The primary goal of this research is to identify more precise and enhanced methodologies for the use of transformer-based models in the classification of news and the identification of topics on digital news platforms. The research investigates the utilization of advanced deep learning architectures, including BERT, RoBERTa, and DistilBERT, to classify news and identify the underlying themes within extensive text corpora. The semantic relationships and context of news articles are occasionally difficult to understand using conventional machine learning methods. In contrast, transformer models implement attention mechanisms to improve the understanding of complex linguistic patterns. The proposed system includes data preprocessing, tokenization, feature extraction, model training, and evaluation, which are assessed using accuracy, precision, recall, and F1-score as performance metrics. Experimental results have shown that transform-based methodologies substantially improve the precision of topic identification and classification efficacy in comparison to conventional techniques. This work facilitates the organization of content, the eradication of misinformation, personalized recommendations, and real-time information management, thereby contributing to the development of intelligent news analysis systems in contemporary media applications.