yaxincui/swcc4event

Code for our ACL2022 paper "Improving Event Representation via Simultaneous Weakly Supervised Contrastive Learning and Clustering".

SWCC: a Simultaneous Weakly supervised Contrastive learning and Clustering framework for event representation learning.

---

Official implementation of our paper "Improving Event Representation via Simultaneous Weakly Supervised Contrastive Learning and Clustering".

Note.

The Event triples we use for the training data are extracted from the New York Times Gigaword Corpus using the Open Information Extraction system Ollie. Our event representation model is implemented using the Texar-PyTorch package. Our model starts from pre-trained checkpoints of BERT-based-uncased and we use the CLS token representation as the event representation. We train our model with a batch size of $256$ using an Adam optimizer. The learning rate is set as 2e-7 for the event representation model and 5e-4 for the prototype memory. We adopt the temperature $\tau=0.3$ and the number of prototypes used in our experiment is $10$.

Dataset

We recommend using gdown to download our data from Google Drive:

pip install gdown
gdown https://drive.google.com/u/0/uc?id=1FSZq0HM_rS2GKt0IDoMZlh6REytVrtFi&export=download

Quick Start

conda create -n swcc python=3.8
conda activate swcc
pip install -r requirements.txt

Training/Testing

To train and test a specific model, run the bash files train.sh and test.sh. For example, to train a new model and test a specific model, do the following:

// Training
// sh train.sh
CUDA_VISIBLE_DEVICES=0 python3 main.py --do-train 

// Testing
//sh test.sh
CUDA_VISIBLE_DEVICES=3 python3 main.py --do-eval --checkpoint ./models/checkpoint.pt 

Citation

@inproceedings{gao2022improving,
 author = {Jun Gao and Wei Wang and Changlong Yu and Huan Zhao and Wilfred Ng and Ruifeng Xu},
 booktitle = {Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics},
 title = {Improving Event Representation via Simultaneous Weakly Supervised Contrastive Learning and Clustering},
 year = {2022}
}