Word error rate
The standard score for transcription accuracy: the share of words a speech recognition system gets wrong.
Word error rate, almost always shortened to WER, measures how well a system turns speech into text. It counts the words that were wrong, missing, or invented, and divides by the number of words actually spoken. Lower is better, so a WER of 3 percent means roughly three words in every hundred came out wrong.
It is the number to look at when comparing transcription services, but it hides a lot. WER treats every mistake equally, so misspelling a name counts the same as dropping a “not”. And it is always measured on a specific test set, which may sound nothing like your recordings. Accents, background noise and technical vocabulary can push real-world error rates well above the advertised figure.