I see this all the time: “That’s” misspelled as “Thatus” and “Don’t” as “Donut” and similar mistakes. I see these on the closed-captioning for advertisements, so it’s not a simple transcription error.
My understanding is the these systems work phonetically, but it seems absurd to not fix some of these errors (how many times does the word “don’t” get said compared to donut?).
Bump.
No TV geeks on the SDMB?
The only time I’ve ever seen errors like that is during the news or other live broadcasts, where it has to be either done by a fast typist or by an automated system. Commercials and other pre-recorded broadcasts don’t seem to have any more typos than any other printed matter.
I’ve wondered this too. I even see awful closed captioning on old programs on MeTV. Some 50-year-old show will be on, and the CC will be all over the place. Sometimes I even see the word appear, then disappear as though the typist is backspacing over it. I can’t figure it out. Surely the CC isn’t being done live, is it? Isn’t there a decent CC track by now for these old shows?
As previously mentioned those mistakes occur when a court reporter is typing the captions live or software is being used to “hear” the program and type the captions.
The fcc has recently http://www.fcc.gov/document/fcc-moves-upgrade-tv-closed-captioning-qualitymade effort to improve captions since those errors are pretty common.
Edit: old tvs use to have special telecom adapters to decode for captions. While newer boxes after 1980 are made with these pieces really old showd have terrible cc. I assume its because of this but who knows
Some captions arehilarious. Or at least mildly amusing.
I can understand CC for news broadcasts or live events being sort of chancy. For pre-recorded stuff like series TV, they probably have the post-production time to do it right. For televised movies, well, it’s a tossup, I guess.
Two different CC systems, “digital” and “old”. The old system put the captions into the analog signal, in the spaces between the pictures. No reason why the content of the Captions should be different: I suspect that if old shows have terrible CC, it’s because the captions were done less carefully.
Intereseted to see the comment above about CC in commercials. We don’t get CC in commercials at all. Sort of irritating when they are playing trailers with no captions in a 12 minute break between programs, since if they were captioned, the trailers are probably better than the programming content.
Is there any chance the code for apostrophe is similar to the code for U, and it’s getting scrambled up?
Rather than being close to the ‘U’, I would think it might be improper special character handling.
Another error I’ve seen sometimes is when the news has anchors talking about current events while showing archival footage: Sometimes the captions will be for the original sound of the footage, not the current commentary.
I’m specifically asking about these “phonetic” mistakes. I watch the TVs at the gym when I am on the stationary bike, and it is the rare day that I don’t see one of those errors. Especially “Donut” for “Don’t.”
It just seems odd to me that whoever wrote the software hasn’t flipped the default lookup so that the text is rendered as “Don’t.”
There are quite a number of these that I see on a regular basis - the next time I catch one, I’ll post it.
I don’t completely understand it, but I can shed a little light.
Transcription for captioning is done by real time transcribers, like those used in courtrooms. In court reporting school, each transcriber evolves her own personal “language” so that a combination of simultaneous keystrokes results in a phonetic “sound,” which is immediately translated to the screen. It’s instantaneous, and the transcriber has no opportunity to clean up what is published. If the speaker is going slow enough, the transcriber may be able to insert a correction before the “sound” is published in real time, but often not. It really depends on how fast people are talking. No two transcribers take down what they hear in exactly the same way, and sometimes they even have trouble reading another transcriber’s notes. The “donut” sound may serve as a transcriber’s phonetic sound for a number of different words. Another transcriber’s combination of keystrokes might yield a sound more like “dontia.”
Court reporters or transcribers are trained to take down up to three speakers simultaneously, but the transcript in such circumstances won’t be very “clean.” Before a transcript is published in printed form, they spend hours cleaning up the mistakes made based on the phonetic nature of their personal language.
I hope an official court reporter or transcriber will be along to clarify the exact process, but this is an overview based on many years spent working with them in courtrooms.
Over here, you see a lot of hilarious mistakes, but that’s usually due to non-native speakers doing the captioning. Why you’d have that in the US with native-speaking Americans doing the captioning, I’d guess just pure sloppiness.
If only it were so, then these spelling errors would be the only problem. And for live broadcasts, I suppose it is indeed the major problem. But from what I see, the captions for prerecorded shows and movies are taken directly from the script. If an actor goes slightly off script – use a synonym, changes the word order, adds a word or two – you don’t see it in the captions. The spelling is perfect, but it’s not what was spoken. This happens to 40-year-old reruns and to first-run dramas. If they wanted to, one individual could fix all of these in an hour of post-production. But for some reason, it doesn’t happen.
Kind of sounds scary about the accuracy and integrity of courtroom reporting in an actual, y’know, courtroom. Is there a possibility that the final transcripts donut all come out right?
No, not really. Their accuracy when they have time to clean up their transcripts is excellent. They remember the context of what was being said and generally know what their strokes were meant to say. It’s challenging, difficult and very physically demanding, but they usually get it right.
If there is a substantive disagreement as to what was said, a motion is made by the party who disputes the language in the official transcript, the parties present their cases and get a ruling from the court. In the fourteen years I worked with official court reporters, I only saw this raised once, and it was in a death penalty case.
Recent thread about this:
http://boards.straightdope.com/sdmb/showthread.php?t=743913&highlight=subtitles
The time, yes. The will to bother with high quality, no. A computerized TTS system is cheaper than two humans (one to create the CCs, one to review them), but they need to be “trained” for a specific voice in order to work any decently and they have the same problem as autocorrect systems in that they pull from a limited dictionary; the word they pull as “closest” may be phonetically close to what’s being written but it won’t necessarily make sense. Names are one of the things they tend to be shit at, since names are not going to be in an off-the-shelf dictionary.
And, as Keeve said, often the subtitles are pulled from the script. If you have access to media with both a “normal” CC and a “Hard of hearing” CC, check both at any point where the “normal” version is mistaken. The Hoh version will normally be taken from what is actually uttered.
Not exactly related to the OP, but why do a lot of CC’s censor curse words? The show being captioned says the “bad” word but the captions change that word to a less offensive version. Are hard of hearing and deaf people more sensitive than the rest of the population? Or, are the folks doing the captioning delicate flowers that can’t handle typing such obscenities? 