Captioning for Classroom Lecture Videos

Abstract

Tablet PC-based lecture videos are widely used by students at the University of Houston. The goal of the ICS (Indexed, Captioned, and Searchable) Videos Project is to enhance user experience and improve usability of classroom lecture videos. The goal of this thesis is to assess the value as perceived by students, of having captions for classroom lectures and to find methods of generating captions as automatically and efficiently as possible. In this thesis, we established through survey and focus groups, that students highly valued having captions for their classroom lectures. The accuracy of currently available speech recognition tools was assessed with the assistance of three faculty members. This revealed that the output of these tools for spontaneous speech cannot be used directly as captions due to its low accuracy level. This text needs to be manually corrected. To edit these automatically generated captions, we designed, implemented, and evaluated a web-based crowd-sourcing caption editing tool. This tool can be used by a group of people to correct the captions simultaneously. This reduces the time and effort required to correct the captions, which otherwise is a monotonous task. We evaluated the caption editor through a study and a survey with 28 students from two classes. The students in the class were able to effectively combine their efforts such that a full class lecture was captioned with a very modest effort from each student. In this thesis we established a semi-automatic process to generate corrected captions for classroom lectures efficiently. The captioning module is integrated with the ICS Videos player, thereby allowing navigating the videos from text captions. Focus groups and other research show that captions and transcripts are highly valued by students, especially those with hearing disability and whose first language is not English.