Voice Recognition for Teachers

Voice Recognition for Teachers

Summary for if you're busy: Voice recognition works brilliantly now, even in noisy environments. By far the best software is Dragon Naturally Speaking. Get the cheapest version "Standard". Click here for educational pricing. You're crazy not to buy a special microphone designed for voice recognition as well. Click here. Watch out! Even though it works fantastically, psychologically you'll find it quite uncomfortable, and if you don't work alone, consider privacy and annoyance.

Read on for a bit more detail:

Here's me dictating at almost 200 words per minute with 100% accuracy. I am using a headset designed for voice recognition, which I think is very important if you're going to bother buying the software.

But that's just a gimmick. In real life I don't each those speeds because I can't think of what I want to say that quickly!

I use voice recognition quite a bit to dictate emails and documents. Here’s some information for other teachers curious about it.

 

Why would I use it?

Not for speed. The claim that you can dictate text with voice recognition quicker than you can type it is not necessarily the case, especially if you can already type at a decent speed. I find that by the time I’ve made corrections, it is about the same speed as I type.

Because you’re sick of typing. This is my main reason. I just get sick of typing. My fingers get tired or sore. Or mentally I get annoyed at the effort and clunkiness involved with tapping each individual key for every last letter in every word.

Because you hate the paperwork of teaching. I use voice recognition mainly for the boring bits, e.g. writing teaching programs, giving feedback to numerous different students about the same task. Emailing various people on a similar topic. There is little creative fun in this sort of typing. It is repetitive and dull. Voice recognition is at its best for this. It lets me fly through it as quickly as I can talk, without getting tired fingers. Take, for instance, essay marking. Students make the same mistakes over and over again. It is soul-destroying writing “start each paragraph with a topic sentence” or “ensure you integrate quotes into your sentences rather than just putting them by themselves” over, and over, and over again. I suppose you could use macros for this. And of course typing these comments is easier than handwriting them. Easier than all these, is voice recognition.

You’re in danger of repetitive strain injury. Or some other muscular problem. Or perhaps it’s too late, you’ve stuffed your fingers but are still in a job that requires heaps of text-production. I enjoy playing computer games in the evenings, and it’s too much for my fingers if I’m also typing.

How well does it work?

Very, very well. The technology is much, much better than it was 5 or 10 years ago. Here's a less gimmicky demonstration than the one above:

Accuracy is not 100%, but it isn’t far off (there are some words it always gets wrong with me). You can speak as quickly as you want to without hurting the accuracy, providing you’re speaking clearly. In fact, speaking in phrases or full sentences is much more accurate than dictate one or two words, because it gives the recognition program a context to help figure out what you said.

Correct by typing. The best way to use it is to have your fingers ready to correct bits here and there as you go. The program would rather that whenever you make a mistake, you retrain the program to avoid the error in future. Fair enough, but you come up against a law of diminishing returns. The program is already very accurate and it’s not worth me stopping to retrain every few sentences to push accuracy from 96% to 98%.

DANGER! DANGER! I often don’t notice errors. DANGER! DANGER! It is very difficult to accurately proof read your own dictated text. So many times I’ve reread my dictation carefully, and missed huge errors that make you look careless when your document is published, email received, or whatever. I was intrigued that a few years ago I noticed quite a few errors in a colleague’s emails, only to discover he was using voice recognition himself.

 

Which program to use?

Without a doubt, Dragon Naturally Speaking, which is miles ahead of any competition. The current version is 9.0 but they've patched it to 9.5 for free. There are three versions: Standard, Preferred, and Professional, but all have the same accuracy and the same speech engine, so you can get the cheapest version, Standard, and get the same performance, with fewer bells and whistles in the software.

The student price for the Standard version is as cheap as $60.

 

The teacher price for the Standard version is $100, and about for the Preferred version $200. That particular website offers to bundle Standard with a proper voice recognition headset (which I imagine is very good, like my recommendation below) for $179 or Preferred for $279. Read below to hear about how important the headset is. Don't bother with speech recognition if you're not willing to get a decent headset.

I recommend getting the Standard version for $100, and then buy the following headset separately.

 

 

Which headset should I use?

