Voice Recognition

There are some individuals and organizations that see voice recognition software as one solution to their transcription needs. For some applications and needs, voice recognition software may be a reasonable choice. For many others, however, it is inadequate. The more demanding the expectations for accurate and clear transcriptions, the more voice recognition may not be the solution you seek.

The benefits of outsourcing your transcription requirements are well known:

• It saves money by decreasing personnel and overhead costs
• It provides increased capability by providing superior practices, job skills, and technology
• It lets you focus on your core business or practice
• It frees office and floor space
• It offers accuracy, speed and reliability

Can these benefits be replaced or exceeded by voice recognition software?

While the benefits of outsourcing transcription services may be well known, the problems in implementing a voice recognition solution that can meet the same standards may not be well known.

Voice recognition works by having the computer recognize what a user is saying. The computer then reacts accordingly. The software identifies the sound pattern that the user produces; it then associates those patterns with patterns in its database. When the association is made, the software returns a response and types that response for the user.

The demand for voice recognition grew as personal computers became commonplace and more powerful. Windows XP introduced speech recognition in 2001.

Users can find both free standing voice recognition software and as part of an operating system pre-installed on computers.

The software attempts to learn the speech patterns of the user. So, if more than one user is providing input, the problem of learning speech patterns increases. The software must learn those speech patterns well to avoid mistakes in transcription and command execution.

Problems begin with the pre-programmed database of sounds. A user’s speech will differ from those patterns to some degree. The quality of the input microphone may be a problem. Ambient noise is a problem.

There is an obvious learning curve as the software learns one or more patterns of speech from users. Initially there will be many mistakes in the translation from input to text on paper.

Voice recognition software will not filter out the “ums” and “aahs” of common dictation. It cannot separate sentences at times or organize sentences into paragraphs. Conference settings are problem for voice recognition since so many different speakers make take the microphone and speak.

Compared to professional transcription services, voice recognition cannot provide the level of accuracy demanded by professionals. It often cannot deal with the specific vocabulary of a profession or the more narrow vocabulary of specialists within a profession. For best results, voice recognition still requires careful proof reading, an editor and review by someone familiar with the user’s profession.

The bottom line is that the filtering process is too detailed and content is too critical to accept the level of errors common to voice recognition.

If you are seeking the highest quality in professional transcription services in Canada and throughout North America, contact SpectraScribe at info@spectrascribe.com