Is the Future Speech Recognition or Human?

Technology is great and can save a lot of time, but there is one important fact that it is vital never to forget, it is not human. When Seneca said, (4 BCE), ‘to err is human’, he had no idea speech recognition would be trying to take over the world and automated transcriptions would make humans look infallible.

As long as you are using speech recognition to record your own notes, ideas, reminders or a shopping list it could be a great solution. If you want an accurate transcription for business, this is not the way to go.

Here is a frightening point; since speech recognition software was used in the healthcare sector, errors in prognosis and diagnosis have risen. Now this could impact on someone’s life very badly.

Your business may not deal in life and death, but a simple error could result in a disaster of another type for your company. (Note; speech technology just transcribed that as ‘Your nosiness may not deal…’ Granted I changed my mind about the word I intended to say, from company to business, but neither resemble the word the technology chose, and I have used the word ‘business’ before and had it transcribed correctly.)

Technology is only as good as the human using it. In the case of speech technology this means you will need to thoroughly check every utterance transcribed, every time. This makes it neither time nor money saving compared to a human transcription.

Companies making speech recognition software admit their products are 80 – 88% accurate, (you may find it is not even as high as this). Human transcription done by a reputable company, such as Singapore Transcription is 99-100 % accurate. This 12-20% discrepancy may not sound too bad, but it depends what is inaccurate. Take the error mentioned above. Had you sent an email to a company to say, ‘Your business has terms we find incompatible with our company. Can we discuss reaching a compromise?’ You might be able to negotiate something advantageous. If you had sent, ‘Your nosiness has terms we find incompatible with our company’ it is doubtful they would have kept reading and communication with them would be over. Rightly so, as this is offensive.

There are many other reasons why humans are better; they distinguish background noise as just that; they have no difficulties changing abruptly from one accent to another; and there are many more reasons why human transcription is better.

As long as an automated transcription needs human intervention, it will not be worthwhile using machines to do this job. You are simply paying twice for the same work.