Compatibility Note: At present, the Knox Center has Dragon on two Windows computers in room 2064A (the side room): version 14 on the computer near the entry door, and version 12 on the computer near the back wall. Documentation for version 12 is below, and documentation for version 14 will be added shortly. This software works best with standard software such as Word, Internet Explorer, and Firefox. It will have variable compatibility with other programs.
Required Equipment Note: To use Dragon at the Knox Center, you will need to bring your own USB microphone. A good choice is the Cyber Acoustics AC-850 USB Headset, which is available for $28 at the U-M Computer Showcase.. Headset microphones are preferred, since they tend to pick up less ambient noise than desktop models. If you would prefer to use a desktop or wireless microphone, please contact Jane at firstname.lastname@example.org to discuss options.
The documentation below is not intended to be exhaustive, but to provide sufficient information for most people to use basic Dragon features successfully. Other resources include the information available under the “Help” menu in Dragon. All U-M students are welcome to contact Jane at email@example.com with questions or to schedule 1:1 training sessions with Dragon.
Speech recognition software has improved dramatically over the last several years. But even Dragon, the best product currently available, is still not the wondrous technology shown in Star Trek. With any user, it will make periodic recognition errors, due in large part to three factors:
The English language is full of homonyms (words that sound alike, but have different spellings and meanings). Dragon is unreliable in its ability to correctly choose the right homonym, so it may hear "grease spot" and put up "Greece spot," or vice versa.
Dragon expects a consistent voice, which no one has. Our voices change naturally over the course of a day, so that even between morning and afternoon there can be enough variation to affect accuracy. Stress, fatigue, and other factors can also cause vocal modifications and consequently reduced accuracy.
The longer and more unusual a word, the more likely Dragon is to recognize it correctly. It tends to have trouble with short, vowel-intensive words such as "an," "and," "it," etc.
The Dictation and Mouse Commands sections of this document contain tips on how to maximize your likelihood of success with speech recognition. In addition, , there are two general practices that can help most people improve their experience as Dragon users:
Maintain your sense of humor. Staying amused rather than frustrated at Dragon's errors will keep strain out of your voice, thereby increasing recognition accuracy and decreasing the likelihood of injury.
Take some type of vocal training—singing, oratory, drama, storytelling, etc. Ask for warmup exercises and other tips to avoid vocal strain. Then apply these when using speech recognition.
Unfortunately, the software still has difficulties performing well for people who have severe speech disabilities or extremely heavy accents.
Dragon users can experience effects such as hoarseness and discomfort if they overuse or improperly use their voice when dictating. There is even some research evidence that people with carpal tunnel and other repetitive stress injuries may be more susceptible to vocal strain. It's therefore important to make sure you're taking care of your voice when using Dragon:
- Use your natural voice—not too fast or slow, not too loud or soft, and vary your pitch rather than speaking in a monotone. (The voice children's librarians use when telling a story works quite well.) No need for exaggerated enunciation.
- Drink plenty of water. Avoid drinks containing caffeine or lemon, since these will cause faster dehydration.
- Take brief but regular breaks—a few per hour; more if you feel your voice straining.
- Avoid using eucalyptus and menthol lozenges. And if you smoke, this is a good reason to quit.
At a minimum, you will need to set up your voice file and go through a microphone check. It is strongly advised that you go through the "Short" training to give Dragon a better sense of what your voice is like.
Plug in your USB microphone. (Remember which USB port you used; you will need to use this same port every time you work with Dragon.) Then open Dragon by going to the start menu and typing “Dragon”. The "Open User" dialog will appear. Click on the "New" button on the right side.
You will then go through a series of prompts to indicate the following:
- Your user name (may be whatever you like)
- Your age (helps Dragon predict some of your likely vocal characteristics)
- Your geographic region (to determine some default spellings, etc.)
- Your accent
- Your speech device (be sure to choose “USB” if you are using a USB microphone)
You will then be given a chance to review your information. Click “Back” to go back and change anything. When you are satisfied, click “Create.”
After the profile is created, you will be prompted to position the microphone so that it is parallel to the corner of your mouth. When you have done this, click "Next."
You will now see the following dialog box prompting you to do a volume check:
Click "Start Volume Check." The greyed text will turn black. Keep reading and re-reading this text until you see the word "Finished!" in the box to the right of the "Start Volume Check" button. Then click "Next." (If the program reports unsatisfactory results but gives you the option of continuing, do so; usually there is no real problem.)
Repeat step 4 for the Quality Check on the next screen:
You will then see the following dialog box for beginning voice training.
- Most people will choose “Show text with prompting.”
- “Show text without prompting” is helpful if you have difficulty with reading; it allows you to print out the text or otherwise access it in different ways. Contact Jane at firstname.lastname@example.org if you need assistance with this.
- “Skip training” is not recommended.
When you are ready, click “Next”. Steps 7-9 below assume you have chosen “Show text with prompting.”
On the next screen, you will be asked to click the Go button and then read two sentences: “Welcome to general training” and then “Training is about to begin.” At any point during this training exercise, if Dragon doesn’t recognize something, it will put a yellow arrow at the point where it wants you to begin re-reading.
When both sentences have been recognized, you will be given a choice of readings. I recommend "What to expect from speech recognition," since it gives good advice about using Dragon. If you prefer an easier (and charming) read, try "Pigs, wolves, owls and more." Then click "OK."
