Sometimes when I am on site with a client working with their speech recognition software, they ask “Is there any way to make this
thing any more accurate?” or ” The system just doesn’t seem to understand me as often as it should”. So today I want to go over a few different ways to ensure you are getting the most accurate recognition while using our favorite Speech Recognition software Dragon Medical Practice Edition 2. Note that this applies to most any version or edition of Dragon Naturally Speaking as well.
Before we start, if you’re more of a visual person, you can view our YouTube video on Accuracy Optimization for Dragon as well.
1. Check Microphone
The first thing on our list is to check the microphone. This can often be the reason for poor accuracy. I would highly recommend you make sure the device is up to date. Here at VTEX we always recommend to keep up with firmware updates. You can find these firmware updates for any brands and models we carry on our knowledge centre. These updates have made big differences in accuracy for some of our clients.
No matter which type of device you are using you want to ensure that you are using it at a consistent distance from your mouth. This will ensure the software has a steady vocal reference to transcribe from.
2. Speed vs Accuracy
There is an option in Dragon that allows the user to define how long system analyzes your words and phrases, in order to find a word match. If you go to Tools –> Options –> Miscellaneous there you will find the Speed vs Accuracy slider. The further the slider is to the left, the faster Dragon will analyze your speech and produce a match. This can be beneficial if you have a system the is slow or just meets the minimum requirements to run dragon. The further you move the slider to the right, Dragon will take more time to analyze your speech to find a match. This can improve your accuracy especially on a faster system. This feature is intended to allow each user to find the perfect balance between speed and accuracy.
3. Accuracy Tuning
The final piece to the puzzle is in Tools –> Administrative Settings –> Scheduled Tasks –> Accuracy Tuning. Here you can configure both the Acoustic Data and Language Model analysis.
What is Acoustic Data?
When you train words, make corrections, and perform additional training you are creating new acoustic information or data that is added to your User Profile. This acoustic data is the sound of your voice as you pronounce each word.
What is the Language Model?
A Language Model is a statistical record that predicts which words are most likely to occur in the context of your sentence. As you speak Dragon is using the Language Model to anticipate what words you will say next depending on the context of your current spoken words.
In the Accuracy Tuning configuration we recommend you run both on a monthly basis as he process requires a fair amount of content to effectively improve accuracy. This process can be time consuming and CPU intensive. We also recommend that you have it run overnight while the computer is not in use.
Well that wraps it up for today. Just follow these steps to ensure Dragon is the most accurate as possible. Have fun and happy dictating!
Post By: +Paris Vaughn