Academic Emotion Classification and Recognition Method for Large-scale Online Learning Environment-Based on A-CNN and...

Area of Science:

• Educational Psychology
• Computational Linguistics
• Human-Computer Interaction

Background:

• Subjective well-being is crucial for quality of life, with emotional measurement vital for research.
• Academic emotion, specific to education, impacts learners' well-being in online settings.
• Accurate, rapid classification of learner emotions in large-scale online environments is challenging.

Purpose of the Study:

• To develop a dimensional classification system for academic emotion aspects in online learning comments.
• To create an aspect-oriented automatic recognition method for academic emotions.
• To enhance the measurement of subjective well-being in online learning environments.

Main Methods:

• Literature analysis and data pre-analysis to build an academic emotion aspect classification system.
• Development of an aspect-oriented convolutional neural network (A-CNN).
• Implementation of a long short-term memory with attention mechanism (LSTM-ATT) for emotion classification.

Main Results:

• The A-CNN model achieved 89% accuracy on the test set.
• The LSTM-ATT model achieved 71% accuracy on the test set.
• The developed models provide quick and effective identification of academic emotions.

Conclusions:

• This research offers a novel approach for measuring large-scale online academic emotions.
• The findings support research into students' well-being within digital learning contexts.
• Automated academic emotion recognition can improve the online learning experience and support student well-being.