Student engagement assessment using multimodal deep learning.

Area of Science:

• Educational Technology
• Artificial Intelligence
• Computer Vision

Background:

• Student engagement is crucial for academic success and effective pedagogy.
• Accurate assessment of student engagement is challenging, especially with diverse data sources.
• Existing methods may not fully capture the nuances of student engagement.

Purpose of the Study:

• To propose a novel multimodal deep learning framework for accurate student engagement assessment.
• To develop a method integrating video, text, and log data for engagement evaluation.
• To validate the framework's effectiveness and practicality in real-world educational settings.

Main Methods:

• Utilized a multimodal deep learning framework incorporating video, text, and log data.
• Implemented engagement indicator extraction and asynchronous data fusion techniques.
• Employed deep learning models and gradient magnitude mapping for engagement level evaluation.
• Explored the application of deep Convolutional Neural Network (CNN) models.

Main Results:

• The multimodal framework demonstrated effectiveness in assessing student engagement.
• The proposed method accurately quantified engagement levels, distinguishing subtle differences.
• Empirical studies validated the reliability of the engagement quantification results using statistical methods.
• Deep CNN models showed applicability within the developed framework.

Conclusions:

• The multimodal asynchronous data fusion and deep learning approach is effective for student engagement assessment.
• The framework offers a practical solution for enhancing teaching methods and student performance.
• This research contributes a robust method for nuanced and reliable engagement quantification.