Abstract
Currently, the development of deep learning (DL) technology provides new avenues for educational sentiment analysis. Building on this, this study introduces an Attention Feature Convolutional Neural Network (AFCNN) model for sentiment recognition in university foreign language classrooms. The model captures students' emotional changes in real time, providing teachers with timely instructional feedback and supporting their professional development. To validate the model, comparative experiments are conducted with traditional models, including Visual Geometry Group 16 (VGG16) and ResNet18. The results show that AFCNN outperforms these models in recognition accuracy and maintains robustness under occlusion. Specifically, this study studies the professional development path of foreign language teachers based on the above background. First, this study explains the direction of professional development of foreign language teachers in the context of organizational change and DL. Second, the influence of students' emotions on education and teaching during the course of class is explained. Foreign language teachers can determine students' emotional changes in time and make corresponding positive feedback by establishing a model. Finally, the AFCNN model is established through the DL network to recognize students' faces in the classroom and feedback to the foreign language teachers. Comparative experiments are carried out to verify the effectiveness and rationality of the proposed model. The experimental results show that when there is no obstruction, the emotion recognition ability of the AFCNN model is significantly higher than that of the traditional recognition model, which can reach up to 81%. The recognition rate of the two emotions of being happy and neutral is better. During the lesson, the student's face capture will be occluded, which may reduce the recognition accuracy of the model, so the image will be occluded again. Finally, it is compared with the traditional model. It is found that both the traditional model and the present model are affected. However, the recognition rate of the AFCNN model is still the highest, so the effectiveness of the model can be verified. In summary, this study proposes a professional development pathway for foreign language teachers that integrates psychological sentiment analysis. It also introduces a new perspective for cross-disciplinary research linking educational technology with human-computer interaction.