Interactive approaches to video lecture assessment

Meer over het boek

A growing number of educational institutions record videos of classes to provide additional study resources. However, these recordings often lack the editorial polish of well-prepared educational videos. Issues such as prolonged silence, unnecessary repetitions, spontaneous student interactions, and corrections of mistakes are common. Unlike scripted and edited educational videos, these plain recordings typically lack summaries or tables of contents, making them less effective as e-learning resources. This thesis presents a system designed to bridge the gap between simple video recordings and valuable e-learning tools by generating automatic summaries and offering an interactive lecture browser. This browser visualizes automatically extracted key phrases and their significance on an augmented timeline. The system relies on four main tasks: automatic speech recognition, extraction and ranking of key phrases, extractive speech summarization, and visualization of phrase salience. Each task is thoroughly described and evaluated using a newly acquired corpus of academic spoken English, the LMELectures. A preliminary user study indicates that students using the lecture browser can complete a topic localization task approximately 29% faster than those relying solely on the video.