<![CDATA[In the digital news production process, editorial teams face challenges in manually categorizing news articles and generating summaries, which are time-consuming, inefficient, and prone to inconsistencies that affect content management quality and Search Engine Optimization (SEO). This study aims to design and develop an automated system for news classification and summarization on the radartegal.com platform using the Transformer-based IndoBERT model for automatic news category classification and the Large Language Model (LLM) LLaMA 3 for abstractive text summarization. The research methodology consisted of problem identification, literature review, dataset collection, text preprocessing, IndoBERT fine-tuning, LLaMA 3 implementation, pipeline integration, and system evaluation. Classification performance was evaluated using accuracy, precision, recall, and F1-score, while summarization quality was evaluated using ROUGE metrics. Experimental results showed that the IndoBERT model achieved an accuracy of 84.00%, precision of 84.58%, recall of 84.00%, and F1-score of 83.95%. Meanwhile, the LLaMA 3 summarization module achieved ROUGE-1, ROUGE-2, and ROUGE-L scores of 0.4672, 0.2732, and 0.4122, respectively. The integrated system successfully automated editorial workflows, improved categorization consistency, generated informative summaries, and supported SEO optimization. These findings demonstrate that the proposed system can improve editorial efficiency while maintaining content quality in local digital news publishing.]]>
Copyrights © 2026
