{"id":24645,"date":"2021-06-18T10:41:42","date_gmt":"2021-06-18T08:41:42","guid":{"rendered":"https:\/\/www.oneword.de\/?p=24645"},"modified":"2023-01-05T11:53:07","modified_gmt":"2023-01-05T10:53:07","slug":"why-mt-needs-the-pe","status":"publish","type":"post","link":"https:\/\/www.oneword.de\/en\/why-mt-needs-the-pe\/","title":{"rendered":"Why MT needs the PE"},"content":{"rendered":"

18\/06\/2021<\/small><\/p>\n<\/div>

Why machine translation needs post-editing<\/h1><\/div>

Machine translations that rely exclusively on artificial intelligence instead of professional post-editors can be a complete flop. This is shown by numerous, usually unintentionally entertaining examples from real-life scenarios. This even happened to an e-commerce giant recently when it shouldn’t have. But, one thing at a time!<\/p>\n<\/div>

Read more here<\/span><\/i><\/a><\/div>
<\/div><\/div><\/div><\/div><\/div>

We told you so!<\/h4><\/div>

Yes, it’s true… Machine translation (MT) now makes it possible to produce texts in many languages quickly and inexpensively. And MT systems such as DeepL or Google Translate are even sufficient in the business world for retrieving information quickly and for short bits of in-house news. But not for official corporate communications! To date, none of these practical everyday helpers have managed to translate specialist texts, technical documentation or product descriptions consistently and without errors. They lack specialised terminology, industry knowledge and context. That’s why artificial intelligence needs human expertise.<\/p>\n<\/div>

<\/div>
<\/div>
<\/div>

Error with an announcement<\/h4><\/div>

When companies resort to MT systems anyway to save money and time, the translation results can sometimes be entertaining, and at worst, embarrassing and incorrect. Even in 2020, you may still come across operating instructions translated from English to German that ask you to “press the strength key” when you should actually press the power button. Or you might read about “patriotic sprayers” in technical documentation instead of “patriot sprayers<\/a>“, high-performance field sprayers in the agricultural sector.<\/p>\n

An “anti-jack knife feature”, a feature between a towing vehicle and a trailer, has also been called an “anti-snatch knife feature” by machines. This can be vitally important in road traffic. And when it says “the airline must remain connected”, it’s crucial to know that it doesn’t mean an airline company, but instead that the air line must remain connected. Ignoring this mistake and identifying it as a correct translation can truly be a matter of life and death.<\/p>\n

In English, this kind of thing is perhaps still easy to understand. Also that “right lighting” and “left lighting” explains which side the lighting is on, rather than meaning “correct” and “abandoned” lighting. Even with digital topics and the possibilities of cloud computing<\/a> the machine is not faultless, but almost poetic: You sometimes see “cloud native” rather than “cloud-based”. Of course, many people would think themselves capable of correcting this type of a translation faux pas.<\/p>\n

In French, however, this soon gets difficult for most people. Although, when it comes to this language, a missing accent in the source text is enough to offend a lawyer: The subtle difference between “arr\u00eat\u00e9” (“regulation”) and “arr\u00eate” (“stop it!”) led to a legal text stating “Don’t make a fool of yourself!” instead of talking about the regulation in question.<\/p>\n<\/div>

<\/div>
<\/div>
<\/div>

Old Swede!<\/h4><\/div>

However, when the language is a bit more “exotic”, translations can slip through the net of amateurish user checks. Amazon is currently demonstrating that you don’t have to go too far to find an example. Just a few days ago, the e-commerce giant launched its own Swedish-language website in Sweden, which attracted attention with “astonishingly unprofessional” translations<\/a>. The example that got the most laughs was the word used for “rooster”, which was translated into the Swedish colloquial term for male genitalia.<\/p>\n<\/div>

<\/div>
<\/div>
<\/div>

Conclusion: That’s<\/em> why machine translation needs post-editing<\/h4><\/div>

It can’t be repeated often enough: artificial intelligence needs human expertise. And machine pre-translations (!) must not be used on a one-to-one basis in official corporate communications, but must instead be checked, corrected and standardised by trained translators (post-editors). It takes someone who knows the language and specialised terminology and who does not lose sight of the economic advantage of MT: the solution is machine translation + post-editing (MTPE)<\/a>.<\/p>\n

We’ll be happy to explain<\/a> how, when and to what extent this process works best. We’re also happy to explain why a potential collaboration using MTPE should always start with a consultation with MT experts to achieve translation results that work, rather than them being remembered for making people cry with laughter. Just get in touch with us.<\/p>\n<\/div>

<\/div>
<\/div>
<\/div>

You may also be interested in:<\/h4><\/div>
<\/div>
<\/div>
<\/div>

\u2013 What is MTPE?<\/a><\/p>\n

\u2013 What’s the difference between human translation and MTPE?<\/a><\/p>\n

\u2013 Case Study: How HELLA finds success with MTPE<\/a><\/p>\n<\/div>

<\/div><\/div><\/div>
\n\t\t
\n\t\t

Recent Posts<\/h4><\/div>\n\t\t