BERT pretrain for KGAT

The BERT pre-trained codes for KGAT.

Introduction

• The fact verification shared task contains three steps: Document Retrieval, Sentence Retrieval and Fact Verification.
• We utilize claim-evidence pairs to pre-train BERT (Same codes of BERT + Prediction in retrieval_model).

Pre-train BERT for fact verification

• Pre-train the BERT model:
  • Run bash train.sh to pre-train the BERT model.