Can Voice Recognition Technologies Make Transcription Services
Redundant?
Many businesses need to convert recorded voice to text and have
long been looking for ways to do it quickly and inexpensively.
Transcribing medical dictation is a prime example.
Some years ago, when voice recognition software became
commercially available, most people expected that the solution
had finally arrived. Businesses looked forward to cutting down
on transcription costs and everyone who hated typing looked
forward to getting rid of their keyboard.
Unfortunately, the reality turned out to be rather different.
Voice-to-text technology has been a big let down so far.
The fact is, voice recognition software is easily thrown off
track by many different factors. If you don't speak clearly and
distinctly, it may not give you the right output. If you try
using it in a noisy place, it will fail more often than not. If
you have an accent, it may not understand you. Even if you have
a bad cold, you'll find that the software may give incorrect
results!
In other words, voice recognition software works reasonably well
under ideal, laboratory conditions, but not in a typical home or
business setting!
Healthcare professionals who attempted to use voice recognition
technologies to eliminate transcription services found that they
need to "train" the software to function well. That takes a long
time and a lot of work. Most wound up continuing to outsource
their medical transcription work.
Of course, there are many other types of situations where
transcription is needed. Examples include recordings of
seminars, teleconferences, interviews and classes that need to
be converted to text.
In natural speech, people tend to use lots of "aahs" and "umms"
as well as unnecessary phrases like "you know". Current voice
recognition technology is just not capable of filtering out such
irrelevant sounds or words.
In addition, people also string together several sentences using
"ands". The software can't break up such speech into meaningful
sentences. Nor can it break up speech into meaningful paragraph
units the way a transcriptionist can.
And if the recording is filled with background noise, or if more
than one person is talking at the same time, the software will
not function reliably and consistently.
Maybe sometime in the future someone will invent voice
recognition technology that can handle all the above issues.
Till then businesses will need to use transcription services,
particularly for work like medical transcription, where accuracy
is critical.