Speech Recognition for Live Closed Captioning

摘要

"Closed-captioned broadcasts", which provide audio information in TV programs through its transcribed text, have become important medium to ensure information accessibility for the hearing impaired and elderly. It had long been difficult technologically to provide closed-captions in a live broadcast program in Japanese. However, steady advances in speech recognition technology and Japanese stenographic keyboards have recently made it possible to provide live closed-captions for programs such as news and sports. STRL researchers are continuing with their research and development of speech recognition systems to realize closed-captioning more efficiently.