On social platforms like Twitter (now, X), users frequently share opinions on daily matters, sometimes aggressively due to online anonymity. These interactions may lead to “controversies" where differing views on a topic arise. This paper builds on an existing controversy detection framework that relies on interaction dynamics within trending topics. We enhance this framework by incorporating content-based features, such as the distribution of URLs, hashtags, and mentions, to assess whether these additions improve model performance. To evaluate the impact of both interaction dynamics and content-based features, we trained three classifiers-a decision tree, an SVM, and a random forest-using two distinct feature spaces. Models trained solely with interaction dynamics served as the baseline, while models with the extended feature space demonstrated the effect of content-based features on performance. Our experiments reveal that integrating content-based features significantly enhances model performance across nearly all metrics on the tested tweet datasets, outperforming baseline models and demonstrating the value of our hybrid approach.