.png)
AI as a starting point
We use machine translation to generate initial drafts where appropriate. This supports faster turnaround and cost control. Certain content may be translated using AI and machine translation only.
Human involvement depends on risk
Human review is applied based on how the content will be used. Higher-risk content requires deeper human involvement, including one or more review stages.
Controlled AI Output
We offer more than basic translation. We maximize AI and machine translation output by selecting the right engine, applying terminology, and using translation memories, style guides, and predefined settings. AI and machine translation (where applicable), along with engine selection, are configured in advance as part of the setup.
Translation memory (TM) priority
If previous translations are available and approved by humans, they override AI output. This includes both client-provided materials and previously validated translations.
Terminology guidance
AI is guided by predefined terminology to improve consistency and reduce errors.
We use multiple AI engines such as Google, DeepL, and Microsoft. We select the MT/AI engine based on:
- language pair and direction
- subject matter
- past performance (based on human-verified results)
Standalone machine translation (MT)
We offer MT-only as a separate service. This is suitable for internal use, reference, or high-volume content.
AI with human post-editing (various stages)
AI output can reach near human translation-level quality when combined with full post-editing. This follows ISO 18587:2017, which defines requirements for post-editing of machine translation. Additional stages may be applied depending on content sensitivity, subject matter, or client preference.
Despite latest advances in the technology human involvement is still required for most content. The level of involvement depends on risk, purpose, and use. Some content can be used with minimal review, while other content requires full post-editing or full human translation.
.png)
