PAT wrote:
> [TELECOM Digest Editor's Note: Regards closed caption, since I
> sometimes these days do not hear as well as I would like, I
> frequently leave closed caption turned on (it is an on/off
> option on my television set) even though I am also using sound
> as well (closed caption allows me to keep up with words I miss
> or do not understand occassionally.) But has anyone else
> noticed how they really _blow it bad_ sometimes, with trash
> symbols instead of the words, etc,...
The "trash symbols" you see are the result of corrupted CC data. In an
NTSC video signal, the CC data is encoded on Line 21 of the picture
raster (if you slightly underscan your TV set, you can see it dancing
along at the top of the screen as a horizontal line of black and white
dots and dashes). It other words, the CC data is part of the video
signal, not a separate data stream. Consequently, it's subjected to
all of the various degradations that also affect the video signal:
noise, distortion, non-linear heterodyning, scrambling/descrambling,
interference from outside sources. Furthermore, if the video signal
is sent digitally (as on digital cable or DBS), Line 21 is sampled
right along with the rest of the video. Even if the sampling data
stream is error-corrected, the underlying CC data is not.
> ... or sometimes just approximations of the phrases used
> instead of the actual words?
Sometimes this is done intentionally; sometimes not. An ad libbing
speaker often speaks in a hesitant manner, sometimes repeating himself
or inserting extraneous or meaningless words. An effective
closed-caption track captures the essential meaning of the speaker's
words, but it isn't necessarily a literal word-for-word
transcription. Indeed, a word-for-word transcription can be difficult
to follow, and sometimes downright annoying (it certainly isn't
necessary to transcribe every "...uh...uh...").
> And in the case of VCR or DVD movies, I assume the closed
> caption is just encoded right on the finished product, is that
> correct?
Yes. It's also true of prerecorded network programming; in fact, it's
true of just about everything you see on television except live or
live-delayed programming. I once observed a captioner at work at the
National Captioning Institute; at the time of my visit, he was
captioning a sitcom. He already had the original script as a text
file; his job was to copy the script to the CC track, match it to the
sound track, and modify it if necessary (as he explained, actors who
really know their roles tend to adlib their lines, especially in
sitcoms). I was amazed at how fast he worked: fingers flying across a
standard PC keyboard, one eye on the video monitor, one eye on the
text monitor, half-listening to the sound track. I asked, how he could
do it so fast? He just shrugged and said "all sitcoms are alike" --
once you know the basic rhythm, you just know intuitively where to
insert the CC data.
> And if it is a 'live program' such as a newscast instead of
> some pre-recorded stuff, it appears they also create the
> closed caption live, since it drags behind the audio by a few
> seconds. PAT]
During live events, CC data is created on-the-fly using the same
22-key "stenotypes" that court reporters use. Some community colleges
now teach stenotyping, emphasizing the training for, and the future
job prospects of, both court reporting and closed captioning.
Of course, doing this type of work accurately requires more than just
stenotyping: it also requires knowledge of the subject at hand. Court
reporters obviously have to understand legal terminology;
closed-captioners have to understand the terminology of whatever
they're transcribing.
Inevitably closed-captioners make mistakes. Among my favorites:
Senator Daniel Anyway of Hawaii
Felix Mendel's son
a feat snobs
John Maynard Cain
Nevertheless, considering the vast range of topics that
closed-captioners have to deal with, I think they do an amazing job.
Neal McLain
[TELECOM Digest Editor's Note: Your comment about transcribing an
extemporaneous speaker was a good one. For a few years in my former
life in Chicago, IL I had a job of tape-recording then transcribing
a lecture series given at the Chicago Public Library. It was stressed
to me, 'do not transcribe him literally, transcribe what he _meant_.'
So I would listen to the tape three or four times, jotting down some
notes. Then I would listen to the tape again, and type it up mostly
'word for word' (these notes were given out to persons who requested
them later on), but I would *make complete sentences* where the
speaker had made partial sentences; I would eliminate the extraneous
'ah, um' things. But then before I took the finished transcription to
be mimeographed in a large quantity, I took my finished product to
the speaker and said "if this correctly states your lecture, _please
sign off on it for me_" then when the speaker and myself were in
agreement (we almost always were), then I would get the copies run
off. PAT]
Neal McLain (nmclain@annsgarden.com)