AI & Automation ยท Workflow
MT & Post-Editing
Machine translation (MT) produces a first-pass translation automatically; post-editing (MTPE) is the human work of correcting that output to the required quality. The workflow only pays off when content, engine and error tolerance are chosen deliberately.
Also known as: machine translation, mtpe, post-editing, post editing
The MTPE pipeline
- Source contentScreened for MT suitability: repetitive, unambiguous, structured.
- MT engineRaw output generated, often with terminology constraints.
- Post-editingEditor corrects output to the agreed quality level.
- QAAutomated checks plus sampling of edited segments.
- Updated TMCorrected segments are stored so leverage improves over time.
Light vs full post-editing
Light post-editing fixes only what blocks understanding: mistranslations, wrong terms, broken meaning. Full post-editing brings the text to human-translation quality. The choice is a business decision per content type. Help center articles and legal terms should not get the same treatment.
When MT is the wrong tool
Highly creative copy, culturally sensitive content, and anything where a wrong sentence creates liability. MT output is cheap to generate and expensive to repair when it fails silently. A fluent but wrong translation is worse than no translation.
Measuring whether it works
Track edit distance (how much the editor changed), quality scores on sampled segments, and time per segment. If editors rewrite most of the output, MT is adding cost, not saving it. That is the honest signal to renegotiate the setup.