Hybrid FastText-LSTM for Fake News Detection: A Multilingual Approach with a Focus on Kurdish and English

The spread of mis/disinformation, also known as fake news has become a paramount challenge in digital media, thereby proving the need for efficient machine learning (ML) and deep learning (DL) models. This study examines different models, including classical ML methods like Support Vector Machines (SVM) and Logistic Regression (LR), as well as DL strategies like Long Short-Term Memory (LSTM). The proposed model is Hybrid FastText and LSTM, which uses the Kurdish Fake News Dataset (KDFND), with balanced data and met with a remarkable data accuracy of 94.25% for the overall Kurdish data and 92.11% for English. Results emphasize the importance of ensemble and hybrid methods for improving fake news classification and can serve as a baseline for low-resource languages such as Kurdish. Future Work — Other potential future directions may include using transformer-based architectures and multi-lingual fake news detection (FND) techniques.