Maithi-Net: A Customized Convolution Approach for Fake News Detection using Maithili Language

Abstract

Online news consumers now face serious difficulties due to the widespread distribution of fake news on social media platforms. To distinguish fake news from real, the paper suggested a customized CNN model named "Maithi-Net". The model being suggested is composed of five convolution layers which are capable of automatically acquiring the distinguishing features essential for identifying fake news. Both the CGU-Maithili and ISOT fake news datasets have been used to successfully validate the proposed model. The efficacy of the model is verified with several evolution metrics like accuracy, specificity, sensitivity and F1 score. The model provides the detection accuracy 96.85 % for CGU-Maithili and 97.28% for ISOT fake news datasets. The experimental results show substantial gains over prior state-of-the-art results in the area of fake news detection and validate the potential of our method for categorising misinformation spread via social media.