Recommendation systems are essential to improving user experience throughout e-commerce, media, and many other sites.—However, traditional recommendation approaches typically do not capture the complex and contextualized interests expressed in textual data, including reviews. Therefore, this study presents a hybrid recommendation framework which synergizes sentiment analysis and LDA-based topic modeling for enhancing recommendation quality. Sentiment analysis extracts the emotional tones of users’ reviews and LDA extracts the latent thematic interests of users which are then used to enrich the user-item interaction matrix. They showed significant improvements over the baseline models, with a 12.7% decrease in RMSE and a 12.1% increase in NDCG@10 using a large-scale e-commerce dataset. The results demonstrate the importance of leveraging both sentiment and topic features in order to provide accurate and interpretable recommendations. This work offers a general and scalable solution for enhancing recommender systems, with wide-ranging impact across user-oriented domains.