Media consumption patterns have increasingly transitioned to a multi-screen paradigm, where, through multitasking, viewers can search for additional information about the content they are engaging with and share their perspectives with others. This paper focuses on the sentiment classification of tweets related to live sports events published during their broadcast. By analysing these tweets, it becomes possible to understand audience sentiment throughout the event and provide insights to content creators, production teams, and other stakeholders about audience perspectives. Additionally, complementing existing sentiment analysis approaches, the study demonstrates that incorporating Wikipedia knowledge graph embeddings can enhance sentiment classification performance. The proposed approach achieves an accuracy of 94.7% while processing and classifying over 500 tweets per second, making it both accurate and scalable. The conclusions and methodologies outlined can have broader applicability across domains such as e-commerce, where customer sentiment analysis plays a critical role.