Why are missing letters in closed captioning missing in pairs?
November 14, 2011 8:13 PM   Subscribe

Why do missing letters in closed captioning always appear in pairs?

Sometimes I watch TV with the closed captions on. Sometimes there are missing letters.

It seems like, when they're missing, they're missing two at a time. For example, I might get "metafilter: counity weblog" (with two consecutive letters "mm" missing), or "mefilter: community blog" but not "metafilter: commuity blog", with just one letter missing.)


(I'm not asking how to fix this -- I don't care enough to do anything about it. I'm just curious.)
posted by madcaptenor to Technology (11 answers total) 3 users marked this as a favorite
Closed Captioning is live (at least the first time the program airs) and the person types in everything live.

The machine is not like a typewriter, but more like a stenographer machine to speed up the captioning.
posted by Yellow at 8:27 PM on November 14, 2011

A stenotype is laid out very differently from a normal keyboard and I think that there are keys that include multiple letters that are commonly used in combination (e.g., there might be a "mm" key, and maybe a "ta" key). I think most closed captioning is automated these days, but live TV might still be done by an real-live stenographer, so that might explain this.
posted by asnider at 8:29 PM on November 14, 2011 [2 favorites]

Closed Captioning is live (at least the first time the program airs) and the person types in everything live.

Definitely not true. It may be true for news, sports, academy awards, etc...but its not true for all tv.
posted by hal_c_on at 8:53 PM on November 14, 2011

Closed captions, whether live or preproduced, are encoded as sixteen blinking white bars in the video signal outside the normal viewing area. The pattern of bars corresponds to four 8-bit bytes, which is enough to encode two letters. If the signal drops for a frame for whatever reason, the caption decoder will miss four bytes and thus two letters.

More info here.
posted by infinitewindow at 9:00 PM on November 14, 2011 [27 favorites]

It's covered here:

"The captions for prerecorded programming are prepared before the broadcast. The typists can take their time using a regular computer keyboard. They may pause, restart or even rewind the program they are transcribing to improve accuracy.

Live broadcasts, such as the nightly news, are a different story. Live closed captions are done in real time by a trained stenographer using a court reporter's machine modified for closed captioning. The machine permits the stenographer to transcribe the spoken word a syllable at a time. Accuracy can suffer depending on operator experience, the voice quality and enunciation of the speaker being transcribed, and the overall audio quality. A stenographer who starts to lag behind may skip words or even whole phrases in order to catch up. Miskeys often cause "garbage" characters to appear in the captions, and misunderstandings of what was said can lead to odd-looking spellings or turns of phrase, often with humorous results."

posted by peagood at 9:01 PM on November 14, 2011

Just to clarify the question: I was asking specifically why characters would be dropped in pairs. Hence infinitewindow's answer is the one I wanted.

That also agrees with something I noticed after asking the question: I could have, say, the fifth and sixth characters of a caption being dropped, or the seventh and eighth, or generally any odd-numbered character followed by the next even-numbered character -- but not the other way around.
posted by madcaptenor at 9:11 PM on November 14, 2011

When I was a traffic reporter, I would have the local news shows on my four TV monitors, on mite with captions on. I collected my favorite bloopers (and am thinking of writing a book with them.)

One of my favorites should have said "NBC's Lisa Meyers will have the story on NBC Nightly News."
The two-letter dropouts got her name badly, and the caption came out as "NBC'S LIERS" etc.
A lot of the captioners monitor the audio by phone (the captioner for one of the shows I was on in Dallas was from Colorado) and phone static was bad some days.
posted by Gridlock Joe at 9:29 PM on November 14, 2011 [1 favorite]

Great name for a traffic reporter, Traffic Joe!

Also, when people are taught transcription, I believe they are taught to ignore the context as much as possible and just get the sounds right. Once they are able to transcribe all the sounds in real time, then they add context back in so they can get homophones and word bunching ("dollars" versus "doll ears" and "Lisa Meyers" versus "Lee Sameiers") mostly correct. Depending on how experienced the transcriptionist is, you will see mistakes like this.

Adding a little to infinitewindow's comment, the NTSC video signal is very simple. Think of it like how a dot matrix printer works. Line of text, carriage return, line of text, carriage return. Each "page" of text is 486 lines long, but the signal is 525 lines long. This was done so that the TV sets had a chance to move the mechanism back up to the top. Then people realized that you could put other data into that empty part of the signal. Sometimes you can see the blinking bars if you are watching standard definition on a digital TV or a computer.
posted by gjc at 5:43 AM on November 15, 2011

Closed Captioning is live (at least the first time the program airs) and the person types in everything live.

I don't know how it is in the US, but stenographers are rare in the UK now - there are fewer places that train them. When I worked in subtitling, the majority of live programming was done via a process called 'respeaking', where the subtitler would read short form/macro commands based on the audible text into a machine which would then translate them into subtitles on the broadcast. (My job was to produce word lists of the tricky words that wouldn't necessarily be in their system, so names, places, unusual slang that might come up.) This was why there were a lot of homophones appearing in subtitles. It's much faster and easier to train someone to do than stenography!
posted by mippy at 9:25 AM on November 15, 2011

And, yes, non-live programming (or at least the stuff that was made far enough in advance - certain topical programmes were classed as 'live' for our purposes) was transcribed by a different department who would use the script and the video file together and work at adding subtitles that conveyed appropriate meaning where the sentences were too long, or where there were certain sound effects that needed to be captured. This was essentially a combination of audiotyping and interpretation.
posted by mippy at 9:27 AM on November 15, 2011

posted by mippy at 9:29 AM on November 15, 2011

« Older From software to a trade in as little time as...   |   Beyond American Lit Newer »
This thread is closed to new comments.