FactCheckPro: Automated Fact-Checking Using BERT to Classify the Veracity of Claims

A fact-checking model is presented, utilizing Natural Language Processing (NLP) to examine various claims and sorts them into distinct levels of verification, namely, False, Mostly False, Mostly True, True, or Unverified/Mixed. For its claim verification model, which utilizes the ’multi fc’ repository’s data, BERT (Bidirectional Encoder Representations from Transformers), which is regarded as one of the best NLP model available, is used. The others are data scrapping, label adjustment, and tokenization, which are done to enlarge the dataset and allow for better model classification of features. Based on the evaluation results, where an overall accuracy of 41% was achieved, the performance is considered to be satisfactory. The tag distribution resulted in discrepancies between precision and recall but given the outcome, it shows that the model is suitable for automated fact checking processes but requires some level of improvement.