Post-editing (human translation)

Workflow for human correction of machine translation

---

Human post-editing is the review and correction of machine translation output. With the rise in machine translation quality, human translation often refers to a post-editing workflow, rather than to pure human translation from scratch.

When post-editing, the human translator checks the original input and the machine translation output, and fixes errors in the output.

Post-editing started to replace pure human translation in the 2000s because it is faster and cheaper, and became the standard for many content types in the 2010s.

Flavours

With light post-editing, the goal is that readers understand the idea of a text. Light post-editing provides the gist of a message. It is usually aimed at internal communications.

With full post-editing, the goal is to match human quality. Full post-editing is suitable for publishable information.

Metrics

Post-editing distance is a measure of how much the human translator post-edited the machine translation.

Post-editing effort is an estimate of how much time the human translator spends post-editing a machine translation output.

Post-editing metrics are used both for a specific translation and in aggregate - at the document, project or workflow level. They can be length-normalized and given as a percentage or score.