AI translation systems, whether based on neural machine translation engines such as DeepL & Google Translate, or on large language models such as ChatGPT, have made amazing progress in the last 2-3 years.
However, we shouldn’t lose sight of the fact that even the best AI-translation is based on algorithms and algorithmic learning. So, in common with all AI systems, they have certain vulnerabilities. These can be summed up as frequent inability to grasp cultural context and linguistic nuance.
So, in addition to more glaring translation errors which arise due to the AI misunderstanding source content, AI-translations are usually not completely free of sense errors, omissions, missed nuances, inconsistencies and cultural faux pas. Generally, the more complex, nuanced, business critical or culturally nuanced the content, the less suited AI translation tools will be to the task.
The good news is that a hybrid approach can offer the best of both worlds. By introducing human post-editing as a second step in the AI-translation process, the weaknesses of AI can be largely compensated for. With our hybrid approach, AI-assisted translation, the AI-translations are checked before publication by a qualified human proof-reader or “post-editor”, specialised in editing and correcting AI translation outputs.
We offer three service levels of post-editing for AI translation:
| AI-Translation Post Editing: Lite (2 eyes) | AI-Translation Post Editing: Full (2 eyes) | AI-Translation Post Editing: Full + (4 eyes) |
|---|---|---|
| Rapid workrate, lowest pricing. | Quick workrate, low pricing. | Medium workrate, medium pricing. |
| Suitable for non-core internal comms, such as intranet pages, staff emails and employee surveys. | Suitable for non-core external comms, such as low-priority website pages and social media posts. | Recommended for internal and external comms, such as company newsletters, customer service materials, employee handbooks. NOTE: for core, business-critical content we would always recommend human translation, plus revision where appropriate. |
| The source text is fed into an AI-translation tool. After which, it is quickly proofread by a qualified translator who will make any necessary corrections to meaning. | The source text is fed into an AI-translation tool. After which, it is post-edited by a qualified translator whose aim is to make the text sound more fluent, as if it had been written by a human. The post-editor will also make any necessary corrections to meaning. | The source text is fed into an AI-translation tool. After which, it is post-edited by a qualified translator whose aim is to make the text sound more fluent, as if it had been written by a human. The post-editor will also make any necessary corrections to meaning. The text will then be proofread by a second qualified translator, whose role is to ensure that the translation is accurate, that its tone matches the source text, and that it reads fluently, as if written by a human. |
