TER
An automatic machine translation evaluation metric that computes the ratio of the minimum number of edit operations required to transform a hypothesis into a reference, relative to the reference length; lower values are better and reflect the post-editing cost of the translation.
Timeline (last 90 days)
The study used conventional reference-based metrics such as BLEU, chrF and TER.
On the selected models and languages, ReciLex augmentation lowered TER from 77.14 to 75.12.
TER computes the ratio of the minimum number of edit operations (insertions, deletions, substitutions, shifts) needed to edit the translation into the reference translation to the reference length, where lower is better.
All Facts (3)
The study used conventional reference-based metrics such as BLEU, chrF and TER.
50%UnverifiedOn the selected models and languages, ReciLex augmentation lowered TER from 77.14 to 75.12.
50%UnverifiedTER computes the ratio of the minimum number of edit operations (insertions, deletions, substitutions, shifts) needed to edit the translation into the reference translation to the reference length, where lower is better.
50%