Dragon voice recognition resources and news

These are some tips for achieving high recognition accuracy with Dragon NaturallySpeaking:-

A proper computer hardware system and microphone that produces clear sound input

Although Nuance recommends an Intel Pentium 1 GhH Processor or higher and 512 MB of RAM, I would recommend at least a 1.6 Ghz processor or higher and 512 Mb RAM as minimum. Ideally 1 Gb RAM is preferred. This is because speech recognition is processor and memory hungry as it has to do many calculations to process speech. Approved operating systems are Windows 2000 with SP4, Windows XP SP1 or SP2, XP Home. It is very important to have a good sound card. Acceptable sound cards include the Creative Labs® Sound Blaster® 16 or equivalent sound card. For a full list of Nuance approved sound cards, see:-

http://support.scansoft.com/compatibility/default.asp

You can obtain the above information for your computer as follows:-

1. “Start” (bottom left of the Windows desktop)
2. “All Programmes”
3. “Accessories”
4. “System Tools”
5. “System Information

A good quality noise – cancellling microphone is recommended for optimum accuracy. The microphone that comes bundled with Dragon NaturallySpeaking should be considered a starter microphone only. Investing in a high quality microphone will pay dividends in reduced time making corrections. 

Laptops and certain desktops may need a USB sound pod. The sound card may be of poor quality or it can be of good quality but it can be crowded by adjacent electronic circuitry. Both issues can mean noise is generated by the sound card to the detriment of good speech recognition. A USB pod is basically an external sound card. To determine if an external USB sound pod is needed, you can run a Windows Recorder test in order to check your sound quality. If you hear a lot of noise in the playback, the chances are you may have poor speech recognition. Instructions for doing the Windows sound recorder test:-

1. Plug in a good quality headset microphone into the 3.5 mm line in jack of your computer 
2. Go to Windows Sound Recorder – Start – Programs – Accessories - Entertainment – Sound Recorder.
3. On the File Menu click “Properties” and then click the  “ Convert Now button.
4. In the drop down list under “Name”, select CD Quality. Each time you make a new recording it will default back to the previous setting.
5. Click back to Sound Recorder.
6. Click the Edit Menu to bring up the “Audio Devices” dialogue box. Make sure that your sound card is selected as the Preferred Recording device in the drop down list. Check the Volume and or Advanced settings to adjust more or less audio input. Some microphones require that you place a check in the box (if you have one) for Microphone Gain or Boost. Only use this Gain setting if your recording level is very low.
7. If you wish, click on “volume” and adjust the microphone volume slider up or down for the sound level.
8. Click out to Sound Recorder.
9. Make a recording by clicking on the Red button. Dictate for 5 seconds. If you have an On/mute switch make sure it is in the “On” position. As you record, the flat green line should show bubbles as the sound comes through. Hit the rectangular black button to stop recording, then hit the playback arrow to play back the sound.
10. Make sure your speakers are at maximum and listen for any pronounced humming, hissing, crackles or “electronic” noise in general.
11. If the playback is “clean”, the chances are that your sound card will be fine for achieving good accuracy with Dragon NaturallySpeaking.

Correct and consistent microphone placement

Correct and consistent microphone placement is essential for good speech recognition. Whilst this is particularly true for a hand-held microphone, it also applies to a headset microphone where you would expect the microphone elements position relative to the mouth to be constant. for a headset microphone, the microphone element should typically be one thumb’s gap away from the mouth and slightly below the mouth to avoid picking up breath sounds. Check the microphone position throughout the day and adjust as necessary.

Clear enunciation

Speak clearly but at your normal rate of speech. Pay attention to good enunciation of words and phrases. un the audio setup wizard again when conditions change. It is surprising how audio conditions can change throughout the day. We may speak different volumes throughout the day, we can get tired so that our voice sounds different, we may have a cold etc etc. if accuracy deteriorates, run the audio setup wizard again so that Dragon knows how you sound at that particular time.

Add words from documents and let Dragon adapt to your writing style

This is a very powerful feature in Dragon NaturallySpeaking. Basically, you can analyse a collection of your documents sometimes referred to as “concatenated” documents. The software will add unknown words to the dictionary and will analyse your “writing style” so that it can make more educated guesses of what you are saying next time.

