Loekalization Blog

Here’s the deal: Thorsten Hendriks, a highly skilled human translator, just finished a mammoth 100,000-word translation from English to Dutch for Huawei via GienTech, an LSP. A month of hard work, expertise, and, let’s face it, probably an unhealthy amount of caffeine went into this project. And how does GienTech respond? Oh, you’re going to love this. They’ve taken his beautiful, human-crafted Dutch text, run it through machine translation to back-translate it into English, and are now comparing that machine-generated Frankenstein’s monster of a text to the original English. And because, shocker, the two don’t match perfectly, they’ve asked Thorsten to “review and implement changes.”

Let’s pause for a second to appreciate the sheer lunacy of this. GienTech hired a human—a real human, with decades of language evolution baked into his brain—to carefully and thoughtfully adapt English into Dutch. Then they decided to QA his work by running it through an algorithm that doesn’t know the difference between nuance and a stale bagel. This is like hiring a master carpenter to handcraft a bespoke table, then running it through a wood chipper and asking why the sawdust doesn’t look like the original tree.

Let’s be clear: back-translation as a QA method is already about as reliable as asking a parrot to proofread Shakespeare. But using machine translation for it? That’s a whole new level of absurdity. Of course the back-translated English doesn’t match the original—it’s not supposed to! Translation is about capturing meaning, intent, and context, not parroting word-for-word equivalence like some malfunctioning phrasebook.

And it gets better—or worse, depending on how much patience you have for this nonsense. They’ve asked Thorsten to fix the discrepancies their machine created. Fix what, exactly? The fact that machines can’t replicate human intuition? The fundamental misunderstanding of how languages work? Or perhaps the gaping hole where their QA budget should be? And the cherry on this sundae of incompetence? They expect him to do it for free. Free! As if his time, expertise, and sanity were just bonus features in this Kafkaesque circus.

Let’s talk about GienTech for a moment. This is a multi-billion-dollar tech behemoth, a global giant that, on paper, should be able to afford actual quality assurance. Instead, they’ve opted for a penny-pinching, corner-cutting approach that screams, “We don’t understand how words work.” It’s embarrassing. No, it’s tragic. Actually, scratch that—it’s hilarious, if you’re not the one being asked to clean up their mess.

The logical way to QA a translation is to hire a native-speaking, human reviewer—someone who understands both the target language and the artistry of translation. But no, that would require common sense and a willingness to invest in quality. Instead, GienTech has chosen to Frankenstein their way through the process, leaving chaos, confusion, and back-translated gibberish in their wake.

Thorsten, to his eternal credit, has refused to participate in this farce. And who can blame him? His work deserves to be reviewed by someone who actually speaks Dutch, not a machine with all the linguistic grace of a drunk toddler. GienTech’s approach is not only an insult to translators everywhere; it’s a waste of time, money, and whatever shred of credibility their QA department had left.

So here’s the takeaway: if you’re a company with global ambitions and a reputation to uphold, don’t treat language like an afterthought. And for the love of all that’s holy, stop using AI to clean up after humans. It’s not working, it’s never worked, and if this debacle is any indication, it’s just making you look foolish. GienTech, do better. Or at the very least, stop asking people to fix your mistakes for free.

Thorsten has better things to do—like not dealing with your nonsense.

Addendum: Update (February 7, 2025): After GienTech requested the removal of their name from this post and initiated discussions about improving their processes, they invited me to a private meeting on January 23 to “clarify” the situation. I declined, requesting a written response instead—ensuring transparency for my readers and maintaining accountability. Since then, GienTech has not responded. Given their silence, I am have decided to reinstate their name in this blog to reflect the full context of this situation. If GienTech truly aims to improve its approach to quality assurance and translator relations, meaningful public engagement—not private, undocumented meetings—would be a more constructive step.