Google translate, the machine translation (MT) service is believed to translate in excess of thirty trillion sentences/year (Kuczmarski, 2018)[1] in more than 100 languages. But would you let it translate your IFU into Japanese?
MT applications may successfully handle non-technical texts in the most widely used languages. However, medical and technical translation demands stringent quality assurance, particularly when regulatory requirements have to be met. This task still requires skilled linguists with medical, pharma and tech sector expertise. Documents, where accuracy is more important than speed, require a human touch.
Of course, we would say that, as it is our business. But there is also extensive research to suggest that there are significant risks to using MT in health and scientific contexts. Nunes Viera et al [2] suggest that a cost-benefit calculation is needed where MT usage is considered. It is obvious that when a medical/technical text is translated it must employ the correct terminology in the target language. Otherwise, people, time and reputations are at risk.
Recent research by Khoong et al,[3] into the use of Google Translate in emergency department discharge instructions, showed that 8% of the Chinese translation was highly problematic and concluded that there was potential for serious harm.
Finally, Interpreting and Translating guidance from NHS England simply mentions that MT ‘should be avoided’ due to quality concerns (NHS England, 2018, p. 11).[4]
Clearly, the use of MT in healthcare settings could have serious implications. There is likely to be a breach of Duty of Care, not to mention potentially catastrophic legal consequences.
Here are some specific examples of a human linguist’s superiority over MT:
- Translation of an adverse reaction pharmacovigilance report requires specialist knowledge of specific medical concepts
- Drug composition information requires an understanding of the correct chemical terminology
- Medical translators are comfortable with Greek and Latin and understand specific nomenclature
MT applications don’t have all of this functionality
Obviously, many of us are focused on pricing. If your text is not crucial – it is not patient or clinician facing and is not a part of a regulatory submission – then an MT application may allow understanding of the main points of a text. In the early stages of a project cycle, MT translation may help you distinguish between documents that are simply interesting and those requiring professional translation. However, if you have an IFU, a PIL, a research article or some medical marketing material to translate and you want to get it right, then you should employ an expert human linguist. Your business will benefit from the quality translation that they produce.
MT may appear to be free at the point of use. However, its usage may not be consequence-free. Users are liable for their use of its output. Some organisations offer a simple disclaimer to disown the consequences of inaccurate translation; caveat emptor. (These disclaimers usually state that ‘any differences created from the translation have no legal effect’.) However, a professional translator is liable for the quality of their work. If you have ‘skin in the game’ then so should your translator, just as you would expect for your lawyers and accountants. As is often the case, the internet looks like it offers something for nothing. But it only looks that way.
When using a human linguist, remember:
- Linguists can usually translate between 1,500 to 2,000 English words per day.
- The translation is usually priced by the source word
- EU language medical translations cost from €0.14 per source word
- Asian or other non-standard language medical translations such as Arabic and Russian start at about €0.28 per source word
- The volume of source text will probably differ from that of the target text: 1000 Danish words = 1200 English words
- Quality is assured by a 2-step process- translation followed by a separate review of the initial translation by another linguist
- Translation does not usually include formatting or proofreading services – these come at an additional cost
- Where possible, provide the linguist with background information: terminology lists, previous translations or relevant materials in the source language
[1] Kuczmarski, J. (2018, November 28). A new look for Google translate on the web
[2] Lucas Nunes Vieira, Minako O’Hagan & Carol O’Sullivan (2020) Understanding the societal impacts of machine translation: a critical review of the literature on medical and legal use cases, Information, Communication & Society, DOI: 10.1080/1369118X.2020.1776370
[3] Khoong, E. C., Steinbrook, E., Brown, C., & Fernandez, A. (2019). Assessing the use of Google Translate for Spanish and Chinese translations of emergency department discharge instructions. JAMA Internal Medicine, 179(4), 580–582.
[4] *NHS England. (2018). Guidance for commissioners: Interpreting and translation services in primary care. NHS England / primary care commissioning