TER (Translation Edit Rate)

TER (Translation Edit Rate) is a widely used metric for evaluating Machine translation (MT) quality by measuring the amount of editing required to match a human reference translation. The calculation is straightforward: it counts the number of edits—insertions, deletions, substitutions and shifts—needed to align the machine output with a professional human translation.

A lower TER score indicates higher quality; for example, a score of 0 means the output is identical to the human reference. In the localization industry, TER is essential for benchmarking Neural machine translation (NMT) performance and estimating post-editing effort. A low TER suggests that linguists will spend less time correcting the output, leading to faster turnaround times and lower project costs. RWS uses TER within its Language Weaver platform to monitor engine performance and guide training strategies for custom MT models.