Emotion recognition and achievement prediction for foreign language learners under the background of network teaching

Abstract

At present, there are so many learners in online classroom that teachers cannot master the learning situation of each student comprehensively and in real time. Therefore, this paper first constructs a multimodal emotion recognition (ER) model based on CNN-BiGRU. Through the feature extraction of video and voice information, combined with temporal attention mechanism, the attention distribution of each modal information at different times is calculated in real time. In addition, based on the recognition of learners' emotions, a prediction model of learners' achievement based on emotional state assessment is proposed. C4.5 algorithm is used to predict students' academic achievement in the multi-polarized emotional state, and the relationship between confusion and academic achievement is further explored. The experimental results show that the proposed multi-scale self-attention layer and multi-modal fusion layer can improve the achievement of ER task; moreover, there is a strong correlation between students' confusion and foreign language achievement. Finally, the model can accurately and continuously observe students' learning emotion and state, which provides a new idea for the reform of education modernization.