Classificação automática de estilos de videoaulas

Abstract

Although video lessons are often used in diverse areas that cover a wide range of studies and applications, the lack of a common approach to defining and classifying their styles results in the use of many different models for these purposes. There are different proposals for classifying these styles, but with differences between them. Consolidating production styles makes it possible to understand the possibilities of producing materials and facilitates communication between researchers or content producers. To develop this common approach, there is a need to build a framework through which these styles can be defined and classified. In addition, much has been done to investigate the effects of these styles on student engagement and learning outcome. These studies suggest that video lessons styles affect academic performance and that students learn better through a certain video lesson style compared to others. Based on this, the objectives of this work are to propose a unified model for classifying styles of video lessons based on the nomenclatures and definitions used in the literature also, to propose an approach for automatic classification of 4 video lesson styles (Talking Head, Voice Over Slides, Presentation Style and Khan-Style) using visual characteristics of these styles. The proposed classification model allows classifying video styles based on 2 visual dimensions which are human embedding and instructional media. It is believed that this model will make possible the correct characterization and formation of a common understanding of definitions of video lessons styles based on existing scientific studies. The automatic classification of styles can be used by recommendation systems to suggest styles that are more consistent with student preferences and the intended learning outcome. The approach presented for the automatic classification shows that the features extracted from the videos are capable of accurately classifying the set of styles presented in this study and this classification is carried out through simple and easy-to-extract features.