Machine Translation Post-Editing (MTPE)

Time doesn't stand still with translations. Technology for working more efficiently with large amounts of data is constantly being developed. The quality of machine translation is nowadays much better than two years ago, since the quantity of global translation data that the software can access has increased exponentially. As a result, we would now like to offer this service to our clients. We currently offer all language combinations with German, English, French and Spanish. We are working on adding further languages to our selection of machine translation.

What does MT mean?
Machine translation refers to the automatic translation of content with a software. This makes it a sub-area of artificial intelligence (AI). A rough distinction is made between the statistical method (SMT) and neural networks (Neural MT). At the heart of both methods is a translation memory (MT engine), in which commonly recurring sentences are stored with the relevant translations. SMT calculates how similar the entries in the engine are to each sentence from the source text. Neural MT uses an artificial neural network to train the contents of the engine, thereby understanding connections between the source and target language.

Neural MT appears to prevail, since the translations sound better from a linguistic perspective. Even Google Translator has switched from SMT to Neural MT since November 2019. We work with a professional DeepL Pro Engine, which uses neural networks combined with AI and machine learning.

What does MTPE mean?
MTPE means machine translation post-editing, where machine-provided translations are revised and corrected by human specialist translators. For this, we work partly with our core translators and partly with some specialist translators acquired for this particular service. As with classic specialist translation, all translators chosen for PE are trained linguists, native speakers of the target language and have at least three years of professional experience in the relevant area of expertise. Another requirement is a good assessment of an MTPE test. The objective is to create a translation that is almost comparable to the result of a human translation.

Advantages and disadvantages
We don't see MTPE as competition to traditional translation. It is simply another option, with advantages and disadvantages: MTPE is quicker and cheaper, but also more prone to error. Traditional translation is still much more fluid in linguistic terms, since the translator understands the text, recognises connections between sentences and can interpret and structure the text accordingly at their discretion.

Client-specific glossaries and stored translations  
MTPE also uses your company-specific specialist terminology and stored translations (if you are already a client of ours). This means that the translation software accesses simultaneously the MT engine’s huge database and the glossaries and TMs that we have created for you. The client-specific data is prioritised.

For instance:
The German term “PSA” (Persönliche Schutzausrüstung - personal protective equipment) is found in the source file.
English MT translation: PSA (the engine doesn't recognise the acronym and treats it like a code that should not be translated).

If the correct English translation for PSA (PPE - personal protective equipment) is stored in our client TM, our TM automatically has priority.

Cases in which we recommend MTPE
If you need a translation very urgently, have a low budget and the translation is only for internal use, MTPE is the ideal solution. This applies in particular to tender documents, legal texts, financial and general technical texts.

However, we do not recommend MTPE with demanding journalistic texts, sensitive documents, texts in the image/marketing sector, very specialised texts (e.g. for the textile sector) and texts that will go to print and be published.

We are happy to advise you and look forward to your request.

Data protection
Texts that you consign to us for MTPE are subject to our usual data protection regulations.

Our provider DeepL Pro is subject to European Union data protection regulations and pledges to not save texts in either the source or the target language. The connection to the DeepL Pro server is encrypted and texts cannot be seen by third parties.