In the AI era, machine translation (NMT - Neural Machine Translation or SMT - Statistical Machine Translation) is more and more popular. As a result, so is the Machine Translation Post-Editing service - MTPE.
Are you tired of PEMT (Post-Editing of Machine Translation) resources claiming they can post-edit but they do standard revision instead? Would it help to use a provider with ISO18587 standard certification, the project managers trained with TAUS and millions of words post-edited in Polish and other languages?
Full Post-Editing means post-editing of Machine Translation to obtain a target text comparable to human output. Basically, to make the translation look as good as if it was done by a human translator.
Light Post-Editing means post-editing of machine translation to obtain a merely-comprehensible text, without attempts to make it comparable to human output.
Content with low visibility, like internal communication or any content that does not impact your brand's image on the international market is suitable for LPE.
From the quality perspective, it is crucial that the PEMT projects are handled by highly-skilled professionals with long-term experience in the Machine Translation Post-Editing area. These projects require a specific set of skills that, when supported by experience, guarantee the required outcome.
No. QA checks in the CAT tools should be done for both Light and Full Post-Editing. After all, it matters how correct and comprehensible your content is, regardless of whether it's marketing translation, website localization or translation of your internal communication.
No. When a client sends a Machine Translation for post-editing, the text has to be assessed as to whether it is at all suitable for post-editing in terms of the quality of the MT output and the amount of time that a reviewer will spend on it. In short, to save your money, make sure you use the right Machine Translation engine so that the outcome quality is suitable to be post-edited, rather than translated again and then revised and reviewed.
The free of charge MT engines like Google Translate of DeepL use the stock repository for the translations, hence the quality is there, but may not be as you'd expect it to be.
The paid ones, like Globalese, Systran, Tilde or Kantan use their own translation repositories as well as client-based or domain-based repositories o their clients which makes the engines highly customized. If a client does not have the required minimum of translations to build the repository sufficient to fully train the engine, they can use their own translations as master training material, and then complete the volume with a stock repository. Since the engines constantly "learn", the quality of such translations is much better and more fitted to the client-specific requirements. This also means the better and cheaper end result of the MTPE process.
Yes. Whatever guidelines or terminology the client wants to be implemented in the translated content, the professional post-editor will address those and use them in the full post-editing process. This also ensures consistency throughout the content, regardless of its type. This is relevant especially in highly creative content such as marketing collateral but also in more technical content translation to make sure terminology, style, tone, and voice are consistent.