How To Use Windows Speech Recognition on your PC

If you use Microsoft Windows and have a headset microphone, as many gamers do, you could probably double your typing speed immediately by using a little-known Windows feature.

Microsoft is low-key about the fact that Windows has a full-blown speech recognition system. But instead of hunting and pecking while trying to remember what you wanted to write, you could use Windows Speech Recognition (WSR) to decide what to say, say it, and see the words appear on the screen. However, some user skill is involved—don't expect an immediate nirvana of effortless inputting.

Getting Started

Put on your headset microphone and hook it up. Click Start, then Control Panel. You'll see a window with a list of applets. One of them is Speech Recognition. Click it. You'll see a new window with five more applets. Select Set Up Microphone and follow the directions to make sure your mic will work with WSR.

When that's done, return to the Speech Recognition window and click Start Speech Recognition. The WSR control bar will appear on the screen, probably in the top center, and should look like this:



The circle on the left, surrounding the microphone icon, is initially gray, meaning that WSR is not listening. The hatched vertical icon to its right is the volume indicator. The gray rectangle in the middle that says Off is the status readout. On the right end of the bar, the X is the Close command and the bar is the Minimize command.

Talk-to-Type

Open a word processing document and put the cursor where you want to start. Click the microphone icon in the WSR control bar so that the circle around it turns blue and the status readout says Listening.

Start talking. Remember, you'll need to actually say the punctuation marks aloud so the system will know where you want them. Say something such as: "When in doubt, comma, scream and shout, comma, wave your arms, comma and run about, period." If it worked, you should see this:

Editing

Now let's try some editing. For obvious reasons, you have decided to change "about" to "for Congress." Say "select about," and the word about will be highlighted. Then say "delete," and the highlighted word will go away. Then say "for Congress." You should see this:



Now let's try something harder. Say "new line new line." This will insert two carriage returns. Then say "you scream, comma I scream, comma we all scream period," to get this.



Notice how it tripped over a homophone, mistaking ice cream for I scream. Changing the mistake with WSR's Correction feature should help the software avoid the problem in the future. Say "correct ice cream." Notice that you need to pronounce the words that appear on the screen (ice cream) not the words you meant to say originally (I scream.) You should see something like this:



As you can see, WSR has generated a pop-up screen with possible alternatives. In this case, one alternative involves different capitalization and the other involves a homophone. Speak the number of the alternative you want. It will become highlighted. Then say "okay." (If you don't like any of the alternatives, say the correct text again until, hopefully, it's included in the list. Alternately, you can say "spell it," and speak the letters aloud, followed by "okay." And if you really get stuck, you can use the keyboard.)

Menus

With a correction made, it's time to save your deathless prose. Say "show numbers." You should see something like this:



Notice that the menu items and actionable screen elements are now covered by numbers. (The numbers will slowly fade in and out so you don't have to remember what they are covering.) You'll want to click the file item, which is covered by the number 5. To do that, say "click five." You will then see the pop-up window with the file commands. You can use the show-numbers command in that window to complete the action. Also, from the original Notepad window, you can simply say "file" and then "save," or "click file" and "click save," but using the show-numbers command will give you more precision in windows that are crammed with numerous possible commands.

Beyond

Now that you have seen what the WSR can do, go back to the original Speech Recognition window that you accessed in the Control Panel and peruse the other applets. Take the tutorial, print out the reference card and study it, and conduct some of the recognition training. Try using WSR with your favorite software.

Meanwhile, give yourself at least a couple of weeks to get used to this new mode of composition. You can't expect to become productive without some effort and practice, though it undoubtedly pales in comparison to the effort and practice it took to learn to type.

The biggest adjustment will probably be the habitual employment of accurate and precise pronunciation to enhance the machine's recognition accuracy. Remember, you are controlling a machine with your voice, not talking to a person. It doesn't matter if you sound oddly stiff.

The only other serious alternative for general-purpose desktop speech recognition is Dragon Naturally Speaking (DNS) from Nuance Communications. WSR is already present on your Windows machine, and it requires only the acquisition of a microphone. DNS must be purchased and installed but comes with a microphone. DNS has more reliable recognition than WSR. Also, when you select text and then dictate replacement text, DNS just writes over it rather than going through the review procedure that WSR uses, making its use smoother. On the other hand, DNS tends to bog down the machine and consumes more than 3 gigabytes of disk space. WSR is part of Windows and has no visible effect on performance.
Continue Reading