Research In Semantic Annotation On Video-Recordings Of Classroom Teaching

Lecture videos are important media for dissemination of teaching behavior and skills in the education field. When used as teaching cases for professional development, lecture videos are mainly focused on its high-level semantic. The high-level semantic is subjective feelings on teaching behavior and skills. The existing metadata standards and semantic technology both have difficulty in describing high-level semantic of lecture videos. The existing video management systems seldom support searching teaching skills from lots of lecture videos for teachers'/pre-service teachers'training or education. According to pedagogy and education evaluation knowledge, a semantic annotation model and video annotation tool have been constructed to describe the high-level semantic of lecture videos. The video annotation tool support teachers/ pre-service teachers to observe and annotate their own or peers'lecture videos. The annotation data help improve searching efficiency for video-recordings of classroom teaching.Firstly, we survey existing theories and technologies of semantic annotation, especially for video annotation. Based on the practical requirement of classroom teaching, the tasks of semantic annotation are proposed for the video-recordings.Secondly, videos'information features are analyzed from the points of video structure and content, and a semantic hierarchical model is proposed for lecture videos. Teaching behavior and skills are the knowledge of pedagogy and education. Based on the education theory and classroom observation & evaluation knowledge, four classes (teaching behavior, teaching skill, teaching process and summative evaluation) are designed to construct and describe the semantic model. The metadata has been defined based on existing metadata standards. According to the extended requirements, the formal expression and evolution strategy of semantic model are presented.And then, we design the annotation strategies, functions, structures and flows of lecture video annotation tool. The annotation prototype is implemented and the annotation database is designed based on multiple-table method.Finally, the semantic model, annotation data and annotation tool were tested by investigation and experiment. We applied correlation and partial-correlation analysis to test the correlations of subject or teachers'working age and teaching skills, teaching behaviors, teaching links and classroom teaching types. The correlation analysis results show that the classes and properties of semantic model are universal. Some teachers and pre-service teachers were invited to use the annotation tool to test its usability and its application context. Search effectiveness and satisfactoriness are also tested in different search ranges. The experiment shows that the annotation tool supports web-based classroom teaching observation and evaluation, and annotation data is helpful in searching high-level semantic of lecture videos.