james mckay dot net

Over the past few months I’ve been rather intrigued by some of the reports that I’ve read about Windows Vista speech recognition.  For example, Scott Hanselman claims an increase in speed from approximately 72 words per minute when typing to about 125 words per minute with voice recognition—an improvement of approximately 75%.

Now Hanselman works for Microsoft, so it is only natural that he would give something to do with Windows Vista a glowing report, and let’s face it, a typing test is actually a very artificial way of trying out this kind of thing—you’re not chopping and changing all the time but reading verbatim off the screen.  Other people are less complimentary.  Microsoft’s ill fated demonstration of Vista’s speech recognition went down in history, with the immortal phrase “Dear aunt, let’s set so double the killer delete select all” appearing on the evening news and geek T-shirts.

I haven’t actually used Windows Vista that much until recently.  My old laptop only had Windows XP on it, and I use Windows Server 2003 at work.  However, a couple of weeks ago I got a new laptop, complete with Windows Vista, so I thought I might as well put it through its paces, and so today, I’ve been trying it out.  I’ve set myself the goal of writing a complete blog entry without using the keyboard or the mouse: opening Windows Live Writer, navigating Firefox and Google Reader, finding other pages that I want to reference, and inserting hyperlinks, using speech alone.

My experience so far has been more along the lines of the ill fated demonstration than Hanselman’s glowing report, but to be fair, it’s still early days.  Initially it was painfully awkward, and after a few hours it’s still pretty clunky, but it does seem to be learning from its mistakes and it does get it right about 80% of the time.  The problem is that the 20 per cent of the time when it doesn’t get it right, or when you want to chop and change things, it is very slow and fiddly.  Some things don’t even work: “show numbers” doesn’t show numbers correctly for hyperlinks on Ajax enabled sites in Firefox, and your blog’s category names don’t appear in the “categories” drop down list in Windows Live Writer.

Yes, it’s more comfortable than using the keyboard, but it does take a lot longer, and thoughts only trickle from your brain into your document rather than flowing.  It is also pretty frustrating if you keep chopping and changing things, as I do when I’m writing. Perhaps if I persevere at it I might find it improves, but I don’t think we’re going to be dispensing with our keyboards anytime soon. 

Comments closed (1) »

Posted at 20:06 on 12 April 2008.
Tagged: speech recognition, typing

I got a new phone today. My old one died at the weekend but fortunately there was a spare one available at work that I have been given the use of. My mobile number is unchanged.

Unfortunately it is not an iPhone, but it does have some rather interesting features. One that I rather like is handwriting recognition, which even makes a reasonable attempt at interpreting cursive (joined-up) handwriting. It is accurate enough to be usable most of the time, and though it does make enough mistakes to slow you down, it seems to get more accurate the more you use it. It is certainly much easier to use it for texting than a tiny numeric keypad.

With technology like this, one wonders whether this means that the writing is on the wall for the keyboard as we know it. Speech recognition may have had a bit of a bad press, but it is improving all the time. We now have futuristic technologies such as Microsoft Surface and the iPhone either out or in the pipeline. Computer scientists at Cambridge University have even produced a device that can decode facial expressions.

Sure, all this may still be slow and resource intensive at present, but the technology is improving all the time, and while voice recognition may not be suitable in all settings (you wouldn’t make many friends using it in an open plan office environment), handwriting recognition is certainly rather promising, and it could well be a serious rival to the humble keyboard as our main data input device.

Having said that, I don’t think it’s curtains for the keyboard just yet. It comes into its own when you need to combine speed and accuracy. Skilled touch typists can reach speeds of eighty words per minute or more — speeds at which your handwriting would rapidly become illegible even to human readers. We developers will probably be the last ones to hang up our keyboards in the end of the day — the requirement for both speed and accuracy is paramount when you are writing code.

Comments closed »

Posted at 23:44 on 23 July 2007.
Tagged: handwriting recognition, speech recognition, typing