There are source codes for Our KGAT model.

• The fact verification shared task contains three steps: Document Retrieval, Sentence Retrival and Fact Verification.
• We come up a joint probability to aggregate evidence and do inference over several pieces of evidence.
• Kernel is used for Node and Edge attention for better inference.

• You should pre-train BERT model for bette performance.
  • Go to the pretrain folder for more information.
• Train the retrieval model.
  • Run bash train.sh to train the KGAT model.

• Run bash test.sh to get the test set perfomance.
• Go to the output folder to prepare your submission for Codalab leaderboard.
• Note that the predictions.jsonl file is the result we submit.

We compare our model performance with GEAR and keep all same experiment setting. The same ESIM based sentence retrieval and BERT (Base) encoder.

• Development set.

• Testing set.