Just eight years after YouTube launched its automatic captioning system, 1 billion videos on the site have subtitles, an accomplishment that showcases the vast strides that its parent company Google has made in transcription technology.
YouTube announced the milestone in February, noting that there are more than 15 million views of its captioned videos every day, providing an indispensable service to hearing-impaired users, according to TechCrunch. The company added that its technology is approaching a level of accuracy equal to that of human transcribers.
It would be hard to imagine YouTube attaining such a successful and large-scale deployment of automatic captioning when it started the service in 2009. The initial version of the system was infamous for producing mangled transcriptions. These malapropisms occurred so frequently that the phrase “YouTube Caption Fail” actually became an internet meme.
YouTube itself was well aware of the shortfalls of system, with the company blog observing, “There were limitations with the technology that underscored the need to improve the captions themselves. Results were sometimes less than perfect, prompting some creators to have a little fun at our expense!”
However, with Google having made improvements to its transcription technology over the years, the captioning service has improved its accuracy by 50 percent, YouTube says. The company specifically cited its work to improve its speech-recognition and machine-learning algorithms as well as expanding its training data.
To attain the scalability required to get to the 1 billion mark, YouTube had to combine its captioning system with Google’s automatic speech recognition (ASR) technology.
For YouTube, 1 billion videos are just the beginning, with the company foreseeing a future when everything is captioned, allowing all hearing-impaired people to appreciate video content.
Tyler Schulze is vice president, strategy & development at Veritone. He serves as general manager for developer partnerships, cognitive engine ecosystem, and media ingestion for the Veritone Platform. Learn more about our platform and join the Veritone developer ecosystem today.