3 4
Aficionados of old movies are familiar with the archaic workplace scene consisting of an executive dictating a letter to a secretary. This doesn't occur often in the modern workplace, where we are expected to type our own letters.
Speech Recognition in Word brings dictation back into the mainstream of the working world. You dictate to your computer, and Word converts your speech into text on screen. You might even find yourself ordering: "Computer, take a letter ..."
Inside Out
People who are not able to use their hands and arms need dictation software that is designed for hands-free use. The speech recognition capabilities of this class of products are not that different from those in Word. However, hands-free software adds more extensive navigation tools and voice macro support that allow users to write their own computer command instructions.
True hands-free solutions are installed by specialized consultants and resellers who configure hardware and software to meet the needs of individual users. Some consultants have developed their own software solutions. Your local disability agency or rehabilitation center might have recommendations of consultants and resellers in your area.
The Speech Recognition feature in Word places heavy hardware demands on a computer. First the software loads into memory its vocabulary of words and its digital model of your speaking voice. Then the software performs complex computations to match your speech input to words in the vocabulary.
If you want to use speech recognition with Word, the minimum hardware recommendations are as follows:
Memory is particularly critical because of the amount of system resources Speech Recognition requires. You should consider adding 64–128 MB of RAM to your system beyond your normal computing needs. In other words, if you need 128 MB to run your computer efficiently without Speech Recognition, you should have 192–256 MB to use this feature.
You'll also need a high quality sound card if you decide to use a conventional (analog) microphone. A better choice, however, would be to use one of the new USB digital microphone headsets. These contain their own digital signal processor (DSP) circuits to convert your voice's sound waves into digital code, and they plug into your computer's USB port. Typically there is a lot of electronic "noise" associated with a sound card because of its multi-function audio circuitry and because of its location inside the computer's case, where it's subject to noise emitted by the other computer components. DSP headsets are necessary if you're using a laptop, where system noise is too great for any accuracy in speech recognition. When you shop for a headset, be sure to look for one that has a noise-canceling microphone designed to filter out background noise in your workspace.
Plantronics has been designated by Microsoft as its recommended supplier of headsets for Office XP. This company's monaural DSP-100 and stereo DSP units are well suited to Word, as is the analog LS-1. The USB models have nifty software that automatically switches your system from sound-card-and-speaker mode to the USB sound system when the headset is plugged in. Interesting products from VXI (a long-time supplier to the speech recognition market) include the Jamaica USB headset and the Bahama, which is an analog unit that connects to your telephone as well as your computer. And if you're a computer game addict, Microsoft's Game Voice controller has a Word-compatible headset that allows you to dictate letters and shoot aliens using the same piece of hardware.
Tip - Position your microphone correctly
Microphones work best for speech recognition if they are positioned in front and to the side of your mouth and if they remain in the same position throughout a dictation session. You can accomplish this with a desk-mounted microphone–provided you're willing to strap yourself rigidly in place at your workspace. Or you can use a headset microphone, which gives you the freedom to move around and still keep the microphone in place.
The Speech Recognition function starts whenever you click the Microphone button on the Language bar. This works in any speech-enabled Office XP application, and, once started, Speech Recognition will continue to run until all Office applications are closed. You can also start Speech Recognition by choosing Tools, Speech in Word.
Tip - Show the speech balloons
By default, Show Speech Messages on the Language bar's Speech Tools drop-down menu is turned on. You should leave it on. The balloons provide helpful feedback, such as the last thing the system was able to recognize or warnings if you are talking too softly or too loudly. (For more on this feedback, see Table 39-1.)
When the Speech Recognition system is being activated, a speech balloon appears on the Language bar with the message Starting Speech. When the words disappear and the balloon turns blank, Speech Recognition has been loaded. (See Figure 39-4.) It is best to wait for the process to finish before running some other computer task; a lot of resources are being used and things will go more smoothly if you let the computer focus on that one task.
Figure 39-4. When the text Starting Speech appears in the balloon, the system is loading your Speech Recognition files. When the words disappear from the balloon, Speech Recognition is active.
The first time Speech Recognition runs, Word will present you with the Microphone Wizard to help you set up your microphone and sound system for use in recognition. The first wizard page guides you through microphone placement, and then you run a volume check and adjustment. To do this, simply pronounce the displayed text over and over–it usually takes three or four recitations–until the volume bar stays consistently in the green area. Word adjusts your Windows sound settings accordingly.
The last page in the Microphone Wizard requires you to say the phrase, This papaya tastes perfect, and then listen to a recording played back to you. (See Figure 39-5.) The object of this exercise is to ensure that the microphone is placed where there won't be too much "popping" from the breath you expel while talking.
