• Mohammed I. SalihComputer Information System Department, Technical College of Zakho, Duhok Polytechnic University, Duhok, KRG, Iraq
• Salim M. MohammedComputer Science Department, College of Science, University of Zakho, Duhok, KRG, Iraq
• Asaad Kh. IbrahimComputer Information System Department, Technical College of Zakho, Duhok Polytechnic University, Duhok, KRG, Iraq
• Omar M. AhmedComputer Information System Department, Technical College of Zakho, Duhok Polytechnic University, Duhok, KRG, Iraq
• Lailan M. HajiComputer Science Department, College of Science, University of Zakho, Duhok, KRG, Iraq
  

https://www.etasr.com/index.php/ETASR/article/view/10625

This study focuses on improving automated news classification using a fine-tuned BERT (Bidirectional Encoder Representations from Transformers) model. The authors applied transfer learning on the Reuters-21578 dataset and compared the performance of BERT with traditional machine learning models such as Naive Bayes, SVM, and Random Forest.

The results show that the fine-tuned BERT model achieved 91.77% accuracy, significantly outperforming both classical models and non-fine-tuned BERT. The study concludes that fine-tuning allows BERT to better understand contextual relationships in text, making it highly effective for modern NLP tasks like news classification.