An Al-BERT-Bi-GRU-LDA algorithm for negative sentiment analysis on Bilibili comments.

Area of Science:

• Educational Technology
• Artificial Intelligence in Education
• Natural Language Processing

Background:

• Increasing adoption of online self-learning and lifelong education programs.
• Student commentary provides valuable insights into learning experiences, opinions, and complaints.
• Need for effective sentiment analysis tools for educational feedback.

Purpose of the Study:

• To develop and evaluate a hybrid algorithm for sentiment analysis of student online learning feedback.
• To identify key themes and issues from student comments on high school mathematics courses.
• To improve curriculum and teaching quality based on student feedback.

Main Methods:

• Utilized a dataset of student comments from 10 high school mathematics courses on the Bilibili platform.
• Constructed a hybrid algorithm: Artificial Intelligence-Bidirectional Encoder Representations from Transformers (BERT) + Bidirectional Gated Recurrent Unit (BiGRU) and Linear Discriminant Analysis (LDA).
• Conducted comparative analysis with alternative algorithms on educational review datasets.

Main Results:

• The proposed hybrid algorithm demonstrated higher precision and lower loss rates compared to other algorithms.
• The algorithm maintained a consistently low loss ratio during testing.
• Topic mining of negative comments revealed issues with messy content, course complexity, and video quality.

Conclusions:

• The AI-BERT+BiGRU+LDA hybrid algorithm is effective for sentiment analysis in online education.
• Student feedback highlights areas for improvement in course content, delivery, and video production.
• The findings provide a foundation for addressing student concerns and enhancing educational offerings.