Layout of a stenographic keyboard
By Mark Johnson, Realtime Coordinator
There are a number of high-profile screw ups in live closed captioning that have been ridiculed and lampooned on the internet. Setting aside the high-pressure nature of a realtime closed captioning writer’s job–they perform in a live, highly public environment where it’s impossible to go back and fix mistakes–it’s important to understand how their equipment and software works so we can better comprehend why these mistakes can happen.
A stenographer uses a stenographic keyboard. This keyboard is comprised of 22 keys, and words are typed by pressing multiple keys at once to create all of the sounds in a word. For example, to type the word “how,” they would press “H,” “O,” and “U” simultaneously. The software then translates those keys into a word by looking into a dictionary the writer has prepared. This dictionary is essential because a writer cannot type a word that does not exist there. This is why you rarely see spelling errors in closed captioning–unless it’s a proper noun–and why you will more frequently see issues with homonyms.
An example of this dictionary in action can be seen from the Week 16 NFL game between the Carolina Panthers and Atlanta Falcons. The Panthers were turning to a rarely-activated running back, Cameron Artis-Payne. Initially, every time the announcers said his name, the closed captioning read “CAMERON ARTIST PAIN.” The writer for this game did not have that player’s name in their dictionary. Subsequently, when they entered the keystrokes for “Artis-Payne,” their dictionary searched for the closest sounds it could find, “artist” and “pain.” At a commercial break, the writer entered the player’s name, and for the rest of the game it was correctly written as “Artis-Payne.”
It’s unlikely the writer forced their dictionary to spell “Artis” instead of “artist” or “Payne” instead of “pain,” especially in a game as violent as football where “pain” is likely to be said. What the writer probably did was enter a shortcut, known as a brief, for “Artis-Payne.”
This use of briefs for a proper noun is the likely explanation for a much higher-profile closed captioning mistake, when the Boston Marathon bomber’s name was written as Zooey Deschanel. Dzhokhar Tsarnaev is a very unusual name, and the caption writer certainly had to create a dictionary entry for it. The keystroke of “DZ” would make sense, as that’s not a combination of letters that’s likely to be used for something else. But if the writer also does entertainment news and had a dictionary entry for Zooey Deschanel and used, say, “ZD” for her name, there we have our problem. The writer either mistakenly used the wrong dictionary–an entertainment dictionary instead of a news dictionary–or they accidentally pressed the wrong keystroke, using “ZD” instead of “DZ.”
Realtime caption that was supposed to read "Dzhokhar Tsarnaev" reads as "Zooey Deschanel."
Knowing it’s impossible for a caption writer to write a word or name that isn’t in their dictionary, when we see issues like what occurred during the August Republican debate, there’s something else at play. During that debate, the closed captions read as complete gibberish, “HALFRICAN SENLD IF YOUK.TYHOUTOCIAL.” Even if the writer was hitting random keys on their stenographic keyboard–or if a cat walked across it–their software would find the closest matching word in their dictionaries and spit those out, meaning it would still be nonsense, but that nonsense would contain correctly-spelled words. In cases where complete gibberish is what comes out on screen, there is a hardware issue either with the encoder that embeds the captioning, the modem that transmits the data, or the phone lines connecting the writer to the network. Garbling does not occur during the transmission of captions when the connection is IP-based instead of modem-based; if there are problems with the connection from the writer to the network in this scenario, their connection will give out entirely instead of producing garbling. If garbling occurs on an IP connection, it points to the problem stemming from the encoder.
As Senator Lindsey Graham debates on Fox News, the realtime captions read as gibberish.
CaptionMax is 100% iCap compliant, and we offer the fastest and highest security realtime service in the industry with our Rocket™ Realtime Technology. For more information on our realtime captioning services, contact email@example.com.