Voice recognition technology has entered fairly mature stage in its development and implementation cycle. From voice prompts of 800 numbers to professional versions of dictation software, current voice recognition gadgets can replace human operators and transcriptionists in a variety of settings. Of course, some of us chose to adopt it and had success with it, while others may have been introduced to this innovation in a more difficult manner. If you are among those for whom adoption of this technology was not necessarily an easy task, please read on.
This article was written by a radiologist, but the techniques introduced here can easily be applied to other fields. The key to success with voice recognition is flexibility on the operator’s part. The chance of success is not great if you try to use voice recognition as a 100% substitute for human transcriptionist. But, as a fancy keyboard or “writing” device, it works just fine.
The confusion probably comes from salespeople claims that their software is close to being 100% accurate. Guess what, even if it is 99% accurate, all it takes is one mis-recognized word to render a sentence non comprehensible. In addition, there are phone rings, interruptions, and other extraneous sources of noise that result in 85 to 95 percent accuracy being more realistic. The accent, by the way, contributes very little to inaccuracies. The tone and volume of voice seem to be more important in practice.
One approach to success is to use the actual voice recognition feature as little as possible. To do so, you will have to utilize macros, learn how to setup your documents to make editing easier and learn some tricks that are already built into word processing software and operating systems.
First, invest some time into thinking on what sentences and phrases you use the most. There is a good chance that you are saying the same things over and over again if your work is similar between different reports and letters.
Your macros should be broad enough to be useful in a variety of settings. Can you change your common phrases slightly so that one macro can be used in different settings? For example, “The osseous structures are unremarkable” can be used in radiograph, ct, and mri reports and for various body parts. This is very different from “Tibia and fibula radiographs demonstrate no evidence of fracture” that can only be used on tibia and fibula radiographs.
Put some thought into the sentences that you are putting into your macro. For example, “no” in “Acromioclavicular joint demonstrates no degenerative changes” can be easily changed to “mild” without having to say the entire sentence. When you are describing findings in structures that are close to each other, consider changing verbs slightly so that there is a good chance that final report will read easier. “Mucosal thickening in the right frontal sinus is present” reads well together with “Mucosal thickening in the left maxillary sinus is noted” as opposed to “present” in both cases. Something should also be said in favor of skipping complete sentences all together. “No evidence of acute intracranial hemorrhage” can be used in the findings and impression sections of the report, while “There is no evidence of acute intracranial hemorrhage” does not read well in impression section that usually contains bullet points rather than complete sentences.
Once you start accumulating large number of macros, you will need to find a way to name them systematically. For example, you can make a rule that body part or organ goes first and pathology second. For example, “prostate large” can map to “The prostate is prominent.” You will also need to decide whether you want qualifiers such as “mild” “moderate” and “severe”, and “right” “left” and “bilateral” at the end, or at the beginning of a macro name. For example, macro name “moderate pleural effusion right” will be easy to remember if you are consistent with your naming scheme. Some people advocate redundancy in the macro names, such that “prostate large” and “large prostate” both map to “The prostate is prominent.”
Last point about macros. If you are using macros a lot, find out if your software can use a customized word for initiating macros. For example “autotext” and “dictaphone” are much longer words compared to “paste” or “click” or “press.”
There are couple of tricks that you can use in any word processing software and your operating system that may not be a common knowledge. Double click selects a word in most. Triple click in some software results in the selection of entire paragraph. Both of these techniques are easier than voice commands such as “select X” and then “cut that” or “delete that” or “scratch that.” If the triple click is available in your system, consider setting up your macros with each sentence on a separate line. This way triple click will result in selection of that entire sentence.
If possible, consider keeping your reports short. The less you say, the less you will have to correct later. Your state of mind matters. The tone of voice changes when you are frustrated or tired. One day, voice recognition software may be able to compensate for this. Until then, the user will have to be flexible and work around the imperfections.