To effectively detect and filter out spam or irrelevant news articles in real-time, a combination of a rule-based filtering system and a machine learning model should be used. The rule-based system can quickly filter out obvious spam or irrelevant articles based on predefined keywords or patterns. The machine learning model can then further analyze the content to classify articles as relevant or irrelevant with higher accuracy. This approach ensures that the solution can process news articles in real-time while maintaining high-quality content.