If you’ve read any of our previous blogs, especially those about AI, you’ll know that accents and transcription can be the biggest thorn in the side of accuracy. YouTube for Kids has had a number of issues where children’s content experienced what’s known as inappropriate content hallucination where the meaning of a word is incorrectly transcribed. We even spoke about it in our last blog which highlighted this issue for a TikTok user who found her videos kept being removed because the platform’s algorithm didn’t understand her accent!
We’re all different…
Indeed – and not all of us find understanding an accent easy to do. Some people find deciphering an accent much easier than others.
The major factor at play when it comes to understanding an accent is speech clarity. This is made up of a combination of elements:
· the speed of the speaker
· whether they are male or female
· the pitch of their voice
· whether there is any background noise
· how far away you are from the microphone
· the speaker’s accent
The ability to be able to properly decipher what someone says is simply down to the level of exposure of all the elements of speech clarity. The greater the exposure, the increased chance of understanding what’s been said.
Taking the mental strain
Another factor is the cognitive effort it takes to decipher the accent.
When we listen to someone speak as we do, the mental effort it takes to process it isn’t that taxing on the brain. But the opposite is true when they speak in a way we’re not used to hearing, we need to make more of a mental effort to understand them. When making that effort isn’t something you’re familiar with, maintaining it becomes increasingly difficult.
With AI, accents and transcription are even more tricky
All of the above is fair enough, but throw in AI and transcription, and the situation becomes even more problematic! We found this article gave an excellent example of a real life case of this: smart speakers – such as Google Assistant and Alexa – that have dramatically increased in popularity in homes around the world. But for people with accents — even down to regional lilts, dialects and drawls — the artificially intelligent speakers can seem very different: inattentive, unresponsive, even isolating.
These speakers are taught to understand accents and the same principle of listening with the human ear applies: the more accents they are exposed to, the more likely they are to learn to understand them. But as the article highlights, despite the best efforts by the developers, the technology simply hasn’t cracked it yet.
Human transcription is the way to go for accuracy
It really is!
Our team of highly experienced transcribers are skilled in handling recordings in different formats, taken from meetings of all sizes and structures, and understanding the particular terminology of your business or sector. Our editors ensure the finished product is accurate and its meaning clear; the final copy is a precise record of your meeting, interview or conference – whatever the accent!
For a professional transcription service that delivers an accurate, quality record of your meeting or event, please contact us 01737 852 225 or email alex@fionashipley.com.