Detecting Fake News in Hindi Using Machine Learning and Deep Learning Techniques

This study investigates the detection of fake news in the Hindi language using Machine Learning (ML) and Deep Learning (DL) techniques. Hindi, a morphologically rich and widely spoken Indian language, presents unique challenges for automated text classification due to limited annotated resources and linguistic complexity. Using a dataset of 17,124 Hindi news articles sourced from credible news portals and fact-checking websites, several ML models—Logistic Regression, Support Vector Machine (SVM), Naïve Bayes, and Random Forest—were implemented with TF-IDF features. Additionally, a Long Short-Term Memory (LSTM) deep learning model was trained using tokenized sequences. Experimental results show that the LSTM model outperformed all ML models, achieving an accuracy of 83% and an AUC of 0.89, compared to 77–79% for the best-performing ML approaches. This work highlights the importance of contextual sequence-based models for Hindi fake news detection and demonstrates the limitations of traditional ML methods in capturing semantic relationships. The study concludes with recommendations for future improvements, including transformer-based architectures, multimodal datasets, and advanced language-specific resources for low-resource Indian languages.