Automatic Teaching Plan Grading with Distilled Multimodal Education Knowledge

With the development of artificial intelligence, more and more technologies are used in teaching design. The teaching plan is the direct guidance document of teaching design. Through the teaching plan, teachers can effectively evaluate their teaching design level to improve teaching methods and promote improving students’ core literacy. Current qualitative research and manual coding are challenging to analyze for the teaching plan grading. There are factors such as subjectivity and time-consuming in manual grading. Hence, the manual grading of teaching plans is irrational to some extent. The evaluation process of manual grading needs to be further strengthened and improved. This study adopts the deep learning method and proposes a multimodal education knowledge distillation model (MEKD) to analyze multimodal data to solve the above problems. In this study, the experimental analysis of more than 50,000 teaching plans verifies the MEKD's effectiveness in helping alleviate the problem of automatic teaching plan grading.