With technology advancing in leaps and bounds, we are often asked to address the myth that traditional transcription is dying, with voice recognition software primed to become the dominant technology.
When assessing this myth, it’s important to consider two significant factors. First, there is the technology itself, and how far it has actually progressed. In the case of voice recognition software, there is no doubt that substantial advancements have been made in recent years and that this type of software is, if properly trained for a specific single-speaker, capable of delivering reasonable accuracy (as long as you don’t have a cold).
The second factor surrounds the nature of the recordings that are submitted for transcription. Clear, single-speaker files are in the minority, with most files containing multiple speakers, ad hoc interjections and/or background noise. Frankly, the typical audio file received for transcription is, in actuality, kryptonite to even the best voice recognition technology available.
With so much development required before these issues can be addressed, it’s unlikely that human transcriptionists can be replaced by technology at any time in the foreseeable future. The most likely scenario is that a limited subset of users will invest the time and effort to train dictation software so that it is reliable for their own spoken word, but that transcription services will remain in demand for competent, accurate transcription of multi-speaker audio, recordings of guest speakers, and audio where interjections or background noise render voice recognition software ineffective.
We strongly believe that this myth is, for the time being, still busted.