We propose a multilingual lecture-on-demand system, which searches lecture videos for segments relevant to user information needs across languages. We utilize the benefits of textbooks and audio/video data corresponding to a single lecture. We extract the audio track from a target lecture video, generate a transcription by large vocabulary continuous speech recognition, and produce a textual index. Users can view specific video segments by selecting paragraphs in the textbook for the target lecture, machined translated into the user language. Experimental results showed that by adapting speech recognition to the lecture topic, the recognition accuracy increased and the retrieval accuracy was comparable with that obtained by human transcriptions. Our system is implemented as a client-server system over the Web to facilitate e-education.
Cite as: Fujii, A., Itou, K., Ishikawa, T. (2003) LODEM: a multilingual lecture-on-demand system. Proc. ISCA Workshop on Multilingual Spoken Document Retrieval (MSDR 2003), 13-18
@inproceedings{fujii03_msdr,
author={Atsushi Fujii and Katunobu Itou and Tetsuya Ishikawa},
title={{LODEM: a multilingual lecture-on-demand system}},
year=2003,
booktitle={Proc. ISCA Workshop on Multilingual Spoken Document Retrieval (MSDR 2003)},
pages={13--18}
}