Glossary
WER (word error rate)
The standard measure of transcript accuracy, counting substitutions, insertions, and deletions against a reference transcript. Lower is better.
WER is substitutions plus insertions plus deletions, divided by the number of words in the reference — it can exceed 100 percent when insertions pile up. Strong systems score in the low single digits on clean read English, while conversational, accented, and noisy speech runs much higher, which is where new training data helps. Scores only compare when text normalization — numbers, casing, punctuation — matches.
Why it matters
WER measured on a held-out sample is how transcript accuracy claims in a dataset spec get verified.