Hybrid Semantic and Contextual Analysis for Multilingual Fake News Detection Using Deep Learning

The growth of fake news threatens the integrity of information and public trust. This work provides a very efficient multilingual system that automatically identifies fake news by unifying standard feature extraction with deep learning and transformer-based models. Datasets of Hindi, Tamil and English are used in the system, which uses TF-IDF, NMF and LSTM for constructing semantic and sentiment-aware embeddings and a fine-tuned BERT for classification. The results are considerably high for all three languages with accuracies ranging from 97% for Hindi, 93<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup> for English and 84 % for Tamil, demonstrating its adaptability to diverse linguistic contexts. Several performance evaluations show the precision, recall and F1-score of the target of the system. This research emphasizes the necessity of sense-based and context-based approaches when dealing with the problem of fake news in a world with many languages.