A computational approach for real-time detection of fake news

Fake news is a severe organizational and societal problem as social media further aggravates its spread. Detecting Fake news in real-time is a critical for tackling this challenging scientific problem as it can help stem its spread and consumption quickly. In this study, we propose a computational approach to detect fake news in a real-time manner. Our proposed method leverages event and topic extraction techniques coupled with a topic merging mechanism to process news data and reduce the number of topics. This approach incorporates a two-stage procedure to optimize the cold-start ratio between initial data batches and other ones to improve memory management during processing streaming data. We conduct various computational experiments in different system settings for benchmarking the proposed methodology. Findings of this study suggest that the proposed approach demonstrates takes less time in detecting fake news and reduces the number of topics by 19.76% and the number of data clusters by 26.92% while comparing with other baselines.