Enhancing Fake News Detection with a Hybrid of Bert and Naive Bayes

This paper introduces the development and assessment of a hybrid model that integrates BERT (Bidirectional Encoder Representations from Transformers) and Naive Bayes for detecting fake news. The model utilizes BERT's deep contextual embeddings and Naive Bayes effective classification to improve detection accuracy and lower computational complexity. The paper highlights the performance of the model on the ISOT Fake News Dataset, testing the accuracy, precision, recall, and F1 score metrics, along with the efficiency of the model in the training and prediction phases. The results show that the hybrid model performs better than single models, with a high accuracy of 98.3 % and low latency inference of 0.02 s, offering a strong solution to fake news detection challenges. Lastly, the study claims the importance of integrating practical natural language processing methods to enhance misinformation management in real-world applications.