In today’s digital world, social media and the internet act as the primary sources of information. It can also contribute to information overload where individuals are overwhelmed by the vast amount of available content, making it difficult to effectively process and utilize it. About 40% of individuals believe that social media is the main source of false information, surpassing even news websites, messaging apps, and search engines, according to a 2020 Reuters Institute study. These challenges highlight the need for personalized and relevant news delivery, ensuring users receive tailored content that aligns with their preferences while minimizing exposure to misinformation. Existing news recommendation systems often rely on collaborative filtering that is affected by the cold-start problem. To overcome the existing challenges, a novel content-based news recommendation system with contextual understanding named NewsReSU was proposed. This news recommendation algorithm uses the SBERT transformer model along with cosine similarity. The model also incorporates a user feedback mechanism to enhance the system and review the quality of recommendations. The NewsReSU model considers an English news dataset with over 200,000 news articles and it achieved a precision value of 0.91. It has shown achievable performance as compared to the traditional NLP technique.