This paper proposes a Natural Language Inference (NLI) model using Knowledge Graph (KG) for automatic fact-checking related to COVID-19 in Indonesian. The model consists of three modules: fact module, NLI module, and classification module. The fact module processes information in the KG, and the classification module connects the representation vectors of the NLI module that processes the semantic relationship between premises and hypotheses to produce the final result. The results trained using the Indonesian COVID-19 fact-checking dataset and COVID-19 KG Bahasa Indonesia achieve an accuracy of 0.8616, demonstrating that the knowledge graph is useful for improving the NLI performance in automatic fact-checking.
