Initial setup and Training

It’s a good idea to read this section before using speech recognition in WordQ 5 – that way you will know what to expect and you will be more successful. Also refer to the detailed help available from the WordQ menu Options > Help > WordQ User Guide.

WordQ’s speech recognition is not the same as modern cloud-based speech recognition that uses much artificial intelligence. Instead, it works offline and it requires some amount of training to be effective. What it is offers is integration with word prediction and speech feedback. It is not intended for complete hands-free operation.

WordQ’s speech recognition is most useful for:

1. Quickly getting your initial ideas down before you forget.

2. Being more productive with phrases.

3. Getting unstuck when spelling a difficult word or phrase that interferes with your train of thought.

Before using speech recognition you must create a recognition profile and train it so that WordQ recognizes your speech. You will read several training texts aloud. Each involves speaking for about 2 – 5 minutes. You will need to do several of these, with breaks in between, so that speech recognition can work more accurately. (You need to train at least 800 – 1,000 words, or about 15 minutes of your speech for best results.) If you find that recognition is poor when you come to dictating, then do some further training, preferably with text that is similar to what you want to write.

Note: The speech training interface described below has been designed for a user who has difficulty with reading. If this is not your case, you can use the standard Microsoft speech training wizard located under Start>Control Panel>Speech. This wizard allows somewhat faster training without speech prompting.

Step 1: Connect your microphone 

• You must connect to your computer a quality noise canceling microphone (certified for speech recognition). A USB microphone is often best. Consider using a headset microphone with earphones that allows you to hear the speech feedback privately without disturbing others. The microphone should be positioned close to the corner of your mouth. You may also use a directional desktop microphone and desktop speakers if you are using the computer in a quiet environment. (In a classroom, it is suggested that you sit at the back facing the front of the class so that the microphone does not pick up sounds from the teacher or other students.)

• You should not change your microphone once you have trained with it. If you do change microphones, or change location, then you should create a new speech profile for each situation and repeat training in each. There are no limits on the number of speech profiles that you may create. 

Step 2: Do a quick test of your microphone

• Choose Speech recognition...> from the Options menu.

• Speak a phrase. You should see a moving horizontal green bar displayed in the microphone section. If you do not see this moving bar while speaking, make sure that your microphone is connected and that the correct microphone is selected from the drop-down list in this section. (You may have several possible microphones listed.)

Step 3: Choose the recognition engine

Ensure that only a Microsoft English Recognizer is chosen. 

• Choose Microsoft Speech Recognizer 8.0 for Windows. WordQ will not work with any non-Microsoft recognition engine that may be displayed. 

Step 4: Create a speech recognition profile

• Click the New button in the Speech recognition dialog to create a new recognition profile. 

• Enter a name for your profile, such as your personal name (e.g., Tom) and click OK. This name will now appear in the list of recognition profiles. 

Step 5: Read your sentence

• A microphone wizard will then be displayed to adjust the volume of your microphone.

• When you are ready, speak a sentence. (You can read the sentence shown, or you can make up your own.) Speak normally. The system will automatically adjust the volume of the microphone as you now speak. 

• Pause after you speak and the Next> button should become available which you should now click. If it is not available, speak another sentence and pause.

Step 6: Train your speech

• Click the Train... button in the Speech recognition dialog to begin training your recognition profile. 

• You will first be prompted to choose a training text. Choose a text that is at an appropriate level (beginner, intermediate, or advanced) and in an area of interest to you.

• A training window will now be displayed. The upper portion of the window will display the training text, one phrase at a time. 

• Click on the Train button to begin your training. You will see a short piece of text at the top of your screen. It will be read to you and highlighted word-by-word. When the highlighting stops and the ‘go’ icon  appears at the top of the training window, say the phrase into your microphone.

• Note It is necessary to wait for the highlighting to stop before you speak – the microphone automatically turns off while text is being highlighted so it won’t ‘hear’ you. Speak in your normal voice – not too fast, slowly, loud, or quietly. 

• After you speak, the computer will write down what you have said in the bottom section of the training window. Once the complete phrase is recognized, you will be prompted with the next phrase. After a while you should find that you develop a rhythm of reading/listening and speaking – your training should then go quite smoothly.

• If you do not say everything or the computer does not recognize your voice, do not worry. After a short pause, the computer will say any remaining words again. 

• Take a break whenever you need one by clicking on Pause. Click Train to continue with training. If you get stuck on a word and it is not recognized, just click on Skip to skip that word. 

• If you happen to say the wrong word, and the computer thinks it is correct, do not stop. Just continue with the training. You do not have to make any corrections now. 

• Along the bottom of the window you will notice a progress bar that indicates your progress through the training. When all of the text is trained the training window will close and save your trained profile.

Step 7: Repeat training

Repeat this training procedure with more training texts. At least three of the full training texts (i.e., all parts) provided should be trained. This is about 800 – 1,000 words, or about 15 minutes of your speech. You can take a break between trainings or do them on different days. The recognition will improve with more training.

Discrete or continuous dictation mode

Discrete: With discrete dictation, what you say is displayed in the word prediction list for your review before entering it into your application. This supports a slower dictation pace. You can type initial letters to refine recognition.

With this approach you control the words that are entered into your documents while improving the recognition accuracy. Consequently, you spend less time correcting errors, and more time on the real task of writing.

Continuous: With continuous dictation, what you say is immediately entered into your application and spoken back.

To reduce complexity, WordQ’s speech recognition emphasizes the task of writing, and eliminates the need to memorize voice commands. You only dictate text