Application of multimedia network english listening model based on confidence learning algorithm for speech recognition

Language is the most important communication tool of human beings, and listening is one of the basic skills of language expression. Without good listening comprehension ability, it is impossible to use language flexibly to communicate. Due to the influence of traditional teaching mode, Chinese students' English listening is generally poor. Therefore, a new English listening teaching mode is needed to help students improve their English listening skills. In this study, the multimedia network technology is used to realize the integrated teaching of English listening, speaking and dictation skills, and an English listening teaching model based on multimedia network and speech recognition confidence learning algorithm is proposed. First, the mainstream confidence method based on Lattice posterior probability is optimized to improve its effectiveness. Second, the obtained confidence score is converted into a discriminant confidence score by Support Vector Machine (SVM) to enhance the discriminant ability of the confidence. Finally, a score correction strategy is proposed due to the imbalance of training data. The experiment shows that the proposed teaching model of English listening based on multimedia network technology can arouse the students’ interest and improve their listening skills. And the optimized mainstream confidence method based on Lattice posteriori probability can effectively improve the recognition ability of the algorithm and the effect of English listening classes.