Emotional Dynamic Modeling in Student Collaborative Learning Based on Natural Language Processing

Read full text Summarise with VIRAT
Authors: Xiaoxia Hao
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Aiming at the problem of emotional dynamic modeling in student collaborative learning, this study proposes a hierarchical attention LSTM-Transformer fusion algorithm (HALTF).The algorithm extracts text features through bidirectional LSTM, the hierarchical attention mechanism focuses on key emotional information, and the Transformer achieves accurate prediction.The experiment collected 5,000 sets of online collaborative learning text data, and carried out comparative experiments after preprocessing.The results show that the HALTF algorithm performs well in the emotion classification task, with an accuracy of 89.6%, an increase of 12.3% compared with the traditional LSTM algorithm; the F1 value is 88.2%, an increase of 9.5% compared with the Transformer algorithm; and the AUC reaches 0.92.The ablation experiment shows that the hierarchical attention mechanism, LSTM layer and Transformer layer are crucial to the improvement of algorithm performance.After removing the hierarchical attention mechanism, the accuracy drops to 81.2%.The algorithm effectively overcomes the problems of weak generalization ability and insufficient dynamic capture of traditional models, and provides an innovative solution for emotion analysis in the field of education.

Fields of Study & Subjects

Sentiment Analysis and Opinion Mining Emotion and Mood Recognition Online Learning and Analytics

Publication details

Journal / source
Advances in Social Science, Education and Humanities Research/Advances in social science, education and humanities research
Pages
453–461
Publication date
25 March 2026
Language
EN