CIKM 2021: A Conditional Cascade Model for Relational Triple Extraction.

Requirements

The main requirements are:

• python 3.6
• torch 1.4.0
• tqdm
• transformers == 2.8.0

Usage

• Get pre-trained BERT model

  Download BERT-BASE-CASED and put it under ./pretrained.

• Train and select the dataset

python run.py --dataset=NYT   --num_train_epochs=100 --batch_size=18 --train

python run.py --dataset=WebNLG  --num_train_epochs=50 --batch_size=6  --train

python run.py --dataset=NYT_simple  --num_train_epochs=100  --batch_size=18 --train

python run.py --dataset=WebNLG_simple --num_train_epochs=50 --batch_size=6 --train

• Evaluate on the test set

python run.py --dataset NYT

python run.py --dataset WebNLG

python run.py --dataset NYT_simple

python run.py --dataset WebNLG_simple

Acknowledgement

Parts of our codes come from bert4keras.