KGQE Models
See CQD\README.md and QTO\README.md for details.
KGQE Models
Download FB15k-237-betae, FB15k-betae, and NELL-betae from BetaE.
Put these three datasets into data/:
data/
│
├── FB15k-237-betae/
│ ├── train + valid + test data
├── FB15k-betae/
│ ├── train + valid + test data
└── NELL-betae/
└── train + valid + test data
Download the Pre-trained BERT from Hugging Face
Put Pre-trained BERT files into PLM/bert-base-cased/
Set "num_hidden_layers" in PLM/bert-base-cased/config.json to 1
Train QIPP with GQE:
python main.py --dataset NELL-betae --model_name geo
python main.py --dataset FB15k-237-betae --model_name geo
python main.py --dataset FB15k-betae --model_name geo
Train QIPP with Q2B:
python main.py --dataset NELL-betae --model_name box
python main.py --dataset FB15k-237-betae --model_name box
python main.py --dataset FB15k-betae --model_name box
Train QIPP with BetaE:
python main.py --dataset NELL-betae --model_name beta
python main.py --dataset FB15k-237-betae --model_name beta
python main.py --dataset FB15k-betae --model_name beta
Train QIPP with ConE:
python main.py --dataset NELL-betae --model_name cone
python main.py --dataset FB15k-237-betae --model_name cone
python main.py --dataset FB15k-betae --model_name cone
Train QIPP with MLP2Vec:
python main.py --dataset NELL-betae --model_name mlp2vec
python main.py --dataset FB15k-237-betae --model_name mlp2vec
python main.py --dataset FB15k-betae --model_name mlp2vec
Train QIPP with FuzzQE:
python main.py --dataset NELL-betae --model_name fuzzy
python main.py --dataset FB15k-237-betae --model_name fuzzy
python main.py --dataset FB15k-betae --model_name fuzzy