<p>This research presents an intelligent tutoring system (ITS) designed for the adaptive learning of Business English idioms. It addresses the problem of limited personalization and static content delivery in traditional language learning platforms. The objective is to design a hybrid ITS that adapts to individual learner performance, knowledge level, and emotional feedback. The proposed system integrates rule-based classification, ontology-driven knowledge structuring, semantic similarity algorithms, and sentiment analysis using a BERT-based deep learning model. Learners are initially categorized using pre-assessment rules, and idioms are then recommended based on difficulty, domain relevance, and previous learning outcomes. Feedback is analyzed in real time to guide dialogue and content adaptation. The methodology was evaluated through controlled user interaction scenarios. Results indicate a 25% improvement in quiz performance and a 30% increase in learner engagement, compared to a baseline non-adaptive version. Learners received personalized recommendations, adaptive quizzes, and emotionally responsive system messages, improving motivation and retention. However, limitations include the scalability of rule-based logic, difficulties in culturally interpreting idioms, and occasional inaccuracies in sentiment detection. Future work will explore reinforcement learning for dynamic adaptation and multilingual support for broader applicability. In conclusion, the system demonstrates that combining symbolic and deep learning techniques in ITS significantly enhances the learning experience. It offers a replicable model for personalized, intelligent, and emotionally aware instruction in domain-specific language learning.</p>