Thursday, September 2, 2010

Controversy re. Unselecting Automatically add words to the vocabulary Option in Dragon NaturallySpeaking

Note: This article applies to version 10.1 of Dragon NaturallySpeaking.
Most of you know that in his Optimizing NaturallySpeaking for Accuracy and Reliability document, Alan Cantor recommends unchecking the Automatically add words to the vocabulary option. This is found in the Correction tab of the Options dialog box, accessible from the Tools menu on the DragonBar. I have been following Alan's recommendation because I know that he's a longtime trainer of Dragon NaturallySpeaking (DNS) users in his Toronto accessibility consulting practice, Cantor Access Inc. But while doing some research on the way DNS handles vocabulary, I read the DNS Help Topic on this option and have changed my mind about Alan's recommendation. Here's my explanation why.
The DNS documentation describes the Automatically add words to the vocabulary option as: "This check box controls whether a word from the backup dictionary is added to the active vocabulary. Select this box to automatically add corrected words to the active vocabulary. If you correct a word you dictated—either by typing over the word, by spelling it (selecting it and dictating "spell that," then spelling it) or by selecting it in the Correction menu—Dragon adds it to the active vocabulary when the correct word is not present in the active vocabulary but exists in the backup dictionary. After Dragon adds that word to the active vocabulary, the word appears with a green star next to it in the All Words view in the Vocabulary Editor."
Active vocabulary was new to me, so I found a definition: "The active vocabulary contains the words that Dragon is most likely to recognize on the first try, without requiring you to do anything extra such as correcting the words with the Correction menu. Active vocabulary words are stored in computer memory and thus always available. The number of words in the active vocabulary always remains the same. When new words are added, words that have not been used recently are removed from the active vocabulary, but remain on the computer's disk in the backup dictionary. Note:You can view the words present in the active vocabulary in the Vocabulary Editor dialog box, which you can open by clicking or saying "View or edit your vocabulary" from the Accuracy Center."
This lead me to seek out the definition for backup dictionary: "The backup dictionary stores all vocabulary words. Dragon moves frequently used words from the backup dictionary into the active vocabulary. The backup dictionary is stored on the computer's disk. If Dragon misrecognizes a word, the word may be in the vocabulary but stored in the backup dictionary. In this case, one way to move the word into the active vocabulary is to use the Correction menu to correct the error. You must save user files for this change to take place permanently."
Based upon the above, it seems like a good idea to leave the Automatically add words to the vocabulary option at its default, which is checked. Otherwise, the next time you dictate a word that was misrecognized because it was only in the backup dictionary, it won't be in the active dictionary, so DNS won't recognize it and you'll have to correct it again.
Further along in Optimizing NaturallySpeaking for Accuracy and Reliability, Alan says that "The best way to add words and phrases is the Vocabulary Editor." I agree that custom words and phrases that aren't found in DNS's backup dictionary should be added in this way. If you look at the list of filters in the Vocabulary Editor, you'll see that there is one for Custom words only and another for Words from Backup Dictionary. So these are two different concepts. I wondered whether Alan was getting them mixed up. I emailed him and asked. His reply was: "I agree that automatically adding words to the vocabulary would be a useful feature, but my experience is that it does not work as advertised. When this feature is enabled, nonsense words and word fragments are automatically added to the vocabulary. You can see these words in the Vocabulary Editor as Custom words. Over the years, I have deleted hundreds, if not thousands of this nonsense words. I don't think that they contribute anything toward accuracy. The feature appears to be almost repaired in version 10.1, but I am still noticing odd words in the vocabulary when I have it enabled."
So, being a longtime user of DNS, Alan rightly has some practices he's been following to work around some misfeatures in the software. After enabling this option again, I'm noticing words starting to appear in my Words from Backup Dictionary list. I've also noticed a few in the Custom words only list that I did not add manually (I'll explain how to do that in a future blog post). Why are they here? Well, the DNS documentation says "Added words are considered custom words, unless they are in the backup dictionary." So, DNS didn't recognize the word I said (couldn't find it in the active dictionary), so I selected it and said "Correct that"; in the Spell dialog box I spelled it out but made a typo (entertaintng) so now it is in the active dictionary as a custom word. Another custom word that I have is "Setup". If I search for it in Vocabulary Editor (use the All words filter to see what's in the active vocabulary), I see 3 variations: setup, Setup, SetUp. So if I had corrected the word to SetUp, there wouldn't be a custom word present (Setup). The moral of the story is: with DNS automatically adding words to the active vocabulary, the user must be careful when performing corrections: no typos and pay attention to case (i.e. capitalization).
Further in his email, Alan says: "I would use the automatically add words to the vocabulary feature if it worked better than it does. On the other hand, some of my clients enable this feature, and get decent performance. I don't think that it is a deal breaker. (I suggest that they review their custom words occasionally, and weed out the nonsense words.) My preference is to use the vocabulary editor; but I also have the perspective that there are more efficient ways to increase accuracy than playing with individual words. For example, dictating long phrases makes a huge difference, as does correcting misrecognitions in context. Neither is easy, but both are effective. Ditto for running the vocabulary and language model optimizers once in a while."
I hope that I've given you enough information to decide whether you want to enable or disable the Automatically add words to the vocabulary option. I plan to leave it enabled, and be careful about the corrections that I make. If I have doubts about the spelling of a word, I'll confirm the spelling before making the correction. Also, to deal with the case/capitalization issue, I will make my corrections in lower case and then say "Cap That" once the Spell dialog box closes. Finally, I plan to review the Custom words only list in the Vocabulary Editor (View/Edit menu item in Words menu of the DragonBar) and delete any erroneous ones.
There's one other comment that Alan makes that I want to discuss: "My sense of the situation -- I know that it is a controversial one -- is that many people add fewer words than they think. Years ago, I created a list of the words, phrases, acronyms, that I actually use, and import it every time I create a new user. This list has not grown substantially over the years. I am guessing that I add 20 words per year; but I also review the file periodically, and remove words/phrases/acronyms that I no longer need. When I am working with a client, I will often help them generate a word list like mine, and show them how to import it."
Keep in mind that Alan is an accessibility consultant and he's been in the field for over 15 years. While the field is changing, he's probably not coming across new vocabulary the way some of you might be e.g. university students on the cutting edge of their discipline and taking courses from a variety of disciplines that have specialized vocabularies. In a future blog post, I will explain how you can manually maintain DNS's active vocabulary to improve DNS's recognition ability for your dictation. But the point I want to make about the Automatically add words to the vocabulary option is that you might find it less disruptive to the flow of your work if you enable this option and add your custom words as part of the correction process versus explicitly opening the Vocabulary Editor and adding the word there.

No comments:

Post a Comment