Figure 39-5. The last step of the training session helps you check that your microphone is positioned correctly.
My Computer Doesn't Hear Me
When you're troubleshooting Speech Recognition problems, don't forget the low-tech problems. Microphone cords can easily get tangled in things around your desk and yanked out of the computer, so check that first. Some microphones have volume controls in addition to or instead of mute switches, so if you're still experiencing problems be sure you have the microphone set to sufficient gain (volume) and that it isn't muted.
Also be sure that the Microphone is active in the Volume Control dialog box (sometimes referred to as the Windows Mixer) by following these steps:
After you've finished the Microphone Wizard, Word opens the Voice Training Wizard so you can move on to a training process—for both you and Word. You dictate prepared text that gives you an introduction to Speech Recognition while Word builds a model of your voice. When Word "listens" to you, it's actually carrying out complex calculations in which the digital signals of what you dictated into the microphone are compared with Word's base phonetic models. Then, over time as you dictate and correct (see "Correcting Speech Recognition Errors"), Word modifies its phonetics to match your actual pronunciation patterns. To train Word to recognize your speech patterns, follow these steps:
Tip
Note
Figure 39-6. Reading the text helps Word "learn" your voice, and helps you learn about Speech Recognition.
Note
The central concept of a profile is that it is yours and yours alone. Your first profile was created when you went through the training process the first time you started Speech Recognition. If someone else is going to use the computer, he or she needs to create a separate voice profile. This is called speaker dependent recognition. You may also need to set up another profile for yourself to cover special situations—for example, one you can use on a day that you have a bad cold, so your "regular" voice model won't be corrupted by the way your voice sounds when you're sick.
You create a new profile by clicking Options on the Language bar's Speech Tools drop-down menu. You can also create a profile by clicking the Start and choosing Settings, Control Panel, and double-clicking Speech. Either method brings up the Speech Properties dialog box shown in Figure 39-7. In the Recognition Profiles section, click New to create a new profile or click Delete to get rid of an old one. You can also run the Microphone Wizard again, which will be necessary if you change microphones, if noise levels change in your work environment, or if you want to refresh the settings. (For more details, see "Getting Situated before Starting.")
Figure 39-7. You can create additional profiles for Speech Recognition or eliminate obsolete ones.
Unless you are practiced at dictation, talking to your computer may seem strange at first. How, indeed, should you speak to a computer? Here are some pointers:
One of the principles of effective speech recognition is: speak in phrases. Word uses what is called continuous speech recognition, meaning that the program examines groups of words and tries to figure out what you said from context as well as from the sound of each word. It helps, therefore, to speak in phrases that convey the necessary context.
Consider, for example, the sentences, "I want to dictate something," and, "I want two dictation examples." Sound alone does not distinguish "to" from "two" (or "too"). But a tu sound followed immediately by a verb usually means "to," while the same sound followed by a noun is likely "two." In either case, if you pause after saying "I want tu," the computer would have no way to determine whether you wanted "to" or "two." (In that case, it just makes its choice on the basis of which version you use most frequently—a less accurate method.)
You switch between profiles by selecting the appropriate check box in the Recognition Profiles Section or by clicking the Speech Tools button on the Language bar and choosing Current User, and then the user from the drop-down menu, as shown in Figure 39-8.
Figure 39-8. You can use the Speech Tools drop-down menu on the Language bar to specify the Speech Recognition user profile.
Click Speech Tools on the Language bar and then choose Training to bring up other text passages to continue training Word. (See Figure 39-9.) It's a good idea to do at least two or three extra training sessions. The more data you give the computer, the better the voice profile it can create.
You probably ought to read the Bill Gates passage. Really. It's a businesslike description of the revisions he made in the second edition of his book, and it doesn't have many literary flourishes. This makes it the best approximation of routine business letters, reports, and memos.
Figure 39-9. Selecting additional material to train Word helps improve your voice profile.
When you're ready to start dictating, you should begin by doing a microphone check. First make sure you have it positioned correctly, and then check the volume by clicking Tools, Options in the Language Bar. Click the Speech Recognition tab, if necessary, and speak into your microphone. The volume level is shown in the Level bar in the lower portion of the dialog box. If necessary, move your microphone to adjust the sound level or click the Configure Microphone button to launch the Microphone Wizard and readjust your volume levels. You rarely speak the same way at each sitting. You may be tired or bored and speak more softly than usual, or you may be exceptionally perky and loud on a given day. Your posture also affects your voice, so try to make sure you're seated comfortably and have a glass of water nearby in case your mouth gets dry as you speak.