AI-assisted Translations with Post-Editing

For us, post-editing is a two-stage process. First, a carefully configured, AI-supported engine translates the source text according to your requirements regarding preferences, style templates and binding terminology glossaries, which can be provided in advance or during the translation process. Native-speaking specialised translators then check and revise the result in accordance with ISO 18587. We only offer this full post-editing service.

For translation and post-editing we work in Trados Studio and Trados GroupShare, combined with a carefully tested, configured, and trained large language model (LLM). The engine is controlled by glossaries, style guides and project-specific requirements, ensuring that the pre-translation is aligned with your terminology and your style from the outset. If you already have a maintained translation memory, we incorporate it into the process: 100% matches and complete context matches are pretranslated directly from the TM. The AI engine handles the remaining text. This means that technology, your language assets and expert control are already closely interlinked from the quotation stage onwards.

Post-editing is carried out exclusively by qualified, experienced native-speaking specialised translators. All post-editors undergo a multi-stage evaluation process to demonstrate their qualifications in accordance with ISO 18587 before working on client projects. This ensures post-editing meets the usual COMLOGOS quality standards, with humans taking responsibility.

Post-editing delivers a consistent translation that is terminologically precise and stylistically well crafted. Spelling, grammar, terminology and formatting are checked and corrected. Consistency and an appropriate style are ensured. It is standardised in accordance with ISO 18587, the international standard for the post-editing of machine translations, for which COMLOGOS holds certification.

If required, the post-edited result can undergo an additional revision in accordance with ISO 17100, for example in cases where particularly high stylistic or accuracy requirements apply. Post-editing on its own is particularly suitable for efficient, scalable translation of large volumes. The additional revision is recommended when there are special stylistic requirements. We can advise you on the most suitable approach for your text.

Yes, an additional revision is possible. For context: Post-editing itself is standardised in accordance with ISO 18587. On request, we additionally revise the post-edited result under the four-eyes principle according to ISO 17100, so that an AI-supported translation can also be processed within this standards framework — we're happy to advise you on this.

Both are carried out by qualified native-speaking specialised translators. Traditionally, post-editing (per ISO 18587) has referred to AI-assisted translation, whereas revision under the four-eyes principle (per ISO 17100) has referred to a translation created entirely by humans. We can combine both processes on request: An AI-assisted translation can also be additionally revised in accordance with ISO 17100.

The effort varies depending on the text type, language pair and quality of the engine. However, daily throughput is noticeably higher than with traditional specialised translation. We calculate the exact throughput on a project-by-project basis and include this information in the quote.

While the raw AI output itself may only take seconds or minutes depending on text length, immediate delivery is not possible. Each order goes through several stages: file analysis, calculation, scheduling of the post-editors, the post-editing itself, and a final internal review. Overall, however, the processing time is significantly shorter than for a pure human translation.

If you order both AI-assisted translations and human translations, we create two separate translation memories for each language combination: one for the human translation and one for the AI-assisted translation. This ensures that the two stocks don't mix and that your human TM remains uncontaminated.

Customer-specific trained engines are part of our development roadmap, but we are currently focusing on other process topics. To ensure consistency for our customers, we already use your language assets today: Translation memory and terminology database, plus the dedicated translator principle, whereby the same person handles your projects wherever possible. If a well-maintained TM is available, 100% matches and context matches are incorporated directly. We are also happy to incorporate your company-specific terminology and reference materials.

We do not offer post-editing of externally generated machine translation output. Without knowing which engine was used, it is impossible to reliably assess the effort involved, and the quality of the initial output cannot be controlled. We provide AI-assisted translations and post-editing as a seamless, end-to-end process from a single source. This way we retain control of the engine, terminology and quality right from the start.

We do not currently offer pre-editing, as it can be time-consuming and would reduce the efficiency advantage of AI-assisted translations compared with human translation. We always check in advance whether a text is suitable for AI-assisted translations. If it isn't, we can advise you on 'MT-friendly writing' to help you prepare your source texts optimally.

It is not possible to certify post-edited, AI-assisted translations. If you require a certified translation, the correct option is to request a human translation — please get in touch.