"Why can't AI do your job?"

When I tell someone about audio transcription, invariably the first question I get asked is, "Why can't you just have voice recognition software do it?" Well, unfortunately your computer just isn't that smart (yet). Here are a few reasons why having audio transcription done by humans is so important for accuracy and quality.

1. Background noise. Despite your best efforts, sometimes the quality of audio recordings isn't great. If you do an interview in a public place, there will often be some degree of background noise and interference. Often, this makes it a bit more difficult to understand what people are saying, and you really do need a human ear to do this.

2. Accents, slang, mispronunciation. There is a reason why you typically have to 'train' voice recognition software - there are so many ways of pronouncing the same word. A human ear can tell whether it's an Englishman saying "beg" or a Glaswegian saying "big", but a computer may not be as clever. 

3. Speaker identification. The biggest problem with voice recognition software is that it cannot tell the difference between speakers, so it's fine for single speaker, direct, clear dictations, but any interviews or focus groups need to be done by a transcriber who can tell who's who.

4. Readability. When we do audio transcription, for most research we do 'intelligent verbatim', meaning that we transcribe exactly what is said, but take out all 'ums', 'ahs', stutters, and annoying repeated words like "kinda" and "sort of". It makes it much easier to read and analyse for researchers while still maintaining 100% of the meaning.

What makes a good transcriber (it's more than fast typing)