Leveraging N-gram Features and LSTM Networks for Enhanced Fake News Detection
Inayathulla Mohammed, Chandrika Bathala, Jahnavi Chowtipalli, Kusuma Kavali, Chandrasekhar Reddy Thimmireddy, Hanuman Ramigalla
This paper presents a comprehensive analysis of fake news, examining its proliferation in the digital
age, detection methodologies, and impact on society. Through an exploration of current research and
emerging technologies, we investigate the challenges and solutions in combating misinformation. This
proposed methodology aims to develop an effective method for detecting fake news using an N-Gram
feature selection technique combined with a Long Short-Term Memory (LSTM) model. The N-Gram
approach is used to capture textual patterns and features that are key in identifying misleading or
fabricated information. The model combines natural language processing techniques with deep learning,
utilizing N-grams (unigrams, bigrams, trigrams) as features to capture contextual information. The work
involves data collection, preprocessing (tokenization, normalization), N-gram creation, and feature
selection (Term Frequency-Inverse Document Frequency, Chi-Squared tests). By leveraging linguistic
features and deep learning methodologies, this approach enhances news classification accuracy,
providing a powerful tool in the fight against misinformation. This approach not only enhances fake
news detection accuracy but also provides a scalable solution for real-time misinformation detection,
enabling timely responses to counter false information, helping to maintain the integrity of information
shared across online platforms.
