Will We Ever Trust Machines with Healthcare Translation?
To those of us old enough to remember the days before the internet, machine translation seems like a miracle straight out of science fiction. Translation mechanisms that work perfectly have long been established in science fiction, from Star Trek’s universal translator to the Babel fish in Douglas Adams’ The Hitchhiker’s Guide to the Galaxy.
As such, many of us have expected that the pace at which technology has been evolving should mean that translation snafus and painfully awkward cross-cultural exchanges have become consigned to the history books.
And at first, it seemed Google Translate would fill that role. Or if not Google Translate, some other more sophisticated software. It especially seemed like a boon for healthcare translation, where medical professionals are often tasked with the difficult process of communicating complex medical terminology and concepts to someone who doesn’t speak their language.
But will machines ever be able to replace a good translator? Will we ever get to that glorious future where translation mishaps and cultural barriers become a thing of the past? Let’s take a look at whether machines can ever really take over medical translation.
The Supposed Promises of Machine Translation Services in Healthcare
Machine translation is a fast-growing industry. In fact, Grand View Research stated that machine translation had a $433 million global market size in 2016, and was expected to continue to grow exponentially from there.
Meanwhile, in 2015, Modern Healthcare reported on a new app from a company called Canopy, heralding it outright as the tech that would once and for all “break down language barriers for patients, doctors.” The company was started by a first-generation Chinese-American who saw firsthand how difficult it was for his Chinese-speaking family to access healthcare.
This followed a 2012 report from the Agency for Healthcare Research and Quality, which stated that patients with limited English language skills have longer hospital stays and are at higher risk of infections, pressure ulcers, read missions and falls. So an app that would smash through the language barrier, possibly even supplanting language specialists entirely, felt like a dream come true. One doctor even switched to the app after frustration with his usual over-the-phone interpreting system that left him with convoluted conference calls.
However, the app worked best for quick conversations geared toward gathering basic information. More in-depth conversations still needed phone interpreters.
Machine Medical Translation Fails
While machine translation has come on leaps and bounds in recent years, with neural networks and deep learning changing that way that machines process language to more closely mirror the human brain, it’s still a far way off from being able to replace professional translators. This is especially the case in the medical industry, where a machine medical translation hiccup can damage or even cost lives.
Medical mishaps due to human translation errors are nothing new. In 2010, NBC reported on a doctor who removed the wrong kidney from a Spanish-speaking patient as a result of a communication failure. In 2015, the Daily Mail reported on a British expat who had a needless double mastectomy in Spain after a translation problem.
On a more humorous note, Sputnik News reported on medical mishaps with Google Translate while British fans were at the World Cup in Russia this year. A man needed a tetanus shot after a dog bite and was using Google Translate to talk to the nurse. Google Translate somehow handled what the nurse was saying as “What kind of pizza did you have for dinner?” Thankfully, somehow, the man did manage to get his shot and survive to tell the tale.
Stories like these show how necessary it is to not blindly rely on machine translations. A single mistranslated word or phrase can lead to the unnecessary loss of body parts.
Why Machines Haven’t Quite Caught Up to Medical Translation Needs
For a bird’s-eye view of just how wrong Google Translate has been, there was a study conducted in 2014 that reviewed the accuracy of Google Translate in medical communication. The results were frankly depressing.
The study, published in the BMJ, translated 10 common medical phrases into 26 different languages. Only 57.7 percent of the translations were correct. The highest percentage of correct translations went to Western European languages, yet even those still only boasted 74 percent accuracy.
Some of those translation mishaps were incredibly serious. In Swahili, “Your child is fitting” translated into “Your child is dead.”
Although the study is from 2014, Google Translate still makes some pretty large gaffes, as highlighted by the World Cup story above.
Perhaps Gülay Eskikaya, an English/Turkish interpreter, said it best in a 2013 editorial for The Guardian: “Machines can only translate words and not meaning and will be unable to grasp concepts, abstractions or cultural references. Ultimately, machine translation fails to differentiate between the language of a literary masterpiece and a car manual, a United Nations convention and a text message. To the machine it’s one and the same.”
Since machines still can’t quite comprehend the nuance of human language, we’re a long way off from other uses like translation of medical forms and records. There are also more difficult aspects of translation, like multilingual typesetting, that call for the skills of graphic designers as well as translator. Suffice it to say, it looks like we’ll need bilingual staffing, over-the-phone interpreting and on-site interpretation for the foreseeable future, without question.
Louise Taylor is head of content for Tomedes, a company that provides professional translation services in 90+ languages to clients all over the world. She has been covering a wide variety of translation and language topics for over five years.