Exploring Advanced Neural Networks For Cross-Corpus Fake News Detection

The rapid dissemination of information has become commonplace in our contemporary world, characterized by the widespread use of the internet and social media. Whether it pertains to momentous victories in cricket matches or the joyous news of a newborn, information travels swiftly and permeates the global network. However, this pervasive connectivity has also created a platform for individuals who exploit it to deceive the public by propagating false or misleading information. The consequences of such actions can be dire, potentially escalating to the point of causing significant issues, including conflicts and wars. To address this issue, this research explores the efficacy of neural networks for fake news detection which involves a comprehensive cross-corpus analysis, where advanced neural networks are trained on diverse corpora. The neural networks, particularly the BERT-based architecture, showcased exceptional proficiency, substantiated by their outstanding performance metrics. These metrics notably included an accuracy score of 0.98, a precision of 0.99, and a recall score of 0.98. Moreover, the F1-score also reached an impressive value of 0.98, affirming the robustness and effectiveness of the BERT-based neural network in accurately predicting fake news.