Sometimes called neural machine translation (NMT) engines, AI-generated translation systems such as DeepL or Google Translate have made amazing progress in the last 4-5 years.
However we shouldn’t lose sight of the fact that even the best NMT systems are based on algorithms and algorithmic learning. So they have the same vulnerabilities as all AI systems. Simply put, their understanding of wider cultural context and linguistic nuance is limited.
As a result, NMT is usually not completely free of sense errors, omissions, missed nuances and inconsistencies, particularly when it comes to more complex texts.
The good news is that by introducing post-editing as a second step in the translation process, these disadvantages can be largely compensated. Here, the AI-generated translations are checked before publication by a qualified human proof-reader or “post-editor”, specialised in editing and correcting NMT translation outputs.
We offer three service levels of post-editing:
| MT Post-editing LITE (2-eyes) | MT Post-editing FULL (2-eyes) | MT Post-editing FULL plus revision (4-eyes) |
|---|---|---|
| Fastest workrate, lowest pricing | Medium workrate, medium pricing | Medium workrate + independent proofreading (revision) |
| Suitable for non-core internal comms / intranet social pages / social media translation | Suitable for non-core internal / external comms / intranet business pages / plainly written website pages. | Recommended for more core content NOTE: for business-critical content we would always recommend 100% human translation, plus revision where appropriate |
