Speech Recognition: Getting Started

Although once the realm of science fiction, computerized speech recognition has truly entered the mainstream.  Although most often experienced in automated call centers, individuals are now taking advantage of this technology to create reports, enter data into spreadsheets, and for virtually any process in which text is entered into a word processing documents.  Are you considering this technology?  Or are you seeking to improve your accuracy or functionality with your existing speech recognition product?  Guiding you through this process and seeking input from those who have successfully implemented speech recognition into their daily lives is what this site is about.

Let me begin with a few comments about myself.  I am not a graduate of a computer science program.  I have learned virtually everything I know about computers, speech recognition software, and microphones by means of about 10 years of dabbling, trial and error, reading, and probably thousands of hours which might more reasonably have been spent with my wife and kids!  No doubt you can relate to the lure of a new technology and the passion it can generate.  My “day job” is employment as a cardiologist at Dartmouth-Hitchcock Medical Center in Lebanon, NH.  About 90% of my day involves seeing patients and documenting the history, physical exam, assessment and plan for each patients.  As you might know, accurate and thorough documentation is essential for good communication … and perhaps also for avoiding lawsuits.  It is this daily grind of medical documentation that has become the “playground” for my interest in speech recognition software.  While many of my colleagues have been pushed to hand-typing, about 6-7 years ago I switched to 100% use of speech recognition software.  In the process, I have saved my medical center countless thousands in transcription costs and have saved my wrists from the ravages of repetitive use injury.  More importantly, using speech recognition software has allowed me to provide timely and thorough documentation which is available for others to review even before my patients have reached the parking lot.  It is truly an amazing technology!  It was my love of the technology and desire to provide a source for microphones, software and helpful information that led to my decision to start “Speech Recognition Solutions” in the fall of 2006.

What do you need to get started?  How can you improve your accuracy and efficiency if you are already using speech recognition software?  What follows is intended to be useful starting point for you.

Key Point One:  Buy the right version:  Here’s  the quick and dirty of what version to buy and where to get it.  Keep in mind that each version uses the same speech engine and the differences pertain to added functionality and added vocabularies for special uses.

  • Standard:  This version is too pared down to be worthwhile for anyone.  We would suggest that even the mildly curious stay away from this and, rather, choose the “Preferred” Version described below.
  • Preferred:  This is the basic version and includes the functionality to create  commands and macros.  For the beginner, for the person simply wanted to dictate emails, general reports, a novel, etc., this is the perfect version.  You can price search this product and typically get it for less than $100.  It comes with an inexpensive microphone which you might eventually want to upgrade, but it’s fine to start out.  Where to buy?  We suggest you do an online price search.  Amazon.com often offers this version for less than $100, which is a great price.
  • Professional:  This is the version for the more serious speech recognition user.  It does not include specialized legal or medical vocabularies, but does have the ability to create advanced commands.  This version is A LOT more expensive than the Preferred version and unless you truly need the  ability to create advanced scripting commands (most folks don’t use these), I would suggest you stick with the much more affordable Preferred version.  Where to buy the Professional version?  We sell it.  You might also do a price search and see if you can find a bargain.
  • Legal:  This version is basically the professional version with an added vocabulary of 30,000 legal terms and the ability to automatically format legal citations.  We strongly suggest this version to attorneys.
  • Medical:  This version is tailor made or physicians and other medical providers and includes over 50 medical specialty vocabularies, lots of medical formatting options, the ability to navigate throughout documents using something called “dragon commands” .  It is the only version of Dragon which fully integrates with the very popular Dictaphone PowerMic II, which is a hand-held microphone with programmable buttons.  Although you can buy less expensive versions of Dragon and add a third-party medical vocabulary (saving money in the process), it doesn’t truly emulate the functionality of the full medical version.  We should also point out that Nuance took the somewhat controversial step of disabling the microphone in the non-medical versions of Dragon when operating in common electronic health records, so be aware that you might run into problems if you have a commercial EHR and don’t use the medical version.

Key Point Two:  Learn all you can about the software.  Although the basics of using Dragon are quite simple (turn on the microphone and talk), there are many nuances about the program that will allow you to improve your accuracy and optimize your productivity if you take a look “under the hood” and learn some keys things, such as:

  • Microphone management (learn how to program a keyboard key to quickly turn the microphone off and on and avoid  extraneous noises from contaminating your dictation
  • Correction:  Learn how to train Dragon when it makes mistakes so it’s unlikely to repeat the same mistake again
  • Vocabularies:  Learn how to add words to the basic Dragon vocabulary.  Learn how to manage vocabularies.  Understand the benefits of switching from one vocabulary type to another as you change your work type.

If you want some assistance in learning the software, we suggest purchasing a DVD training guide.  We offer two such programs, including the “CompuTalk SpeakEasy DVD Training Series” and the “Say I Can” DNS 10 Video Guide”.  Either will serve as a great jump start in your knowledge base.

Key Point Three:  Learn how to optimize accuracy.  Although Dragon is amazingly accurate out of the box and after the minimum 5-6 minutes of training, there are many things you can do to optimize accuracy.  Among these include:

  • Enunciate carefully:  This is clearly the most important determinant of accuracy.  It is not necessary to speak slowly, but rather just distinctly.
  • Speak in a smooth, flowing manner:  Since Dragon makes it’s decision on any given words on the basis of words occurring before and after, it is definitely helpful to give the program lots of  words to evaluate.  Speaking too slowly can actually degrade accuracy..
  • Use the best possible hardware:  Review the “minimum” and “recommended” standards posted by Nunace and  to the extent possible, get a PC at or greater than the “recommended” standards.  A faster processor and more RAM won’t necessarily increase your  accuracy, but it will dramatically increase the speed of operation and improve the overall performance of Dragon.
  • Consider purchasing a higher quality microphone:  the microphone that comes with Dragon is certainly OK, but higher quality microphones will provide better accuracy and better external noise rejection.  Our “quick and dirty” recommendations are as follows:

For more detailed information on the variety of available microphones, please view our Microphone Selection Guide.

For more information about speech recognition software, you might visit some of the following site:

Finally, if you are interested in having additions to this blog automatically sent to you, please subscribe to our RSS feed.  Do this by clicking on the  link entitled “RSS Entries” in the menu on the right and then click on “Subscribe” on the resulting page.

Jon W. Wahrenberger, MD
December 29, 2009

Leave a Reply