Ever since YouTube’s official auto-captioning feature launch in early March 2010 (with a private beta back in November 2009), I can’t stop thinking about the enormous significance for OER and for the web in general. It’s huge.
As you probably know, David Wiley was recently able to convince Sal Khan of the Khan Academy to start using a CC-BY license on his 1200+ educational videos. Since all of Sal’s Open Educational Resources (yes, we can call it OER now) are delivered via YouTube they benefit from Google’s yummy, new auto-transcription and auto-translation features. I have experimented with viewing Sal’s video mini-lectures with both English and French captions. The English audio transcription seems fairly accurate, and the French translation (auto-generated from the English subtitles) has a little more that is “lost in auto-translation.” This definitely deserves it’s “beta” label, but it is impressive nonetheless when you consider it is all auto-generated.
So what does video auto-captioning mean for OER?
- accessibility (readable by someone with a screen reader or braille output device)
- discoverability (easy for others to find you via Google, etc.)
- searchability (easy to find the specific part you are looking for – i. e. searchable video)
- low-bandwidth access (if text transcripts are made available separately)
- translateability (I think I made that word up)
Once you have an easy way to transcribe video content, several new possibilities open up. While not perfect, the auto-captioned content is definitely more accessible to users with visual and/or aural disabilities. (I just hope future iterations of auto-captioning will allow the content author to invite/approve users to edit these auto-captions, similar to the dotSub model.) But increased accessibility is only one way we can benefit. Now that you have captions, you can search them. Auto-captioning will make it possible to find a particular video on the web, or even a particular segment within a video using a keyword search. Soon you will “Google” through video for a particular scene. MIT is already doing video search (see their Lecture Browser and Spoken Media Project). Finally, raw text video transcriptions use less bandwidth than the original video content, which meets another critical need: access in low-bandwidth areas or places where the cost of bandwidth can be prohibitive. I’ll stop there for now, but it is clear to me that with the explosion of online videos and related rich media, video auto-captioning is a major step forward for the web.