This study tackles challenges in evaluating and improving broadcasting and hosting teaching by proposing an intelligent analysis method using deep learning models. It collects multisource data like speech, text, and feedback to build a multi-dimensional feature system. Combining LSTM networks with natural language processing, a predictive model for automatic expression quality scoring is designed. The system tracks individual performance, generating structured evaluations and personalized improvement suggestions. The model showed good accuracy and stability during experiments, and practical tests demonstrated significant student improvements in speech rate, emotional expression, and coherence. Teachers gave positive feedback on the system’s support for teaching. The study enhances personalized teaching strategies and shows that deep learning-based analysis has strong adaptability and potential for broadcasting and hosting courses.
Xiaoyan Zhang (Fri,) studied this question.