Alternative Realtime Careers: Excerpt

This is an excerpt from Chapter 2 of Alternative Realtime Careers, which should give you a feeling for the content and writing style of the book. For more information, take a look at the online table of contents.

Where captions come from

an excerpt from Alternative Realtime Careers

One of the most common mistakes in the captioning business is confusing online captioning with realtime captioning. In fact, realtime captioning is but one of the facets of online captioning, and the captioner who truly masters online captioning must understand and master all of these facets.

Realtime

Realtime captioning is the contemporaneous creation of text from speech. Up until very recently, stenocaptioning has been the only way to accomplish this. Computerized speech recognition has now been used for realtime transcription with acceptable results, although as of this writing there is only one person in the country that has passed the certification for realtime speech recognition. Speech recognition has come a long way in specialized, controlled environments, but it still requires a system dedicated to a specific person.

The individual has to be trained for the system, and the system has to be trained for the individual as well. Speech recognition systems simply can't be turned loose on the audio of a television broadcast or the public address system in a lecture hall and expected to have acceptable results.

Why is this? Because humans know when they aren't able to achieve accurate realtime transcription, and they understand the subject matter. When a speech-recognition system gets lost, all it can do is spew gibberish that approximates the phonetic translation of the input. A stenocaptioner is capable of switching, on-the-fly, from verbatim to paraphrasing, and still capturing the gist of the subject matter.

Even in the unlikely event that the next decade or two produces a speech-recognition system capable of producing 99% or better translation with an unknown speaker and unknown subject matter, the system will not be able to produce a paraphrased transcript of an out-of-control situation. A bold prediction? You bet it is. It is rarely safe in the world of computer technology to declare that something won't be accomplished, especially ten or twenty years in the future. Given, however, that it is almost twenty years since I began work as an operating systems programmer, I feel comfortable with this particular bold prediction.

Scripts

Live display captioning, also know as scripted captioning, is a live "performance" using a prepared script. The most common example would be stories pulled from a newsroom computer prior to a newscast. The stenocaptioner can load these stories into the captioning system and generate the caption output by feeding these stories out a line at a time. Obviously, if the on-air talent (the anchors) follow the script exactly, then this method produces perfect captions. Most of the time, however, that doesn't happen.

A typical newscast will involve anywhere from 35% to 60% TelePrompTer captioning, with the remainder being unscripted. Unscripted segments in a typical newscast include all live remotes, interviews, banter between the talent, and often the weather and sports. The remainder of the newscast must be done using realtime captions. Broadcasts such as sporting events have little or no available scripted material.

As we will discuss later on in the chapter about captioning-related laws, there are now restrictions on when TV stations are allowed to use standalone newsroom computer captioning.

Early stenocaptioners had to reach from their steno keyboard up to the computer keyboard every time they wished to feed a line of script. In some cases, the scripts were actually loaded on a different computer, making the process even more complex (or requiring two people for each broadcast). Nowadays, stenocaptioners can define a stroke on their steno machine to feed scripts, or in some cases even use a foot pedal.

Stock items

There are certain captions which appear so often that they end up programmed as a key on the computer keyboard, such as [applause] or [laughter], or the credits that many captioners display at some point during their broadcast. Modern online captioning systems also allow stock captions to be triggered from the stenotype keyboard so that a stenocaptioner can bring them up without touching the computer.

How you actually go about accomplishing this will vary from system to system. Some don't support stock items at all. Others allow stock items, but they can only be the same style and position as the realtime being sent.

Stock captions can be used like court reporters use auto-includes. Any text that is repeated exactly from one job to the next, such as credits, messages to viewers (one TV station puts up a TTY number every broadcast for feedback from deaf viewers), and sign on/sign off messages.

Electronic Notes (C-Print)

This capability goes by several names, including electronic notes, C-Print, keyboard realtime, realtime minutes, and others, all of which. basically consist of typing captions in realtime.

Why would anybody use electronic notes instead of realtime? There are a number of good reasons.

Perhaps the most obvious is the emergency broadcast. When disaster strikes, many television stations have contracts wi th captioning firms to provide emergency captioning on a standby basis. Typically, there is a time specified in the contract. For example, the captioning firm may guarantee to have someone on the air within, say, 10 to 30 minutes of the emergency.

So what happens in the meantime? Are the deaf and hard-of-hearing viewers left hanging for up to half an hour? Without electronic notes, the answer is "yes." With realtime notes, the TV station can grab any available person, sit him at a computer, and have him type a synopsis of what is going on. It may not be fast; it may not be accurate; it may not be complete; but any information is better than nothing while waiting for the stenocaptioner to come online.

Autofeed

Online captioning systems also have the ability to output a prepared script at a specified reading rate, perhaps in a repeating loop. This can be another solution to the dilemma raised above, during the time between the onset of an emergency and the beginning of realtime captioning. Let's take a look, for example, at what could happen during a flood.

In the newsroom, the assigned individual calls the standby captioner and tells him to get on the air as fast as possible. At the same time, someone is assigned to sit down at the captioning system and type a message. Perhaps something along the lines of, "There has been a flood affecting ABC County. We expect to have realtime captions on the air shortly." As soon as the call to the captioning firm is complete, the assigned individual prepares a short script file containing a summary of the emergency and the required action. The action may consist of, say, a list of areas being evacuated.

That script file is then loaded in the captioning system and assigned a reading speed (160 words per minute is typical). The script is then set to repeat. That message will continue to loop across the screens of the deaf and hard-of-hearing viewers until the realtime captioner gets on the air with actual verbatim coverage of the news.