Over a billion videos on YouTube are accessible to viewers who are hard of hearing or completely deaf, thanks to the video platform’s automated captions. YouTube product manager Liat Kaver has announced the milestone number in a blog post, where he also talked about how hard it was growing up as a kid who’s hard of hearing and having very little access to closed captions. After his team launched automated captions in 2009, they started concentrating on making it more available and improving its accuracy.
Kaver said they made automated captions more available to YouTubers by combining Google’s automatic speech recognition technology with the YouTube caption system. They then achieved a 50 percent leap in accuracy — for English captions, at least — by improving the service’s machine learning algorithms and expanding its training data. We all know how funny caption fails can be, but people who have no other way of knowing what was actually said would end up missing bits of information. In the image above, you’ll see an example comparing the service’s old (left) and current (right) models.
Going forward, the company aims to improve the accuracy of the 10 other languages its caption tech supports. Kaver is encouraging YouTube creators to chip in and review the accuracy of machine-generated captions for their videos, as well. After all, the more data they have on their hands, the easier it’ll be to improve the technology.