With wearables now on the market for language interpretation, like the Pilot earpiece, there’s an ongoing excitement for artificial intelligence to remove the need for human interpreters. A slick marketing video shows a language barrier interaction flawlessly facilitated with immediate phrase assistance and translation. There are smiles and amazement at this new-found, conversational technology.
Added anticipation for interpreting accuracy came from an updated 2016 Google Translate capability that uses a neural machine translation engine for translating whole sentences at a time rather than one word at a time. The demand for language assistance is massive as Google claims to translate 100 billion words per day.
In recent years, much attention has been paid to Google Translate’s support of medical phrases between doctors and patients. What was once limited usefulness in the healthcare setting now has led to claims of 90%-plus translation accuracy according to a 2019 report published in JAMA Internal Medicine.
Technology-augmented translation in casual conversations is one thing, but there’s more at stake when bridging the language barrier for what can be life or death situations.
Consideration for translation in healthcare is not without caution since misinterpretations have reportedly led to major medical errors. Yet, with recent accuracy improvements by Google Translate and others, one might ask:
How will this affect hospital-based interpreters?
Relying on computers to translate medical phraseology and related social context is fraught with risk. This is why hospitals are not replacing human interpreters with AI translation systems anytime soon. Major US hospitals depend on and utilize both in-house and contracted interpreter services accessible by phone, video feed, or in-person. Support is available not just for Spanish, but advertise phone support for over 200 languages. To prevent misinterpretations and risk of malpractice, a level of quality from certification and continuous training allow for interpreters to navigate language subtleties and possible misunderstandings that may arise.
What about cultural nuances in medical conversations? Can these be successfully conveyed by artificial intelligence?
In a recent article from The Economist entitled, “Lost without translation”, attention is given to human translators and their ability to provide cultural context. Language problems during crisis response feature cultural differences as well as linguistic ones.
In 2017, 700,000 Rohingyas escaped persecution in Myanmar and fled to Bangladesh yet language presented an obstacle more difficult than crossing a border. The biggest language barriers surrounded health issues, according to a linguistics researcher with Translators Without Borders. Local interpreters were hired, but many were not effective in areas where over 20 languages were spoken.
Not isolated to Asia crises, language was at the forefront of the most recent Ebola outbreak in western Africa. Difficulty arose because most international advice on prevention management came in English and French, languages that they were less likely for the target population to speak. In the future, according to the article, automated translation tools such as Google Translate could help with rarer languages. For this to happen, what’s recommended is to have translation software with parallel text that has already been translated by humans. Recently, Translators Without Borders is giving some of its language data to companies such as Google and Microsoft for that purpose.
The Economist concludes that, “these problems are better thought of as cultural rather than linguistic: the words exist, but it is vital to have interpreters who know which can be used and when.”
When accuracy and cultural context are required, human interpretation in healthcare settings is the best option for now. According to Fabio Torres, Senior Education & Recruitment Manager at US-based Translation and Interpretation Network, there are important factors to consider in depending on human expertise compared to AI.
“While a computer may be programmed with extensive vocabulary in a particular setting, a trained, certified human interpreter is able go beyond mere vocabulary to identify language use complexities and context. For example, an interpreter who speaks Portuguese may at times be called on to interpret for a patient who is from Brazil, Portugal, Angola, or even Mozambique. Although these countries share the Portuguese language, the culture of these countries are completely different. A trained, certified interpreter is aware of potential pitfalls and can compensate for them using their complex layers of knowledge surrounding language and culture. In the healthcare setting, mistakes made in machine translation can cause ethical dilemmas and may even endanger the lives of patients.”
Hospital researchers who recently tested medical terminology translation found that:
“Google Translate is more accurate than a lot of clinicians believe and is definitely more useful than not providing anything at all. But is best used in combination with human interpreters.
Perhaps there is a role for interpreters in these settings to receive AI augmentation in some way as noted above.
Google Translate and other language barrier tools are becoming a technology standard for conversational assistance. Caution is called for, however, when linguistic and cultural factors are considered in scenarios where terminology certainty is extremely consequential to the doctor and patient relationship.