More than Words: Why Human Translation is Leagues Ahead

by KnowledgeWorks Global Ltd.

The American philosopher and linguist Noam Chomsky once famously said: “A language is not just words. It’s a culture, a tradition, a unification of a community, a whole history that creates what a community is. It’s all embodied in a language.” While these words are revered by many of those who work with languages for a living, decades of attempts to mechanize translation would suggest that the technology industry believes we are not far from finding an automated system which has a deeper understanding of these wider complexities of language.

As far as professions go, working as a translator can be perilous. Consistently high on the list of “professions most likely to be disrupted, eliminated or replaced by AI”, translators spend their lives looking over their shoulders, fretful of the next game-changing technology from Google or Microsoft which could spell the demise of their businesses and livelihoods. And, unfortunately for them, there never seems to be a shortage of innovations in the field of automated translation.

Since the arrival of Google Translate 14 years ago, the tech giant has been particularly active in driving advancements of the tool - from its 2014 acquisition of DeepMind to integrate machine learning functionality, to ‘Translatotron’, its latest proof of concept which will enable users to translate their speech while keeping their voice.

However, although it is free, easy to use across 100 languages and without doubt the most powerful and comprehensive translation tool out there, Google Translate, at least in its current state, cannot realistically be considered a threat to human translators. It has its benefits – quick word searches and simple sentences are frequently translated effectively and accurately. But the system often falls down when it comes to more complex sentence structures, and specifically with expressions and idioms which involve a greater level of cultural understanding. For example, when you input very common English idioms such as “hitting the sack”, “going cold turkey,” and “looking like a million dollars” and translate them into Spanish - one of the tool’s lead languages – you only get literal, and therefore incorrect, translations in return.

So, what is it that human translators do that computers are finding so difficult to replicate? Perhaps the most important skill which translators bring to the table is transadaptation. This is the ability to make directly translated material more relevant and appropriate for readers of the target language, or as Vadim V. Sdobnikov more eloquently puts it “to change various elements of cultural, visual, audio and other codes in order to better integrate the text into the matrix of the target culture.” For example, a Spanish translation specialist would know instantly that an idiom like “hitting the sack” wouldn’t cut it as a direct translation, and would instead seek out a more suitable alternative, such as the colloquial “irse al sobre” or if the context was more formal perhaps simply “descansar” (to rest).

Transadaptation involves a great deal of problem-solving as translators have to make a concerted effort to both remain true to the source text, and accurately convey its meaning, while also finding a way to make the translated text fit as best as possible in the target language.

In creative industries like advertising, branding, cinema and TV, where there is a greater need for words and terminology which evoke an emotional response among target audiences, a much more flexible form of translation tends to be adopted, known as transcreation.

There have been a number of amusing cases of brands making spectacular errors through using direct translation instead of opting for transcreation – KFC’s “Finger lickin’ good” tagline being translated to “We’ll eat your fingers off” in Mandarin, or Coors Light’s “Turn it loose” motif being translated into Spanish as “You will suffer from diarrhea” are just a few embarrassing examples.

As transcreation focuses far more on how target audiences emotionally respond to content, rather than rigorously ensuring that the exact meaning of words and phrases are expressed correctly, it is a particularly difficult human skill to replicate, and one which can only really be carried out by those with a deep cultural understanding of the target language.

Localization is another important aspect of translation, and many people and companies tend to underestimate its value, instead finding out the hard way that there is enormous disparity between translated content that seems like translation and translated content that seems like genuine, original copy. The most successful translation services will use mother-tongue translators in the target language, who have a deep understanding of grammar, language idiosyncrasy and dialect, as well as a strong awareness of cultural appropriateness, in the target language. Unsurprisingly, this is yet another aspect of translation which automated tools have struggled to come to terms with.

Translation is an art form. Beyond simply converting words from one language to another, translators need an in-depth understanding of the intricacies of languages, to decode meaning and untangle cultural complexities, norms and values. There is a great deal of craft that goes into taking content from one language and making it believable, trustworthy, evocative and authentic in another, and that skill should never be underestimated. There is no reason to believe that machines won’t one day be able to rival this expertise, but we are currently a long way off.

KnowledgeWorks Global Ltd. (formerly Cenveo Publisher Services) is the industry leader in editorial and production services for every stage of the content lifecycle. We are your source for language services, content creation, accessibility compliance, digital learning solutions and more. Email us at info@kwglobal.com.