Machine translation: ready to replace humans in corp comms? Our test

Machine translation: ready to replace humans in corp comms? Our test

An illuminated blank billboard at night

When it comes to machine translation, even we translators have to admit that the advances in using artificial intelligence to quickly render text from one language into another have been impressive. So is machine translation a good option for fully replacing human translators in corporate communications? Here at Lennon Language Solutions, we have tried web-based machine translation providers that leverage the knowledge available on the Internet. Our experience shows that there are still plenty of pitfalls along the path to greater language automation.

The first problem is that the online helpers always play by the rules. Obeying conventions might sound good at first, but keep in mind that a lot of our real, day-to-day speech and writing involves regularly making a conscious decision to bend or even break the rules. A machine assumes that all input is correct and sticks slavishly to the wording and structure of the source. The result is untranslated text, incorrect results and a sometimes jarring, unnatural flow. An experienced human translator, however, can rely on intuition, experience and feeling to correctly interpret the errors or conscious rule-breaking inevitable in human writing — or to split up, combine or reorganize sentences and paragraphs to produce a natural flow in the target language. In short: A human translator knows when to break the rules.

The second problem is that the bots are frequently unable to account for regional differences in the target language. What may sound like a minor hiccup can result in major misunderstandings in languages with extensive global variety, such as English or French. Mixing conventions from regional dialects makes professional corporate communication appear amateurish and haphazard at best, and can lead to misunderstandings and cultural affront at worst. These regional differences are something an experienced human translator can easily pick up on to tailor messages to today’s global audience.

The third problem is the lack of consistency that arises when web-based machine translation providers grab whichever translation for a term appears appropriate in the given context without sticking to only one variant. Further to that, they are unable to take basic client preferences into account, such as whether to write “xx percent” or “xx%.” The result is a text speckled with inconsistent terminology and usage, perhaps even placing the author’s attention to detail in question. Again, an experienced human translator knows to look out for consistency issues to ensure a message that goes down smoothly and makes the author look good.

To be fair, there are things web-based machine translation does well. Artificial intelligence is very fast and admittedly capable of picking up on certain nuances. Texts that benefit from a literal translation (anything straightforward with little subtext, or with a relatively small set of clearly defined terminology, such as most technical literature or legal documentation) also come out pretty good. Essentially, anything sounding less inimitably “human” in the source language will produce satisfying results in the target language. Be that as it may, it is also our experience that even these texts can be improved by having a human vet and punch up the results.

But to come back to our initial question: No, machine translation is not a good option for completely replacing human translators in corporate communications — at least not yet. It lacks the human capacity to know when to break the rules, meaning it cannot independently reorganize syntax or interpret errors to rectify them. Remarkably, machine translation also still has a long way to go when it comes to picking up on regional variation in language and consistency. The unedited outcome is therefore still a strange mix of stiff compliance and slipshod complacency. Humans are still necessary to improve the results.

Here’s what we have learned: The more human a text, the more you need a human. Technology can help expedite certain processes. But if your goal is to add even more humanity to a situation, entity or perspective — which is ultimately the point of corporate communications — you still need a human somewhere along the line to act as a conduit for your message and interpret it from person to person. That’s why machines aren’t ready to fully replace people yet in corp comms. Rather than simply buying into the hype, the bare minimum is to have an experienced human linguist vet and refine machine translation and make it sound like what readers expect and need for a true emotional connection to be established.