R-BERT-relation-extraction

Implementation of Enriching Pre-trained Language Model Entity Information for Relation Classification.

Environment Requirements

• python 3.6.9
• pytorch 1.5.1+cu92
• transformers 2.11.0
• tqdm 4.40.1

Device

• TITAN Xp
• CUDA Version 9.0.176

Data

• SemEval2010 Task8 [paper]
• bert-base-uncased

Usage

1. Download the pre-trained BERT model and put it into the resource folder.
2. Run the following the commands to start the program.

python run.py \
    --batch_size=16 \
    --max_len=128 \
    --lr=2e-5 \
    --epoch=5 \
    --dropout=0.1

More details can be seen by python run.py -h.

3. You can use the official scorer to check the final predicted result (in the eval folder).

perl semeval2010_task8_scorer-v1.2.pl proposed_answer.txt predicted_result.txt >> result.txt

Result

The result of my version and that in paper are present as follows:

paper	my version
0.8925	0.8906

The training log can be seen in train.log and the official evaluation results is available in result.txt.

Note:

• Some settings may be different from those mentioned in the paper.
• No validation set used during training.

Reference Link

• https://github.com/monologg/R-BERT