Kernel Graph Attention Network (KGAT)

There are source codes for Kernel Graph Attention Network for Fact Verification.

For more information about the FEVER 1.0 shared task can be found on this website.

Requirement

• Python 3.X
• fever_score

Data and Checkpoint

• Can be found at Google Drive.

Retrieval Model

• BERT based ranker.
• Go to the retrieval_model folder for more information.

Pretrain Model

• Pre-train BERT with claim-evidence pairs.
• Go to the pretrain folder for more information.

KGAT Model

• Our KGAT model.
• Go to the kgat folder for more information.

Results

The results are all on Codalab leaderboard. (The Rank@1 and Rank@2 use XLNet and BERT(large)).

Rank	User	Pre-train Model	Label Accuracy	FEVER Score
1	DREAM	XLNet	0.7685	0.7060
2	abcd_zh (Ours)	RoBERTa (Base)	0.7407	0.7038
3	a.soleimani.b	BERT (Large)	0.7186	0.6966
9	GEAR_single	BERT (Base)	0.7160	0.6710

KGAT performance with different pre-trained language model.

Pre-train Model	Label Accuracy	FEVER Score
RoBERTa (Base)	0.7407	0.7038
BERT (Large)	0.7361	0.7024
BERT (Base)	0.7281	0.6940

Citation

@article{liu2019kernel,
  title={Kernel Graph Attention Network for Fact Verification},
  author={Liu, Zhenghao and Xiong, Chenyan and Sun, Maosong},
  journal={arXiv preprint arXiv:1910.09796},
  year={2019}
}

Contact

If you have questions, suggestions and bug reports, please email:

[email protected]