<![CDATA[In today's digital era, easy access to information is characterized by the rise of online news portals that present hundreds of articles every day. However, the high volume of news available often makes it difficult for readers to identify the essence of various articles on similar topics. Reading articles one by one becomes a time-consuming activity and risks causing information overload. To address these issues, this research develops an automated system that is able to summarize news based on certain topics by utilizing the Relevance-Augmented Generation (RAG) approach reinforced by the use of Large Language Models (LLMs). The system is built using Python programming language and integrated with Flask framework as a web interface. Data collection is done through a scraping process from the Detiknews portal using a special API. The articles obtained were then analyzed using natural language processing (NLP) techniques, including evaluation of sentence length, sentence position in the article, as well as the frequency of keyword occurrence to determine the most relevant sentences. The initial summary generated is further refined with the help of the LLMs model through the Groq API. The implementation results show that the system is able to present information-dense, accurate, and efficient summaries. The summary makes it easy for users to get the gist of the news quickly without losing the main context. Thus, this system provides a solution to the challenges in online news information processing while increasing the ease of access to information for readers.]]>
