Towards efficient information retrival from large lecture capture video collections

TrackTrack 2 (Auditorium 2)
DescriptionThere is a growing trend to capture lectures given in higher education institution. Lectures on video turns out to be of great value for students in many ways. They supplement lecture notes, text books and exercises. Viewing statistics of lecture videos show active use of the video content. Students tend to watch certain parts of certain videos more often, i.e. they wind forward to find and watch again those parts where the lecturer explains very difficult and/or relevant topics.
Sharing lecture recordings openly is also a global trend helped by services like iTunes U and YouTube Edu. As UNINETT is in pilot with centralized services for lecture capture, the idea of sharing recordings on a national basis comes natural.
Given a large national collection of captured lectures combined with students desire to access specific parts of recordings, new improved techniques to access video content is of interest. In cooperation with the iAD-centre headed by Microsoft, UNINETT is developing and testing a search engine which, based on search keywords, attempts to find not only videos, but parts of the videos that have relevance. The result of a keyword search is a video production based on chuncks of videos potentially from many video sources. Hence a student may get a specific topic re-explained multiple times by different lecturers, an ideal setting for most students.
The presentation will explain how the search engine is designed and what requirements it imposes on the video content to operate efficiently. A demonstration will be given, and challenges and future work addressed.

Presentation documents

All talks