📂 Source code is here.
✨ Note: This project is part of my 🎓 M.Sc. Thesis dissertation
-
🛡️ Hybrid Architecture: Combines Transformer and Bi-LSTM architectures to maximize the strengths of both for more accurate fake news detection.
-
🧠 Advanced Tokenization: Utilizes BERTweet tokenization, which enhances the model's ability to understand context and detect nuanced misinformation in tweets.
-
🔄 Ablation Study: Comprehensive analysis of individual and combined contributions of the Bi-LSTM and Transformer components to show performance improvement.
-
📊 High Accuracy: Achieves state-of-the-art results on multiple datasets with superior performance compared to existing models in the field.
-
📈 Benchmarking: Extensive comparative analysis against traditional models like CNN, LSTM, and Transformer architectures to highlight superior performance.
-
🗂️ Robust Text Preprocessing: Incorporates a powerful text-cleaning and standardization pipeline to prepare Twitter data for effective classification.
-
📉 Real-time Detection: Capable of detecting fake news in real-time across a variety of informational settings.
-
📁 Cross-Dataset Validation: Demonstrated high adaptability and robustness by testing across multiple fake news detection datasets, ensuring its reliability in diverse scenarios.
To install, run the following command:
git clone https://github.com/kowshik14/FakeNewsDetection-TweetGuard