Can realtime captioning be done using voice recognition systems?
Speech recognition systems are becoming increasingly accurate and fast. As of today, however, there is no hint of a technology that would actually be able to capture text from multiple speakers in noisy environments with acceptable accuracy. Additionally, much of speech recognition is based on interpreting words using their context. This requires storing up entire sentences, which would leave the captions lagging far behind the speakers. For more details, see the Speech Recognition Technology article that provides a realtime reporting/captioning perspective on the subject. There is research going on now into the use of speech recognition technology for captioning, but you shouldn´t hold your breath. Update: There is now a company in Canada with a speech recognition captioning system. The speech recognition is used only for editing or some data entry, however, and the system cannot be used in a realtime environment. Can captioners replace sign interpreters at live events? Certainly not.
