Combining the 1D CNN Model and Transfer Learning Model in Analyzing Learner Emotions

Sentiment analysis is the process of collecting and analyzing individual opinions and thoughts related to various topics, products, and services. However, for Vietnamese, the number of published datasets is still limited, especially in the field of education. This study used two datasets from student feedback to analyze learners’ sentiments towards lecturers and courses. The first dataset was collected at HUTECH University in the past academic years. The second dataset is UIT-VSFC (feedback from students at the University of Information Technology, Vietnam National University, Ho Chi Minh City). The study used a 1D CNN deep learning model and data balancing solutions, word extraction… And a transfer learning model to analyze learners’ sentiments (6 levels of emotions). With the first dataset, the results were obtained when using the CNN model with an accuracy index (ACC) of about 72%, and after transfer learning, it was about 85%. The second dataset obtained by ACC has a fairly high accuracy. From this result, relevant organizations can apply it to deploy, collect, and analyze user sentiments, thereby having an overview and appropriate planning or as a basis for making appropriate and timely management decisions.