While participating in ACT VII at the Department of Commerce, we realized that many people do not understand the methods by which live closed captioning is created, nor do people understand some of the mistakes seen during a live realtime or captioning session. To that end we offer the following:

Realtime closed captioning is a system based on the talents of the traditional court reporter. As words are spoken, a machine shorthand reporter, or stenotypist, commits the sounds to stenotypy, a phonetics-based system that allows record-making at speeds beyond the best of typists. Traditionally this stenotypy was then read by the reporter and typed, or dictated into a tape recorder for later transcription by a typist.

With the advent of CAT (computer-aided transcription), computers were taught to translate this stenotypy code into English, which a scoper (computer operator) would then edit, teaching the computer to recognize new outlines (stenotypy) so the next time that outline or stroke was used the computer could properly translate it. This is great when you have two weeks to prepare a transcript.

Next, systems were designed to translate these steno outlines on the fly; realtime translation was born. Some systems allowed for editing and printing of this material during translation, and again, for producing transcript, errors or untranslates (undefined steno outlines) could be handled by the scoper before final printing of the transcript.

Enter realtime telecaptioning. Realtime captioning certainly follows this same procedure, except there is little to no time for editing of mistakes. Untranslates can be defined during realtime captioning, but usually not until after errors and untranslates have been viewed by the audience. This means that a new outline will be seen at least once before the scoper can teach the computer to read them, and this, plus what are called mistranslates, is the topic of the present discussion.

How do you recognize an untranslate? Most realtime captioning is presented to the viewer in all upper case characters for easier viewing, otherwise untranslates would be easy to spot, so here are a few pointers:

UNTRANSLATES: If you see a word that is discernable but incorrectly spelled, you have probably just spotted an untranslate. Virtually all realtime captioning software is designed to display a phonetic translation of any steno outlines not already programmed into the computer, otherwise untranslates would appear as a seemingly nonsensical group of characters, e.g., TPOEPB/ET/EUBGS, which if not already defined would phonetically translate something like this: FONE ET IKS.
With this phonetic translation you can probably see that the word that was written was in fact "phonetics." As you can see, the original outline would make no sense at all, while the phonetic translation allows the viewer to understand what was meant, even if not spelled correctly. It doesn't always work out so well, but continued software development will improve this situation in coming years.

MISTRANSLATES: Since the captioner is pressing as many as ten keys simultaneously, miskeying is going to occur. Sometimes one wrong key pressed in a multi-key stroke of the keyboard can cause a mistranslation that in English is not even close to the desired word. These are much more difficult to understand, sometimes impossible if not trained to read stenotypy. For example, PHOER could be defined as the word "mother," while PHOR could be defined as the word "more," and PHOERT as the word "motor." Remember that the P-H-O-E-R-T characters are not typed individually, but all at once, and you can see that mistakes can easily be made, especially when someone is speaking 220 to 280 words per minute. So if you see a caption that says "the new Honda has a more fuel efficient mother," don't be shocked, realize that this was a forgivable error by someone that's working furiously, fingers flying, trying to keep up with the speaker.

BRIEFS: To improve speed and accuracy most captioners use what are called briefs. These are steno outlines that would not make sense even phonetically, and if one is misstroked the phonetic translation will probably not help. As an example, Allison Ash-Hoyman, an accomplished realtime writer since 1985, writes the steno outline PWA*EU for the phrase "by the way." This saves time while captioning, but misstroking this outline by one character could cause a phonetic translation like BAI, WAI, PAI, or possible mistranlates such as bay, pay, weigh, etc. Another example would be the steno outline TKOUR, which would phonetically translate as DOUR or DOWR, but the actual definition of TKOUR is "do you remember." So sometimes it's better to just let the caption pass and try to develop a meaning elsewhere in the program.

MARRIAGE: Sometimes a steno outline that means one thing when it stands alone has many different meanings when it is attached, or married, to other steno outlines, such as the steno outline PRO. Many stenotypists use the outline PRO for the word "proceed," but when married to the outlines PORGS and TPHAT, they form the word "proportionate." But if the three outlines have not been defined together on the computer, what you might see could look like this: "proceed portion gnat," or "proceed portion NAT," and there are other possibilities.

Faulting the stenocaptionist for these mistakes will not change things, it is virtually impossible for a stenotypist to be perfect, just as typists, with no pressure to perform at unbelievable speeds, still make errors. If you rely heavily on live captioned material to receive your local news (although it may have been "cleaned up" from earlier broadcasts) or other information, it is best to learn to live with it by training yourself to understand some of the shortcomings of an imprecise system.

Books on stenotype theory are available and, like ASL, can be intriguing and even enjoyable.

This is just a primer and not meant as a concise discussion of this topic. If you have specific questions or areas of interest concerning live realtime captioning or realtime translation you can write to:
Dave Hoyman
RealTime Reporters
1100 Fetherston Way
Felton, CA 95018,
or
303 Potrero St.
Suite 42-203
Santa Cruz, California
or by
E-Mail: dhoyman@rtreporters.com

Home -- Public Events -- Internet Captioning -- Television Captioning
Experience -- History -- Fun Links and Search Engines