The procedure to do this is as follows:-

1. Cut and paste many of your Microsoft Word documents into one large document
2. Click on “tools”, then “Accuracy Centre”
3. Click on “add words from your documents to the vocabulary”
4. When prompted to add a document, browse to the file in which you have your concatenated documents
5. Tick “let Dragon adapt to your writing style” style
6. Dragon will prompt you to ask if you wish to add the unknown words
7. Check the boxes for the unknown words that you wish to add

Hey presto! Dragon has added lots of new words for you and has “sensed” the way that you write and will accommodate you better (i.e. be more accurate) next time.

Add words and phrases to the vocabulary editor

I (and I know many others) believe that this is one of the most powerful ways to achieve high accuracy. very often Dragon will struggle with a word or a phrase and will have stubbornly misrecognised it every time. if you add the phrase to the Vocabulary Editor (in Accuracy Center), you will typically find that Dragon will usually get the phrase right most times. Some Dragon users have literally thousands of extra customised phrases in their vocabulary and this pays big dividends in terms of accuracy.

Correct mistakes

This is a fairly basic one! Correcting mistakes as you use Dragon greatly increases accuracy. The best way to make corrections is to use the command “correct (phrase)” rather than saying “select (phrase)” and then “correct that”. Doing this places the cursor where it was before you issue the command “correct (phrase)”. More often than not, just correcting the phrase (by spelling with your voice) without voice training the misrecognised phrase is best. This is because correcting misrecognitions by voice forces us to speak the misrecognised phrase in isolation rather than the context of the full sentence and this can actually cause accuracy to deteriorate. Only voice train correct versus misrecognised phrases when the misrecognised phrases are quite “stubborn”.

The following advice on the subject of correcting mistakes is reprinted from a forum entry of Chuck Runquist, Former DNS SDK & Senior Technical Solutions PM for Dragon NaturallySpeaking:-

Quote…….”Training individual words, even in the Vocabulary Editor, has no impact on accuracy because all the words contained in the active vocabulary (160,000 words loaded at runtime) and/or the Background vocabulary (what is stored in C:\Documents and Settings\All Users\Application Data\ScanSoft\NaturallySpeaking8\Data\Enx\enx_aus_general_large) have a fixed probability that does not change regardless of training. Changing the probability coeficeints of these words would result in accuracy degredation on a scale that would render DNS useless. Suffice it to say that this is necessary for the proper recognition of words. Remember that the training script makes reference to “…words by themselves AND in the context of other words.” What this means is that there is a distinction between the handling of individual words that can be modified only by training words in context.

Training words in context alters the context probability coeficients because this is where DNS distinguishes individual words by the context in which they are used, and this is what the AND means in the script phrase above. So if I just say “an” and get “and” it is because in the vocabulary and in common speech “and” usually occurs before “an”. Training these two words individualy will not change this probability. But, traing the phrase “He had an awful cold.” will increase the probability that “an” will be correctly used in this context. Therefore, this is why analyzing documents increases the accuracy of words used in context based on how the user writes/dictates.

It is important to remember that how we say words and phrases during dictation is unique and distinctly different from how we say them when we train them. For example, when you dictate the phrase “Training individual words, even in the Vocabulary Editor, has no IMPACT on accuracy…” and it is transcribed as “Training individual words, even in the Vocabulary Editor, has no IMPORT on accuracy…”, and you say “Correct has no IMPORT on accuracy”, we all pronounce that phrase differently when taken out of the context of the entire sentence. In addition, we virtually never say the same thing in the same way twice anyway. Therefore, when oorrecting this phrase to change “IMPORT” to “IMPACT”, if the playback in the Spell dialog is clear and correct, retraining the phrase alters the acoustic equivalent and, done too often, can result in eventual corruption of the words in that context. Multiply this by 100′s of corrections per day and it is easy to see why some user files eventually become significantly corrupted.

On the other hand, if there is no playback, the user has to make a choice. Did they say it correctly and clearly or not. If in doubt, then train it. No training results in the continued link between the original spoken utterance and the correction. If this is, or was, incorrect, then the correction will be associated with an incorrect acoustical utterance. Training in this context does not hurt because it will only show the one phrase, not the corrected and incorrect words or phrases, and training will create a correct, even if slightly different out of context pronunciation, acoustical/word or phrase link. In addition, no training will result in a correction that will not be saved for later use by the Acoustic Optimizer. Only words/phrases trained during correction or during additional training are recorded for update by the Acoustic Optimizer. Remember that the key word is Acoustic, not Language relative to the Acoustic Optimizer.”