suiliangm / information_extraction_relation-classification-via-convolutional-deep-neural-network

Relation Classification via Convolutional Deep Neural Network

This is an implementation of paper Relation Classification via Convolutional Deep Neural Network in PyTorch

Model

Environment

• Python 3.6+
• PyTorch 1.1.0

Data

1. 'Cause-Effect' 1003
2. 'Instrument-Agency' 504
3. 'Product-Producer' 717
4. 'Entity-Origin' 716
5. 'Theme-Tool' 0
6. 'Component-Whole' 941
7. 'Content-Container' 540
8. 'Other' 1410

Usage

Data preprocess

python process.py --in_filename=ARGS --out_filename=ARGS

Training

python main.py --train_filename data/2010/merge_bin_2010/train/train-1 --test_filename data/2010/merge_bin_2010/test/test-1
python main.py --train_filename data/2010/merge_bin_2010/train/train-2 --test_filename data/2010/merge_bin_2010/test/test-2
python main.py --train_filename data/2010/merge_bin_2010/train/train-3 --test_filename data/2010/merge_bin_2010/test/test-3
python main.py --train_filename data/2010/merge_bin_2010/train/train-4 --test_filename data/2010/merge_bin_2010/test/test-4
python main.py --train_filename data/2010/merge_bin_2010/train/train-5 --test_filename data/2010/merge_bin_2010/test/test-5
python main.py --train_filename data/2010/merge_bin_2010/train/train-6 --test_filename data/2010/merge_bin_2010/test/test-6
python main.py --train_filename data/2010/merge_bin_2010/train/train-7 --test_filename data/2010/merge_bin_2010/test/test-7

python main.py data/2010/split_bin_2010/train-1 --test_filename data/2010/merge_bin_2010/test/test-

Result

Accuracy 67.84%