This study investigates real-time sentiment analysis of geolocated category-based tweets using a pre-trained language model called Dynamic Content Routing (DCR). Using Twitter’s streaming API and keyword filtering, we collect and analyze sentiment during live category-based events. DCR eliminates the need for extensive training and provides an efficient and scalable approach. We evaluate the effectiveness of DCR to capture the nuances of location-based atmospheres without domain adaptation. Focusing on NLP techniques such as sentiment analysis and pre-trained language models, this research helps understand social media sentiment in real time. The framework highlights the potential of DCR and geolocation filtering for broader social media applications.