Read the text on each screen as it comes up. If Dragon doesn't recognize what you've said, it puts a yellow arrow on the word where it wants to you to start reading again. You will read through screens for about seven minutes.
When you have finished, the program will automatically create a user file for you, then prompt you to review your existing emails and documents for vocabulary not already in Dragon's dictionary. Click “Next” to skip this.
Click “Next” on the following dialogue (“Let Dragon automatically improve your accuracy”) to skip it.
On the “Help us improve Dragon” dialog, click “Don’t run Data Collection,” then click “Next.”
On the “Congratulations” screen, either click on one of the two links for more information, or click on the “Finish” button to continue.
An interactive tutorial will come up. While going through this is recommended, if you want to bypass it click on the “Learning More” tab and then the “Finish” button.
Many features of Dragon are controlled using the DragonBar, which looks like this (the view below has been truncated to fit on this web page):
Click on the microphone (outlined above with a red box) to turn the microphone on and off. You can also press the + (plus) key on the keypad. Wait for about five seconds after the microphone is turned on, then begin speaking.
If the checkmark next to the microphone is green, as in the example above, it means you are using an application optimized to work with Dragon. If it's gray, it may not be fully supported.
Tips for successful dictation:
- As much as possible, plan what you're going to say before you begin dictating.
- Some people prefer to correct Dragon's errors right away; others like to get their thoughts down and then make the corrections. Either way is fine. If you forgot what you actually said, highlight the text in question, go to the DragonBar, choose the Sound menu, and select "Play that back." You will hear your voice dictating the passage in question.
- If you get distracted by seeing the dictated text come up, turn off the monitor.
Keep in mind that you will have to speak all punctuation; Dragon will not enter it for you. (There is a setting that allows periods and commas to be entered automatically, but this is based on user pauses and is highly unreliable.)
When Dragon makes an error, it's possible to simply type in the correction. However, by going through the procedure listed below, Dragon is more likely to "learn" the right word or phrase, reducing the likelihood of future errors next time you say the same thing.
To correct an error that Dragon has made:
Say "Correct" and the word or phrase as it appears (not what you actually said). A correction dialog will appear (in the example below, the desired phrase was "One is a top"):
If the desired word or phrase appears in the dialog, say "Choose" and the number associated with the correct option. For example, if the desired formatting had been, "OF THE TOP," you would say "Choose 2."
If the desired word or phrase does not appear, say "Spell that." The following dialog will appear:
Try using the Military Alphabet to spell out the corrected text and avoid confusion between similar sounding letters. (See a chart of the Military Alphabet and its preferred pronunciations.) For example, instead of saying "c a t," you would say "charlie alpha tango."
Using the example above, you could use the Military Alphabet to say, “india sierra space alpha space tango oscar papa" and the dialog would then look like this:
Since the first item is correct, say "Choose 1" and it will be entered into your document.
If you prefer, you may also type your correction and then double-click on the correct item.
Mouse commands are critical for people with little or no dexterity. Most people like to use at least a few common commands so that they don't have to switch between dictating and using the mouse; however, the keyboard and mouse remain fully functional when Dragon is active.
Tips for successful use of mouse commands:
Dragon is skilled but not perfect at distinguishing between dictation and commands. There are also times that you may want to have phrases that are usually commands (such as "Go to sleep") interpreted as dictation; e.g., in the sentence, "Now is the time for all good children to go to sleep." Try pausing less between words when you are saying commands. You can also hold down the Ctrl key while speaking to indicate a command, or the Shift key to indicate dictation.
People sometimes try to improvise commands (e.g., "Wake the heck up!"), but Dragon only knows specific commands. If you have difficulty remembering these, print out a list and keep it with you. You can also use the command "What can I say" to get a list of commands appropriate to the browser, word processor, or other application you're currently using.
|Turn microphone off||"Go to sleep"|
|Turn microphone on||"Wake up"|
|New line (equivalent of pressing Enter once)||"New line"|
|New paragraph (equivalent of pressing Enter twice)||"New paragraph"|
|Move cursor within a line||"Move left (or right) X characters" (X can be any number that is 20 or less)|
|Move cursor among lines||"Move up (or down) X lines"|
|Erase the most recently spoken text||"Scratch that"|
|Emulate a key press||"Press (key name or combination)". Examples: "Press spacebar", "press Control Escape", "press right bracket"|
|Emulate a left mouse click||"Mouse click"|
|Emulate a double click||"Mouse double click"|
|Emulate a right click||"Mouse right click"|
In many cases, simply speaking the name of an interactive item (e.g., a menu or a web page link) will activate it. However, when this doesn't work, MouseGrid is a tool that will work in almost all situations for emulating a mouse click.
Say "MouseGrid." A grid will appear, dividing the screen into nine sections.
Say the number of the section where the item you want to click on appears. A smaller nine-section grid will appear in the section you specified.
Continue saying numbers to narrow down the grid until the mouse pointer appears over the clickable item.
Say "Mouse click." The item should activate.
Periodically, all public computers, including those at the Knox Center, will be updated to install new software, fix problems, etc. This happens regularly just before the start of each semester, and occasionally throughout the year with little or no warning. As a result, your voice file will be removed from the computer. If you want to avoid going through the training process again, you will want to back up your voice files to a thumb drive or other portable storage device.