Why you should avoid machine translation

I remember the days when Google translate first appeared. I thought I was going to be able to slam consistent As in all my French assignments. That was until I used it, and realised it did a dreadful job, and my language teacher would see straight through it.

By Andy Kehoe

Machine translation has come a long way since then and tends not to sound so much like somebody who has only had a language described to them, and never actually learnt it, has done the work.

But how good is it, really? Can a machine ever beat a good, old-fashioned translation by an actual human?

Not quite yet, and here’s why.


Inconsistent translation accuracy

A machine translation can often look good on the surface but, just as, after a conversation with a sleazy car salesman, you initially think the car you’ve bought is incredible, but then find yourself stranded in unknown territory with your bonnet on fire, AI translation can seem better than it really is at first glance.

Occasionally, a machine can produce quite accurate results. However, most of the time, they are way off the mark and, with the amount of intervention and heavy editing that will be required from a language expert, you may as well have just had them do it in the first place.


Writing style

Another problem with machine translation is the inability to replicate the human emotion and soul that goes into writing. Machines don’t yet have emotions, so they can’t create the same sort of connection with the reader.

Machine translation is designed to translate clinically and struggles to convey the unique style of the writer — a real problem in a saturated environment where it’s imperative that your content stands out.  

For instance, a machine would translate the Facebook Company’s new name, Meta, into the phrase ‘she is dead’ in Hebrew. A human being, perhaps one with more Hebrew than Mark Zuckerberg, might translate it differently.


Language is too complex for technology to handle

Machine translation, at least for now, is one of those things that seems too good to be true, and that’s because it is. It can obviously translate simple sentences very accurately but, when it comes to finding a way to say something that exists in one language but not the other, it struggles. For example, the English language has around six times more words than French.

So, until technology can account for the nuances and vast, complex differences between languages, machine learning really isn’t an option in a marketing setting. You will fare so much better hiring a native speaker to translate your content.

So, if you’re a STEM business looking to expand globally, don’t cut corners and rely on machine translation for your content. In a sector which uses a lot of technical, complex language to communicate effectively, this is doubly important. So, hire an expert!


Do you need help with content translation? Are you trying to expand into another country but don’t have the language skills to do so? At Stone Junction, we have a team that speaks 14 languages between them. Interested? Get in touch!

Our Clients
  • Technical PR for industrial computing - Beckhoff
  • Technical PR for construction from Finning
  • Technical PR for graphene from the Graphene Flagship
  • Renishaw
  • Sandvik
  • Sandvik Coromant