Scott Hanselman

Typing Test twice, once with Voice Recognition

September 30, 2007 Comment on this post [13] Posted in Musings
Sponsored By

People give voice recognition a bad name. I use it all the time. I took the typing test that Daniel Moth blogged about (typing test is here) typing the way I usually do. I can't type as fast as I could as my hands are a little rough, but I do OK for old hands.

Here's my results, first try, on a Natural (split) Keyboard:


Number of words typed: 217
Test duration: 3 min
Speed: 72.6 words/min. (363 keystrokes/min.)
Error penalty: 27
Accuracy: 87.6%

Meh. I used to do 100, no longer. (Plus, it's a cheesy test because you have to read, manage the scrolling and spit it back out again on the keyboards. I got flummoxed with the scrolling. And I'm notoriously sloppy.

Here's the SAME test but using Vista Voice Recognition and a cheapo Logitech USB Headset Microphone (I didn't even use my $300 podcast microphone):


Number of words typed: 377
Test duration: 3 min
Speed: 125.6 words/min. (628 keystrokes/min.)
Error penalty: 42
Accuracy: 88.9%

Note the number of errors. I know it's not 98% like some people, but I can almost double my personal typing speed with comparable errors using Voice Recognition.

Note that I only took these tests once each, both times cold, one with hands, one with voice and  the test gave me different text each time.

You have to talk like a newscaster, but seriously, stop giving Voice Recognition a hard time. Videos of Voice Recognition trying to write Perl code are cutesy, but not reality. Yes, it'd be nice to have custom templates out there for writing code with voice (there are some) but CodeRush helps me greatly. Ultimately voice recognition is for things like blogs, email and prose.

Here's a VERY boring, very poor sound/visual quickie video of me actually taking the test. Note that I (gratuitously) stopped to correct two errors, using only voice in the middle. Otherwise I'd probably have gotten 140wpm. Sorry about the poor focus, I probably should have used Camtasia.

And yes, this blog post was written with voice recognition.  So phooey on you. I encourage you to give it a try with a nice microphone, if even only for email and blog posts.

About Scott

Scott Hanselman is a former professor, former Chief Architect in finance, now speaker, consultant, father, diabetic, and Microsoft employee. He is a failed stand-up comic, a cornrower, and a book author.

facebook twitter subscribe
About   Newsletter
Hosting By
Hosted in an Azure App Service
September 30, 2007 6:51
sweet, i tried normal typing and got these:

Number of words typed: 259
Test duration: 3 min
Speed: 86.4 words/min. (432 keystrokes/min.)
Error penalty: 3
Accuracy: 98.8%

not bad i thought...i'll have to find my headset and compare...
September 30, 2007 7:08
Interesting. My issue is that when mistakes are made are made by the voice recognition, they take longer to correct than with typing. This assumes you want to correct the typos - this may not also be true, the occasional mistake may be ok in a casual email. But in those communications where you can't have mistakes, it takes longer to say "error" or whatever, re-say the word, and hope that the VR gets it right this time and go on. This is compared to just hitting backspace a few times, particularly since we're all so good at backspacing, despite its obnoxious location on keyboard.
Also, I've trained over the years to think as a I type, rather than speak while I think (that doesn't sound quite right, does it?). I don't have any science to back me up, but it seems like a totally different cognitive process. Not wrong, but different.
September 30, 2007 7:18
One important thing to note when comparing these results is that a single letter mistake - common when typing - counts as one error. However, when using voice recognition, common errors are words being switched for homonyms - each single "speecho" mistake, in this case a whole word, counts as many errors.

I propose my voice reg. erros would be easy to fix and that my % errors while speaking was quite low.
September 30, 2007 8:29
Yup, did the same test. Had to take it twice because the scrolling part just pissed me off the first time. Not being a quitter type, I went back and took it again but wasn't happy. I've lost 20% of my 9th grade typing speed also (and added a few more errros). haven't tried the voice recognition stuff, but you have me thinking with 120 wpm. (nice live preview BTW, very cool)
September 30, 2007 10:21
Interesting result Scott, but since when do you have a beard? Is that the result of working at home ;-)

September 30, 2007 12:14
Martijn - You've caught me. It's the result of massive laziness. I'll be sure to clean up soon. ;)
September 30, 2007 14:07
Which mic did you use? Was it the Samson, or a headset? I wonder if there would be a noticable difference between mics. Heck, in a somewhat noisy environment I wonder of the Jawbone can work with a laptop's Bluetooth as a heaset and how that would do?

Looks likes I have something to hack around on. :)

October 01, 2007 3:38
I took the test and did 70 words per minute and 90.1% accuracy, 212 words typed. I didn't correct anything along the way - I was more like Ham from Toy Story - had to keep on going! Haven't tried the voice recognition yet, but will.

The thing that I thought about, however, is that I don't really care if I can do more than 70 words per minute by speaking because most of my interactions with a PC are the coding kind, and I'm simply not going to sling code at 70 words a minute. I don't know that I've met anyone that works that way, and for my .NET development, I rely a LOT on Intellisense to complete things when coding. I've never attempted to write any code using voice recognition, but I'm skeptical that it would work all that well - but who knows? I think and type and think and type and then think a little more.

If I were working on a paper or blog post or something like that, voice recognition would be the way to go, particularly if it was reasonably accurate at figuring out my mumblings.

I noticed, Scott, that when you were talking on the video, you had to explicitly say "comma" and "period", and I wondered if that would get old. I think it might "comma" but suppose you get use to that "period"

Ultimately, If I'm dictating to my iMac or my PC, I don't think I'm all that concerned with reading while I speak (after the initial 'golly, that's neat' period wears off), rather I'd want to be able to dictate while I was doing something else that didn't involve talking.

I wonder how Vista voice recognition compares to the Dragon Naturally Speaking tool.

October 01, 2007 4:53
Hi Scott,

Any good Voice Recognition Tools for Windows XP?

October 01, 2007 5:55
Whoever came up with the scrolling algorithm for that typing test should be fed crusty bread and water for a week to pay for their transgression. It was like a penalty for typing more than 60 words a minute.

Anyway, it seems like they could have just made the window bigger.

October 01, 2007 11:15
It's a completely specious comparison: it's the editing that kills you with voice rec.

I spend far, far more time editing my words than typing them. And the same principle applies to code, too.
October 01, 2007 17:39
I'm thinking about getting into this Screencasting thing but I REALLY HATE THE SOUND of my voice when recorded. Maybe it's my crappy $30 headset. Any recommendations? I don't mind spending a little money if it means not sounding like a chipmunk :)
October 01, 2007 22:28
Funny you mention this. I broke my arm at the wrist a couple of weeks ago and seriously thought about going out to get everything I would need to work via Voice Recognition. In the end I sucked it up and suffered thru the pain and awkwardness of typing with an extra 5lbs of plaster on my arm.


Comments are closed.

Disclaimer: The opinions expressed herein are my own personal opinions and do not represent my employer's view in any way.