« AMD Live! A free-to-air TV killer? | Main | Get your Vista beta ... today! »
June 08, 2006
Finally, my computer understands me
In a change from our regular schedule, it’s not Charles, the Chief Toad, here, but Jeremy Howard, the Bleeding Edge toadling. For those of you who have only recently come to the Bleeding Edge pond, you may not know about me – please see my first article to find out more.
The reason I haven’t been seen for a while, is because I have been battling with RSI again. Whilst my health is now much improved, I’m trying to avoid typing so that it doesn’t happen again. Instead, I’m using voice recognition.
Normally, I use Dragon Naturally Speaking. The Chief Toad has written about this before. However, I’ve just arrived in Boston for a three month trip, only to discover that I had forgotten to install it on my new laptop. Rather than purchase a whole new copy, I thought instead that I would try using the voice recognition built into Microsoft Office. Many Microsoft Office users don’t even realize that voice recognition is installed – however it has come free with the software since Office XP. Having said that, the version in Office XP was almost unusable.
I’m now using voice recognition in Office 2003 for the first time. In fact, this entire article was dictated using it. I’m very impressed. In my unscientific opinion, the recognition accuracy is just as good as Dragon Naturally Speaking v7 (v8 has come out recently, but I haven’t tried it). I am dictating directly into Microsoft Word, and the software is providing a full set of correction, selection, and movement tools. Microsoft’s speech recognition continues to learn and improve as you use the software. For instance, this is the first time that I have used it; in the first paragraph that I dictated, and there were three or four dictation errors. However, so far in this paragraph, there have been no errors. So, already the dictation accuracy is improving.
If you would like to try using Microsoft speech recognition yourself, firstly, get hold of a good noise-cancelling headset. Then, open up Microsoft Word, and press F1 to bring up the help window. Type "speech recognition" into the search text box, and follow the simple instructions. If you have any questions, please use the "comments" link beneath this article, and I will do my best to answer them.
Posted by Jeremy at June 8, 2006 10:43 PM
Trackback Pings
TrackBack URL for this entry:
http://bleedingedge.com.au/cgi-bin/mt/mt-tb.cgi/907
Comments
I just came across an interesting comparison of Naturally Speaking vs MS Office at Anandtech. It shows that for fast speech, MS Office has half as many errors as naturally speaking!
Posted by: Jeremy Howard ![[TypeKey Profile Page]](http://bleedingedge.com.au/blog/nav-commenters.gif)
at June 8, 2006 11:09 PM
The punctuation leaves something to be desired. ;-)
cheers, Paul
Posted by: Paul at June 9, 2006 08:31 AM
Have you thought about using something like Dasher?
I found it didn't take me long to get used to it, and it gives my hands a nice break at times.
Posted by: Alex at June 9, 2006 07:59 PM
The punctuation is dicatated manually, so any mistakes are due to me, not the software. Having said that, I can't see any problems with the punctuation in the article - I can only see a stray "and". (That's a common problem with the software - it seems to "hear" conjunctions sometimes that just aren't there!) If I have made some silly mistakes in my punctuation, then I apologise!
Dasher looks very interesting. However, the site says it does 30-40 WPM. Voice recognition does 120-140 WPM!
Posted by: Jeremy Howard at June 10, 2006 01:17 AM
It depends how you use dasher. It also depends on how many wpm you think :)
Where I find it particularly useful is writing technical documents, particularly in LaTeX - I have used one of my works as a training text, so it expects to use the jargon and mannerisms I usually use. Kind of frightening when it tells me that the next word I'm going to use is the one I was going to use though.
Posted by: Alex at June 11, 2006 06:51 AM