lx-bistu/DEP-GAT

DEP-GAT

Dialogue Relationship Extraction with Dependency Relation

Setup

1. Download GloVe vectors from here and put it into dataset/ folder, Then download BERT-base-uncased model from here and put it into huggingface/bert-base-uncased/ folder.

2. Install from requirements.txt by pip install -r requirements.txt and run python -m spacy download en_core_web_sm

3. Insatll dgl library according to your cuda version using the command like below.

pip install dgl-cu110==0.6.1     # For CUDA 11.0 Build

4. download stopwords using the commands like below.

python
>>> import nltk
>>> nltk.download('stopwords')

Run code

Training

Execute the following command to train the DEP-GAN model.

# DEP-GAT-BERT model
python DEP-GAT-BERT.py

# DEP-GAT-LSTM model
python DEP-GAT-LSTM.py

Testing

# DEP-GAT-BERT model
python DEP-GAT-BERT.py --mode test --ckpt_path [your_ckpt_file_path]

# DEP-GAT-LSTM model
python DEP-GAT-LSTM.py --mode test --ckpt_path [your_ckpt_file_path]