[mythtv-users] Caption Operators, was: 'Censoring' decoding 'bad' words off of Closed Captioning

Jeff Simpson jeffsimpson at alum.wpi.edu
Fri Apr 7 21:00:41 UTC 2006


> Just in case you're interested:
>
> "Live" captions are not hand-typed in the way you are thinking. They
> are inputted to a computer using the same type of keyboard that's
> used by court stenographers, using a phonetic alphabet.
>
> Software then translates this into written language, using context
> and some very complex proprietary algorithms. Since the input is
> phonetic "to", "too", "two" and "2" are the same, it can be difficult
> to tell the difference.

I knew they used a special keyboard, but I didn't realize that the
context is done via software. That's pretty neat.

> I hear folks gripe about the inaccuracy of captions, but it is an
> incredibly difficult job. Most operators could make a lot more money
> working in a courtroom with nights and weekends off. Let's give them
> the credit they deserve.

Oh, no, I definitely didn't mean to berate them on the errors, I just
mean that errors are unavoidable, so you really can't rely on the text
for anything that requires it to be perfect (like parsing for
particular words or anything). I'm not sure that the errors I note are
from the captioners or from the transmission - I've noticed even on
plain-old-tv that sometimes noise gets in in a way where it's not
likely to be an actual typo.

> One qualification test I'm aware of for recorders consisted of 20
> minutes of four people speaking at an average rate of 180 words-per-
> minute. Not only did you have to get what was said correct, but who
> said it. The permitted error rate was 2%. How many of us could do that ?

I can barely keep track of what one person says when they are talking
to me unless I'm interested in the subject :-P

 - Jeff


More information about the mythtv-users mailing list