This study addresses the limitation of traditional dynamic pricing models in e-commerce by developing a novel, hybrid Sentiment-Aware Dynamic Pricing (SADP) framework that integrates real-time customer sentiment alongside core transactional and competitor features. A comprehensive, multimodal dataset, including multilingual customer reviews, was subjected to a robust preprocessing pipeline (including SMOTE for imbalance handling) and extensive feature engineering (e.g., competitor price difference, estimated price elasticity of demand). Multiple advanced machine learning models were trained and rigorously evaluated using a Bayesian Optimization strategy and Time Series Cross-Validation. The XGBoost model significantly outperformed all competitors, achieving superior metrics (R2: 0.97, MAE: 1.29, RMSE: 1.65). Crucially, the integration of sentiment features was associated with a quantifiable improvement in prediction accuracy compared to models using only numerical data, demonstrating the ability to capture emotional drivers of purchasing behavior. Both XGBoost and Neural Networks demonstrated low latency, confirming their suitability for real-time, scalable deployment in live e-commerce pricing engines. This research presents one of the first empirically validated dynamic pricing frameworks to successfully integrate sentiment analysis for enhanced predictive accuracy, offering a proven, scalable architecture for next-generation revenue management.