Extracting Multiple-Relations in One-Pass with Pre-Trained Transformers

Notes

This repo contains the implementation for the algorithm in:

@article{wang2019extracting,
  title={Extracting Multiple-Relations in One-Pass with Pre-Trained Transformers},
  author={Wang, Haoyu and Tan, Ming and Yu, Mo and Chang, Shiyu and Wang, Dakuo and Xu, Kun and Guo, Xiaoxiao and Potdar, Saloni},
  journal={arXiv preprint arXiv:1902.01030},
  year={2019}
}

The codes are modified based on the original BERT repo. Some unrelated modules from the original repo have been deleted in order to make it easy to understand.

Data

We provide the processed Semeval2018 data along with the repo. For ACE dataset, we could not share it within this place due to the data policy.

Training (MRE)

The following command will work for training the model on Semeval dataset. For other configurable arguments, please refer to run_classifier.py.

python run_classifier.py \
        --task_name=semeval \
        --do_train=true \
        --do_eval=false \
        --do_predict=false \
        --data_dir=$DATA_DIR/semeval2018/multi \
        --vocab_file=$BERT_BASE_DIR/vocab.txt \
        --bert_config_file=$BERT_BASE_DIR/bert_config.json \
        --init_checkpoint=$BERT_BASE_DIR/bert_model.ckpt \
        --max_seq_length=128 \
        --train_batch_size=4 \
        --learning_rate=2e-5 \
        --num_train_epochs=30 \
        --max_distance=2 \
        --max_num_relations=12 \
        --output_dir=<path to store the checkpoint>

Predicting (MRE)

The following command will work for using the trained model to inference on the test dataset.

python run_classifier.py \
        --task_name=semeval \
        --do_train=false \
        --do_eval=false \
        --do_predict=true \
        --data_dir=$DATA_DIR/semeval2018/multi \
        --vocab_file=$BERT_BASE_DIR/vocab.txt \
        --bert_config_file=$BERT_BASE_DIR/bert_config.json \
        --init_checkpoint=$BERT_BASE_DIR/bert_model.ckpt \
        --max_seq_length=256 \
        --max_distance=2 \
        --max_num_relations=12 \
        --output_dir=<path to the stored checkpoint>

Training (SRE)