BiQUE: Biquaternionic Embeddings of Knowledge Graphs

This is the official implementation for "BiQUE: Biquaternionic Embeddings of Knowledge Graphs" (EMNLP 2021, Main Conference).

Citation

If you find our work helpful for your research, please cite our paper:

@inproceedings{guo-kok-2021-bique,
    title = "{BiQUE}: {B}iquaternionic Embeddings of Knowledge Graphs",
    author = "Guo, Jia and Kok, Stanley",
    booktitle = "Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing",
    year = "2021",
    url = "https://aclanthology.org/2021.emnlp-main.657",
}

Dependencies

Python 3.6+
PyTorch 1.0+
NumPy 1.17.2+
tqdm 4.41.1+

The folder structure is:

|-- .BiQUE
    |-- README.md
    |-- src_data
    |-- data
    |-- codes
    |-- ckpt

1. Preprocess the Datasets

To preprocess the datasets, run the following commands:

cd codes
python process_datasets.py

Now, the processed datasets are in the data directory.

2. Reproduce the Results

To reproduce the reported results of BiQUE on WN18RR, FB15k237, YAGO3-10, CN-100K and ATOMIC, please run the following commands:

cd codes
python reproduce.py dataset

dataset = ["WN18RR", "FB237", "YAGO3", "CN100K", "ATOMIC"]

3. Training BiQUE model

# WN18RR
python learn.py --dataset WN18RR --model BiQUE --rank 128 --optimizer Adagrad --learning_rate 1e-1 --batch_size 300 --regularizer wN3 --reg 1.5e-1 --max_epochs 200 --valid 5 -train -id 0 -save -weight


# FB15K-237
python learn.py --dataset FB237 --model BiQUE --rank 128 --optimizer Adagrad --learning_rate 1e-1 --batch_size 500 --regularizer wN3 --reg 7e-2 --max_epochs 300 --valid 5 -train -id 0 -save


# YAGO3-10
python learn.py --dataset YAGO3-10 --model BiQUE --rank 128 --optimizer Adagrad --learning_rate 1e-1 --batch_size 1000 --regularizer wN3 --reg 5e-3 --max_epochs 200 --valid 5 -train -id 0 -save


# CN-100k
python learn.py --dataset Concept100k --model BiQUE --rank 128 --optimizer Adagrad --learning_rate 1e-1 --batch_size 5000 --regularizer wN3 --reg 1e-1 --max_epochs 200 --valid 5 -train -id 0 -save


# ATOMIC
python learn.py --dataset Atomic --model BiQUE --rank 128 --optimizer Adagrad --learning_rate 1e-1 --batch_size 5000 --regularizer wN3 --reg 5e-3 --max_epochs 200 --valid 5 -train -id 0 -save

Acknowledgement

We refer to the codes of kbc. Thanks for their contributions.

guojiapub / BiQUE