An Intelligent Enhancement of an English Interpretation Online Teaching System based on Deep Learning Algorithms

Online English interpretation education is gaining more importance as students seek flexible, interactive, and accessible language education. Accurate speech recognition and interpretation remain challenging, especially with diverse accents and spontaneous speech patterns. This research presents an online English interpretation system based on Deep Learning (DL), employing Wav2Vec 2.0 for robust speech representation learning and a Recurrent Neural Network (RNN) for capturing temporal dependencies in speech signals to classify interpretation proficiency levels. The Mozilla Common Voice (English subset) dataset, containing diverse real-world speech samples, is used to train and evaluate the model, supporting strong generalization across different speakers and speaking styles. Wav2Vec 2.0 learns high level contextual features from raw audio, and supplies high quality input to the downstream processing. The learned features are used as input to the RNN, which is temporally consistent and predicts the appropriate interpretation level. The end-to-end architecture facilitates real-time processing and adaptive responsiveness in online teaching scenarios. The model achieves a Word Error Rate (WER) of 6.9%, an Accuracy of 98.7%, a Precision of 97.7%, a Recall of 97.1%, and an F1-score of 96.9% on interpretation level classification. These results indicate the model's capability to provide accurate, real-time interpretation and facilitate learner comprehension in online English language learning.