With a reporter’s rough-draft (which is usually audio synch’d as well), you just search for the keywords, and you instantly have the question in front of you — no rewinding, no guessing where it might be — and you’re not costing any lost time in a room full of people who are individually charging hundreds of dollars an hour.

They face a lot of the same obstacles, with being aware of homonyms and “sound-a-likes” in their writing (or speaking), maintaining their dictionaries software entries to deal with new words, word-boundary issues, and other problems that cause errors in the first-draft translation. 

As a stenographer (stenotypist), I have a lot of respect for highly skilled steno-mask writers, as they’re doing essentially the same thing, via a slightly different method. 

The problem with speech recognition is generally that people have different accents, they mumble, and the courtroom or even deposition setting is generally not a “sterile” environment with respect to audio.  There are papers shuffling, people coughing, sneezing, mumbling, talking simultaneously.  In a courthouse, doors are opening, people are shuffling in and out, sometimes in chains; there are side conversations. 

For all of those reasons, it’s still important to have somebody there actually recording the proceedings.  If you search on Google for audio failure in courtrooms, you’ll see the pitfalls of relying on un-manned audio systems.  It can work in a controlled environment and sometimes in general populations, but there is the risk of equipment failure going unnoticed because nobody is sitting there producing a rough draft, monitoring their equipment as they’re doing it. 

With audio systems, you also have to begin transcription with the recording and a blank page, whereas if you’ve had a steno writer in the room, you begin transcription with a rough draft of the translated shorthand, which is often sold to litigants at the end of the day so they can immediately begin bolstering their own notes with quotes from the day’s testimony.

Just a few thoughts about why this ad is not “anachronistic.”

A typo of one wrong key on a regular keyboard will result in a slight misspelling, whereas a wrong key on a machine shorthand keyboard (a stenograph) can result in an entire wrong phrase, or a phonetically similar but entirely different work coming up.  Every piece of shorthand is associated with an English counterpart in the stenographers dictionary, hence the spectacular typos you see when things go wrong.

Still, I’d challenge anyone to sit down and try to even paraphrase a News broadcast and see what their rough-draft looks like before editing.  You’ll have a new appreciation for just how quick and accurate those TV captions actually are, even if they do contain the occasional spectacular mistranslation of the stenographer’s shorthand.

I’d been going to say that it’s obsolete, but a lot of the comments say that the fundamental transcription skills are still current today for some businesses (though the company I worked for from the late 70s – early 90s had a typing pool and accepted dictation by phone when I started, and had gotten rid of it by the mid-80s when it was expected that even managers could do most of their typing using computers, and that’s pretty typical for most businesses.) 

And back in the 60s, the people who really needed a good word-at-a-time typing system were the Chinese – pre-computer Chinese typewriters required the equivalent of a college education’s worth of training to be good at, depending on how much of the language you wanted to be able to cover. 

My point is that the Latin alphabet (and indeed, most alphabets) aren’t really great for taking notes. Shorthand, on the other hand, is. Now, if we could learn shorthand, and have an IME for our phones/tablets/whatever, taking down better notes would be easier.

This is to do with handwriting, but the same argument applies for key input. For example, a chorded keyboard like Stenotype should be much faster with just a few fingers (maybe even just thumbs). Check out GKOS.

The problem would be the learning curve. Fortunately or unfortunately, we’re addicted to QWERTY

(Though one primary characteristic of AI is that any time AI gets good enough to do things, like reading text or theorem proving, everybody says “Oh, that’s not really AI, that’s just OCR or automated logic, real AI is still decades away if it’s even possible!”)

 My wife used to be a court reporter for the U. S. Air Force. She would listen to the discussion and then repeat it into a stenomask. She would then transcribe the recording after the fact. Your own voice is less ambiguous apparently.

Toward the end of her career, she was involved with testing the speech to text machines that they use more often now. For accuracy the speech to text device is trained to understand only the court reporters voice. She would then clean-up the transcription later.

In both cases it’s just a note-taking aid for a court reporter who is physically present and can certify the text matches what was said.

Do a search for “Stenomask” to see some of the devices.

Rome may have been an unusual exception though, since many legionnaires knew how to write letters back home, and the walls of Pompeii show that the ordinary citizen was no stranger to graffiti. However, even with all that, the estimate is 10-20% literacy.

So… magic!

Incidentally, most stenos have gone through a multi-year schooling process to get where they are.  It’s not the sort of thing most people could just pick up in their spare time…

I did this pretty much every time we had a jury trial.

The smaller the community the more interesting it is. Currently, I know what three people on my block were arrested for. I’ve considered moving, but then I’d be surrounded by people who I don’t know what they’ve been arrested for.

The devil you know, I guess.

http://gantopian.com/wordpress/wp-content/uploads/2011/10/Comentarii-Noten.jpg

…Okay, pretty easily, maybe.  But 300 WPM!