Can you use AI translation tools for employee communication?
Language is our communication tool. It’s more than simply putting words, sentences, and phrases together. As communications experts, we work daily to create persuasive and easy to understand content for our clients, in their authentic and unmistakable voice. We succeed in this by using a combination of human skills and AI software. We asked our colleague Lynn Nothegger about AI translation tools such as DeepL, and how to get the most out of them.
Unequity is a specialist in employee communication. It’s our job to communicate HR programs and opportunities to our clients’ global workforces, or to a specific group of employees. It is in the early phases of the project, when we are defining the target group, that we generally learn in which languages we will need to communicate. And it’s then that we are often asked how we deal with translations.
Some customers ask us if we could use free AI translation tools such as DeepL or Google Translate instead of professional translators. It’s a valid question, and our answer is, ‘of course!’. We are all familiar with these tools, whether in a professional setting on in our private life. They are generally pretty good, and as they use artificial intelligence, they are getting better all the time. But are they good enough for professional business communication?
To find out more, we asked our expert. Lynn is Scottish, and our resident copywriter and translator.
DeepL and Linguee
I first heard of DeepL around 2012 when a colleague asked me if I’d ever used it. When I investigated it, I was quite impressed. It was better than the other tools on the market at the time. Over time, I became more aware of the issues with DeepL and other AI tools.
As a German company, DeepL is probably the most popular translation tool in this country. The company’s initial product was Linguee, an online dictionary for language pairs. The Linguee software basically scrapes translations from multilingual websites and offers these examples of how others have translated words or phrases. Clever, I thought at first when I first discovered it. But we must remember that it’s artificial intelligence. As the screenshot above shows, it struggles with words that don’t have a clear translation.
“Compensation and Benefits” has a specific meaning in the HR field. “Compensation” includes wages and salaries, while “benefits” include things like pension plans, vacation time, and health insurance. In the image above, we can see that Linguee presents various translation options, some of them incorrect. When it comes to employee communications, it’s vital to be sure that we are using accurate language.
DeepL, Denglish, and false translations
The translation tool DeepL is based on the Linguee database. It is ‘learning’ from German websites, so it picks up some incorrect translations. These are then being used by other German companies, which becomes a vicious circle of wrong translations. And I’ve noticed that there is a lot of Denglish on DeepL.
Denglish is a mix of Deutsch and English that is often used by German speakers. Some of these words are so common that you might not even realize that native English speakers don’t use them. Some have even been ‘eingedeutscht’ and are used in everyday language, even when speaking German. And sometimes English speakers like me, who live in Germany, use them too because they are so used to hearing them. You may have heard of the more common examples, like home office, handy, beamer, and public viewing, but there are a lot of less well-known examples, such as “offering” to mean “Angebot”, which I’ve written about in my LinkedIn newsletter.
If you use DeepL to translate text from German into English, you might not notice that the translation sounds odd to British or American readers. Or that the literal translation of an idiom from German to English makes it particularly difficult for non-native English speakers to understand. And that’s before we get to translations that aren’t just weird, they are incorrect. And sometimes weird and incorrect. I can’t prove that DeepL translated this one, but it was definitely not written by a native English speaker!
To unravel can mean “to unwind” or “to untangle”, like untangling your headphone cables, but it doesn’t mean “to develop your professional skills”, as the original German suggested. And since “unravel” can also mean a person falling apart and losing control, it’s not just wrong, it’s extremely negative.
Sometimes the translation tool will throw up a few different options. This is completely normal – if you ask ten translators to work on a text, they will word the translation slightly differently. It doesn’t mean one of them is correct and the other nine are wrong. Just that they chose slightly different words. With these translation tools, it can be challenging to know which is the right option. And some options will subtly change the meaning of the sentence, which can be difficult to detect if you aren’t a native speaker.
Data Protection and liability
Another thing to consider when using DeepL is liability and data protection. They explicitly state on their website that they accept no liability for the accuracy of translations. Not ideal when you are communicating highly complex information about, for example, an employee share purchase plan.
Only the Pro version of the site provides end-to-end data encryption, and the translation of personal data is restricted.
Professional employee communication requires professional translation
During our projects, we use professional tools. Alongside project management software, we use DeepL and other helpful tools, such as translation management software which ensure consistent terminology is used. As Lynn demonstrated, we never rely on AI tools entirely. We believe that professional communication needs more than accurate translations. It also means adjusting the tone of voice to suit the client and using the correct terminology rather than just accepting the suggestions from AI, even if it looks right at first glance. We work with various translation partners who combine human skills and experience with modern tools such as AI translation, even for unusual language combinations. The human quality check is not an optional extra—it is a standard part of our process, and just as important as the professional source text in German or English.
Links:
In her LinkedIn Newsletter, Lynn highlights less well-known Denglish mistakes. Subscribe here to improve your English:
https://www.linkedin.com/newsletters/6935228618771988480
DeepL – You can find—and test—the translation tool here:
https://www.deepl.com/de/translator
- This article was published on
Share this page directly: