Rumor detection in networks based on the BERT-CNN approach

Abstract

As of June 2023, China has witnessed a surge in internet users, reaching 1.079 billion, with an internet penetration rate of 76.4%. This rapid expansion of social media and online communication channels has not only enhanced the ease of information sharing but has also intensified the spread of rumors and misinformation. This trend poses substantial threats to social order and public safety. In response to this challenge, this work introduces an innovative approach to detect online rumors, utilizing the BERT-CNN methodology. By harnessing the contextual understanding capabilities of the BERT model and the efficient feature extraction prowess of CNN models, the study has crafted a framework that seamlessly integrates these technologies. This integrated approach proves effective in accurately identifying and categorizing rumors in online data. Through extensive experiments conducted on various real-world datasets, a comparative performance analysis had conducted against existing technologies. The results reveal significant enhancements in key metrics such as accuracy, recall, and F1 score when compared to traditional methods. This affirms the efficacy of the BERT-CNN approach in the domain of online rumor detection. This research not only introduces a fresh perspective to rumor detection technology but also furnishes a robust tool to tackle the challenges associated with the dissemination of rumors in the digital age.