ntunlplab / SEEN

SEEN: Structured Event Enhancement Network for Explainable Need Detection of Information Recall Assistance

SEEN: Structured Event Enhancement Network for Explainable Need Detection of Information Recall Assistance

This repo provides the source code & data of our paper SEEN: Structured Event Enhancement Network for Explainable Need Detection of Information Recall Assistance. If you use any of our code, processed data or pretrained models, please cite:

Dependency

• OS: Linux/Debian
• Python: 3.9.10
• Pytorch: 1.11.0
• CUDA: 11.5

Command of package installation

pip install torch==1.11.0 torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu115
pip install torch-scatter torch-sparse torch-cluster torch-spline-conv torch-geometric -f https://data.pyg.org/whl/torch-1.11.0+cu115.html
pip install -r requirements.txt
# cached-path > 1.1.2 has below bug 
# AttributeError: module 'cached_path' has no attribute 'file_friendly_logging'
pip install cached-path==1.1.2

Data Processing

Please refer to the data/script folder for data constrcion scripts.

Please refer to the preproc folder for data preprocessing scripts.

SEEN Usage

We pre-define four experiments setting for coonvenience.

• SEENLongformer
• SEENLongformerLarge

Train

python main.py \
    --do_train \
    --exp_name=$exp \
    --batch_size=$BATCH_SIZE \
    --val_batch_size=$VAL_BATCH_SIZE \
    --gpus=$GPUS \
    --val_step=$VAL_STEP \
    --epochs=3 \
    --seed=$seed \
    --pretrained_path=$seed # pass this vvalue to utilize the pretrained model)

Validate

python main.py \
    --do_val \
    --exp_name=$exp \
    --val_batch_size=$VAL_BATCH_SIZE

Test

python main.py \
    --do_test \
    --exp_name=$exp \
    --val_batch_size=$VAL_BATCH_SIZE
    --test_model_path=$TEST_MODEL_PATH # pass this value to test specific checkpoint)

Model Checkpoints

• SEENLongformer
• SEENLongformerLarge