We present a two-step method for topic-specific sentiment and emotion scoring by integrating topic modeling and sentiment analysis. We apply BERTopic to a collection of 4.94 million Italian X (formerly Twitter) posts from July 2024 to February 2025 to extract coherent topic representations. Subsequently, we perform sentiment analysis using pre-trained Transformer-based models to assess sentiment polarity and emotions within each identified topic. We focus on international geopolitics discourse to show the potential of this approach in improving the understanding of social media activity compared to standard sentiment analysis and topic modeling. We find interesting trends, particularly in the relationship between topics regarding European Union defense strategies and international conflicts.