each iteration, the code will print the results on test set in the screen, and finally, you can get 5 checkpoints. Note that results on dev set are lower than results on test set, which is normal in this split.

2. Testing

bash evaluate.sh

Similar to training stage, you can modify the parameters to evaluate models in different settings. If you want to evaluate a 5-way-5-shot HCL-TAT model, then the content in evaluate.sh should be like:

  python main.py\
   --model=proto_dot\
   --trainN=5\
   --evalN=5\
   --K=5\
   --Q=1\
   --O=0\
   --distance_metric="proto"\
   --contrastive="Normal"\
   --temperature_alpha=0.5\
   --temperature_beta=0.1\
   --alpha=0.5\
   --beta=0.5\
   --threshold="mean"\
   --test\
   --load_ckpt="proto_dot_fewevent_5_5_xxxxxxxx"\
   --embedding_visualization_path="embedding_visualization/proto_hcl"\

Please specify the "load_checkpoint" parameter to load the model.