Summarization of Educational Videos with Transformers Networks
Leandro Massetti Ribeiro Oliveira, Li-Chang Shuen, Allan Kássio Beckman Soares da Cruz, Carlos de Salles Soares Neto · 2023
This paper presents an approach to summarize educational videos using Deep Learning Transformers models. The approach focuses on educational content by summarizing captions and using the text results to summarize the videos. Tests were conducted using the EDUVSUM dataset, which improved upon the original paper’s results, achieving an accuracy of 26.53% in a multi-class problem, with a mean absolute error of 1.49 per video frame and 1.45 per video segment. Transformer techniques for automatic text summarization have proven effective in creating multimedia learning objects. The results suggest that these techniques can generate more efficient and high-quality digital educational resources, reducing the time and effort required for their creation.