14/04/2020 - Translation
Estimated reading time: 7 minutes
This April, we devote our blog to flops, fails and fumbles in the translation industry. However, what may make us grin while reading a menu on holiday can quickly become problematic in a professional setting. Read more to find out how you can avoid translation fails and what you need to be aware of when using machine translation systems.
Mistranslations: Should you laugh or cry?
Menus at foreign restaurants often serve up a few laughs along with the food thanks to sometimes mild, sometimes glaring mistranslations. If an avocado becomes a lawyer, a pizza quattro stagioni a four-train-station pizza, or Liebfraumilch turns into beloved lady milk, the world won’t exactly stop turning. We smile indulgently and sympathise with the poor soul who had the menu translated into five languages and had to trust the internet to spit out the correct translations.
However, mistranslations can have considerably graver consequences for larger organisations. When the Swedish company Electrolux launched its vacuum cleaner on the UK market, its English slogan “Nothing sucks like an Electrolux” was a risky choice. Although a vacuum cleaner technically sucks up dirt and other debris, in US English “to suck” can also mean to be of inferior quality or a failure. Even though some of its target group might have appreciated the bold and brassy advertising, there is a good chance some potential customers would have been put off the idea of purchasing an Electrolux for good. As we can see, even the most globally recognised brands can fall prey to a translation faux-pas now and then.
Way back in the days before the internet, KFC had its “Finger-lickin’ good” slogan translated into Mandarin for the Chinese market. Unfortunately, someone took something too literally and the Mandarin version ended up reading as “Eat your fingers off”. Not exactly the most appetising way to attract diners. Somewhat more recently, UBS apparently assumed everything when its tagline “Assume nothing” was mistranslated into the equivalent of “Do nothing” in many foreign markets. The company had to invest $10 million in a rebranding campaign after this spectacular flop.
Reputational damage: More than just putting your foot in it
The “Swinegate” case demonstrates more clearly than ever how one wrong translation can result in extensive damage to a company's reputation. The British chief economist of UBS Wealth Management laced his statement about the impact of the swine flu on price inflation in China with some typical British humour by stating that the swine flu “only matters if you are a Chinese pig”.
What appeared to be a harmless joke to us Westerners was perceived as a gross insult by Chinese citizens. The phrase "Chinese pigs" was taken out of context, misinterpreted and translated in a way that equated Chinese nationals with pigs. The result was a wave of indignation and – despite numerous apologies from UBS – the collapse of a deal with a Chinese state-owned company.
It can literally be a matter of life and death if medical terms are translated incorrectly. According to an American study, 1,500 of every 30,000 treatment errors can be traced back to mistranslations. Such incidents can have catastrophic consequences for those affected, e.g. if in a report a case of thyroid cancer becomes pancreatic cancer, or the instructions for the use of a defibrillator contain misleading information.
Yet what is the best way to avoid such debacles in the first place? Do they arise from a lack of professionalism on the part of the translator or the use of unsophisticated machine translation programmes? The answer to these questions is as unsatisfactory as it is logical: It depends. In principle, however, anyone who does not work with a professional language services provider risks putting their foot in it. This doesn't mean, however, that machine translation systems per se are inadvisable.
Opportunities and limits of Neural Machine Translation
Technology began to make its presence felt in the translation industry a long time ago. Specialised computer programmes, known as CAT tools, support language services providers in their work and increase the quality of translations. Machine translation programmes have also been around for a while. When it comes down to it, they are not really any different from electronic dictionaries. In the wake of big data, however, they have developed very rapidly over the last few years, and for some time now people have been enthusing about Neural Machine Translation (NMT).
NMT works with artificial neural networks and is already delivering surprisingly good results. Of course, this does not mean that technology can ever completely replace humans, but NMT does prove useful when translating content with little emotional weight. Emotional weight refers to the emotionality of a text, i.e. how strongly the text should appeal to readers on an emotional level. Texts with little emotional weight include regulations, product descriptions and instructions. The communicative function of the information or instructions is the primary focus of these texts.
NMT can certainly save time when translating such texts, because it reduces the amount of research required. Nevertheless, mistranslations are also ubiquitous in this area. For this reason, anyone using NMT should firstly be proficient in the target language and secondly evaluate the results delivered by the machine very carefully. Technical and key words in particular should be researched thoroughly.
So how emotional can a machine be?
To further complicate matters, there are also some types of text with little emotional weight that nevertheless cannot be translated satisfactorily with NMT. This includes all documents which, for legal reasons, must not contain errors under any circumstances, such as package inserts, instruction manuals, financial and medical reports. In order to handle demanding translations like these, you need a professional translator who is familiar with the relevant specialised domain and can detect any linguistic peculiarities or ambiguities. Otherwise, your thyroid cancer could quickly become pancreatic cancer.
Text types with high emotional weight such as advertisements or website and app content are definitely not suitable for NMT. The reason lies in their communicative function. The purpose of texts with high emotional weight is to convince, form opinions and trigger action. They are characterised by elegance and wit, and they implement rhetorical means such as puns or ellipses. Without these stylistic features the text would not fulfil its communicative function.
And this is precisely where the human factor becomes indispensable. Neural networks can only translate content that is stored in their databases. They cannot create anything new, and they do not have a sense of style, readability or aesthetics. And it only gets more difficult when it comes to wordplay. A specific play on words will very often work in the source language but not in the target language. A professional translator is aware of a text’s communicative function and in such cases will try to incorporate another pun that functions in the target language. NMT cannot do that. It is and remains a data processor. It lacks that all-too-human factor: an eye for the big picture, intuition and creativity.
The best way to avoid translation fails
Contact a professional language services provider
- The safest way is to work together with a professional translation partner. Professional translation companies are ISO-certified and comply with comprehensive data protection standards.. They also employ specialised language technology that ensures uniform terminology and consistent translations.
- Always work with the same language services provider whenever possible, as they are most familiar with you and your requirements.
Use NMT sensibly and carefully
- First determine the purpose of the translation you require.
- A stand-alone NMT text is appropriate if you only want to know what a text is about, and the translation is not going to be published or passed along.
- If you want to use the NMT for other purposes, it is best to have a professional editor/proofreader look over the translation.
- Don’t forget that free neural machine translation systems offer little or no data protection. Your texts will be saved in their database and used as a reference for future translations! It is highly recommended that you contact a professional language services provider for secure machine translations that do not compromise your data.
We are more than happy to help
Do you have any questions? With SwissGlobal on your team, you have an expert language services provider to support you with your individual needs. We all know that lawyers, train stations and beloved ladies are not fit for consumption and that mistranslations really suck. We avoid translation fails by relying on our experience, expertise, ISO-certified quality management and the use of professional process and language technologies. This also includes an encrypted NMT solution to ensure that your data is always protected and secure.