A lecture-on-demand system using spoken document retrieval

摘要

This research proposes a lecture-on-demand system, which retrieves video data in response to users' information needs. For this purpose, we utilize texts and audio/video data for a single lecture. Our system extracts audio tracks from lecture video data, transcribes them by way of a large vocabulary continuous speech recognition system, and produces a lecture database. Users can selectively browse a specific video track by submitting textual queries, i.e., keywords, sentences and paragraphs associated with the text for the target lecture. Our current system is implemented over the Web, so as to facilitate e-education.