March 15, 2010

Murray Stokely has completed his captioning project, which was funded by the FreeBSD Foundation, and provides the following update:

A pilot project to improve the machine generated captions of technical conference lectures from the BSD Conferences YouTube channel has been completed. The 73 videos in this channel have been viewed over 200,000 times since the channel launched in late 2008, and the addition of human-edited transcripts to some of our most popular videos makes this content more accessible to people around the world.

In addition to the benefits to the hearing impaired, captions are very useful for international viewers as well as for the improved discoverability of this content by search engines. The improved quality of the English language transcripts also improves the quality of the automated translation of the captions into over 45 different languages. It is also now possible to search for words and phrases in the audio transcripts and get a link directly to videos that contain spoken content of that word or phrase.

For example, try searching for a famous line from one of Dr. Kirk McKusick’s FreeBSD Kernel Internal Lectures. The above link will take you to the Google Video Search Result page where one of Dr. McKusick’s lectures containing the phrase as long as dinosaurs and mainframes is the first result, along with a snippet of the transcript from his lecture, just as you would see the snippet from text content on a web page. A dozen of our most popular videos of FreeBSD technical content are now captioned and fully indexed allowing users to search for very technical terms and get access to lecture material from BSD Conferences.

The captions were improved by two passes of human editing paid for hire through Amazon Mechanical Turk.