I've always wondered, how are Closed Captions created? I notice on some shows that are obviously taped, the captions are right on with the speech, but other times it is slightly early or late. For live events, the captions are quite behind the speech, which can be expected as it's live.
I look at all the technology available to translate speech into word documents, and other speech to text technologies and would be interested in seeing if it is utilized at all in Captions or even been attempted. Maybe it's the Engineer in me that likes to know how things work
Captioners for taped programs are usually working from a script and they can include time codes, so it's easy to match them up with the speech. For live events, they can only be displayed as fast as the captioner can go. I think average captioning is 225wpm, using specialized equipment, the same thing court reporters use. Human speech is usually 150wpm but some fast talkers go up to 300wpm. So a captioner's ability to keep up (and maintain accuracy) will depend on how fast the person is speaking.
Speech to text is not suitable for captioning. If you've ever used a speech to text program, you might notice that it makes many mistakes. It's easier to have a person get it right the first time, than have that person correct the computer's mistakes. Also, speech to text is commonly dependent on being trained to one person's voice, so it wouldn't work for TV. (CapTel is able to do it because the captioner repeats everything the hearing person says, it's trained to the captioner's voice.)