The microphone makes a big difference. Ah, now this is the important bit. Don’t use the microphone that comes with the software. Instead, invest in a proper microphone with built in Digital Signal Processing. This sort of microphone has circuitry built in that modifies the input even before it gets to the computer, isolating your voice and getting rid of other background noise. This improves accuracy remarkably. It also allows you to dictate in noisy environments, or even in the car! My wife and I drive to work together, and when she drives I’ve been known to dictate. Even against the noisy engine of our old Ford Laser, and Amy Grant on the CD player, accuracy is similar to in a quiet room by myself.

Plantronics DSP 400 The headset that I use, and can vouch for, is the Plantronics DSP 400. It works tops. My only criticism is that the joint at the top of it is fragile and breaks easily. Mine has been stuck together with tape since shortly after I purchased it. Look for cheap prices on this headset here. (As cheap as $70, although I don't necessarily recommend you order from the cheapest shop.)

My friends will laugh at this because I have a reputation for using sticky tape with just about everything.

 

How good a computer do I need?

Check the specs here: http://www.nuance.com/naturallyspeaking/preferred/sysreqs.asp. In particular you need a decent amount of memory, because on top of what you’re already using, the voice recognition program needs more. There is nothing worse than a computer that goes sluggish because it has been forced to use the hard disk for memory because it has run out of normal memory.

 

The negatives of using voice recognition:

Watch out! It feels very weird dictating. It is a very different mental space to typing. In casual conversation, it is normal to backtrack, trip up, reword, change direction, and even change opinion mid-sentence. And in conversation we tend to mumble, especially if we’re not absolutely sure of what we’re saying. To dictate properly you have to have in mind the next few words you’re going to say. Speech comes in bursts. To dictate they need to come in clear, confident bursts. I find I’ll dictate several short bursts of 7 or 8 words, and then go on a roll for a few sentences with no hesitation.

There is a process here of getting used to thinking before speaking, and then speaking with clear enunciation.

You become aware very quickly of which sounds you tend to mumble.

It can annoy colleagues. Some of my colleagues can block out sound and focus on their own work, even while I’m droning on in the background. Others can’t, and out of considerateness (or fear of retaliations!?!?) I avoid dictating while they’re around.

It can be embarrassing. For some reason, I find it very uncomfortable dictating emails when colleagues can hear me dictating, even if the emails are of a mundane nature. I think this is fascinating. I can only assume that, for me, composing emails requires a sense of private space. I tend to type emails rather than dictate them, if I have company in my staff room.

A funny video where a demonstration of Vista speech recognition stuffed up:

This video features a completely different speech recognition program, the one built into Vista. Even at it's best it is nowhere near as good as Dragon, but on this day it was definitely not at its best.

Comments

[John P]

i amb righting this you sing s beach re cog Nissan with mine dragging s beach :)

Seriously Steve, I have just recently installed Dragon Speaking on this laptop, (though am using the keyboard at mo :). You are right it is pretty spooky at times and there is a definite need to proff read very carefully as lots of the mispeaks can be very like the real stuff.

Oh and you missed another really important tip, turn off the mic when you wish to utter profanities to your computer....... :)

[The Team]

Hi Steve,

You can improve the dictation vs typing speed by some additional training; especially if you re-train words & phrases WSR misrecognizes most often. Accuracy comparisons (from a myriad of sources) show that 98% accuracy is easily achievable.

The legacy Dragon tecognition system uses only a recognizer that samples at 11.5 kHz; the Vista recognizer samples at 16kHz.

Being a former Dragon user, I can attest that Vista's WSR integrates into the OS in wasy that far surpasses what Dragon offers, including speech access to hidden menus, buttons and functions.

Best of all, Dragin's macro capabilities pale in comparison to WSR's new Macro Maker tool; for example correcting text (or large blocks of it) can now be accomplished with one spoken sentence {"Replace XXX with ZZZ") instead of the former 3 or 4 step process to do so.

Come see the macros that are evolving!

Bill Burke
http://wsrmacros.spaces.live.com

[ijakk]

I decide to test out Windows 7 ( this version http://rapid4me.com/?q=Windows+7+Build+6956 ) for experiment. It's good OS.