Windows Speech Recognition

This is a quick review of Windows Speech Recognition, which I discovered thanks to this article. Just a few bullet-point thoughts.

  • Compared to the set-up and training for other voice recognition I’ve tried (although I’ve only really used two applications, and one of those a very long time ago), this is relatively quick and painless to get started. Also, it’s free, and it’s already installed in the majority of modern Windows machines. So that’s a good start.
  • The capabilities of the software are pretty powerful. You can tell it to click anywhere on the screen for you, it can highlight and select, it can delete. Say, “correct” and the word(s) it has mistyped, and it will learn from its mistakes. Opening software, switching between programmes, “hitting” keyboard keys can all be commanded. And, of course, in key programmes like desktop email clients and text editors, it enters text. The only issue is getting to the stage where all this works smoothly enough to be quicker than just typing, clicking or hitting the key yourself.
  • This software really, really works better in a quiet environment. Everything from your colleague typing across the room, to sirens going past your office (even with the window shut and from four storeys up) is likely to deceive, or at least confuse Speech Recognition.
  • Speaking of colleagues, if you don’t have your own enclosed space, then even if those you share with are spookily quiet (and never type), you may feel a little silly talking to your computer. Especially when the software is getting used to your voice, forcing you to repeat the same command five times before letting out an exasperated sigh and doing it yourself.
  • The big let down for this programme is that it is just a little too inaccurate, and takes a bit too long to train. During the training phase (training you in the commands, as well as training the software in your voice), only sending the shortest, simplest emails is close to being quicker (or at least easier) by voice than with your hands. On everything else you’ll spend so much time trying to get Speech Recognition to click, highlight or open, or to correct the phrases it’s interpreted incorrectly, you’d have finished the same task in a fifth of the time on your own.
  • All of that said, I’ve only spent an hour or two trying to use this software. If you have, and are willing to invest, the time to learn how it ticks and teach it how you talk, I have a sneaking suspicion this could become quite a nifty little tool. I’m planning to keep at it for a little while longer, at least.

So those are my initial thoughts. I hope you’ll at least have a play if you haven’t already discovered (or discarded!) this possibility. Below is a sample dictation, both after correction and with the original output. Sorry the content’s so uninspiring!

This is a test recording. This is the original version. Everything that has been typed I’m including. So far the software is doing an excellent job! In fact so far it’s only got one word wrong, and that really could have been the right word anyway. Now it’s starting to get a little bit confused. It probably helps that I’m speaking quite slowly, and thinking very carefully about what I want to say before I speak. Sadly, that’s not the way I talk when I’m actually trying to type something for real! Hopefully you can see one or two instances where it hasn’t quite understood, but this example seems to be selling it better than anything I’ve typed so far!

This is a test recording. This is the original version. Everything that has been typed I’m including. So far this software is doing an excellent job! In fact it’s only got one went on, and that really could have been the right word anyway. Now it’s starting to get a little bit confused. It probably helps that I am speaking quite slowly, and thinking very carefully about what I want to say before I speak. Sadly, that’s not the way out torque one actually trying to type something for real! I believe you can see one or two instances that hasn’t quite understood, but this example seems to be selling it better than anything after so far!

The top paragraph has been corrected using the usual correction voice commands. It was more quick and painless and some of my attempts to correct things in the past. So far so good! This paragraph is not needed to be corrected tour so far… Well, until now, that is! Trying to highlight the mistakes in the original version of the paragraph was somewhat more painful, but didn’t quite have me pulling my hair out…

NB. The underlined mistakes were originally highlighted in yellow.

See the article linked at the top for more on activating Speech Recognition…if I haven’t completely put you off!

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s