This repo includes a program that takes a lecture video as an input, and outputs the technical key words that appear in the lecture video. These keywords can be used to provide useful hints to the speech recognition of that video lecture. Note that this program is specifically designed for the speech-to-text recognition system in ClassTranscribe, an educational delivery platform that allows instructors to post lecture videos. However, this program can also be used in the general purpose of extracting technical keywords from lecture videos.
This project is jointly completed by Jiaxi Li, Akhil Vyas, Ninghan Zhong, and Tianhui Cai.
To text the quality of the extracted keywords from lecture videos, we run the Microsoft Azure Speech-to-Text recognition system on the input lectures videos and check how recognition accuracy is changed with/without the extracted keywords from out program.