Translations: man vs machine

Nico Nortjé

Statistical machine approach

We've all no doubt used the services of such machine translation (MT) programmes as Google Translate when searching for a word or wanting to get the gist of information in a foreign language. Wikipedia describes Google Translate as a free, web-based MT application, translating between approximately 58 languages, including Afrikaans. Google Translate follows a statistical machine translation approach whereby patterns in already translated texts are explored for the purpose of intelligently guessing what the appropriate translation should be.

The 'guessing' part is where professional translators enter and claim their worth. According to Dr Nico Nortjé, CEO of South African-based Language Inc. - a translation services company that translates to some 57 languages - it is highly unlikely that machines will take over the work of trained professionals. "While machine translations have evolved significantly, services like Google Translate are often notoriously (and sometimes humorously) unreliable. These programmes undoubtedly have their place and do contribute to diminishing the communication barriers between people and do create a general understanding of to-be-translated text, but have to be used with caution."

Although an incorrect or out of context translation can sometimes elicit nothing more than a chuckle, these errors can have significant communication and even financial consequences. So, for example, a shipment of food products aimed for the Middle East immediately rejected when the unprofessionally online translated list of ingredients included 'motor oil' instead of 'vegetable oil'.

Pros and cons of Google Translate

In an article by Alta van Rensburg, Cobus Snyman & Susan Lotz (2012, US Language Centre) assessing the pros and cons of using Google Translate in an academic environment, they affirmed that humans' innate capacity for language, world knowledge, logic, and the ability to think and correct themselves distinguish them from machine translations. "Our ability to form expressions is not limited to a certain database, whereas a computer can only produce what it has been programmed to do or reproduce data from a database," says Alta van Rensburg, head of language and communication services at the University of Stellenbosch's Language Centre. "Advanced technology allows computers to do extraordinary things, but their performance is still dependent on the input they receive."

Especially in a business environment, correct translations are imperative. "Nowadays businesses need not be bound by geographical boundaries," says Dr Nortjé. "A brilliant idea hatched in South Africa might just as well be a money-spinner in Portuguese-speaking Africa or the East. When you explore markets beyond your own country, language becomes a crucial determining factor for success. And not simply straightforward translation of your marketing pitch, but a true localisation of the language, involving a comprehensive study of the target culture in order to correctly adapt the product to local needs."

But English is the universal language of business, you say! Research will have it differently. According to Lior Cohen, vice president of Net-Translators Ltd only about 15% of the global population speaks English as a first or second language. Apparently most people around the world surf the Internet in their mother tongue.

Some of the challenges of machine translations

Word order: 40-50% of words in a typical English sentence end up in a different position in the corresponding French sentence. Getting the word order wrong not only makes for horrible-sounding sentences but also can change meaning, often in comic ways.

Idiomatic speech: some idioms or concepts are true to just a specific language due to the history of the country, specific events or regional expressions. In the 1960s, the USA's CIA apparently tested machine translations from English to Russian and back to English. The phrase 'out of sight, out of mind' came back as 'invisible insanity'.

"People simply expects too much from such translations," Van Rensburg says. Van Rensburg, Snyman and Lotz (2012 US Language Centre) hold that if a text translated by Google Translate is destined to be published and used as an official document, the revision that such a text would require to be made useful could become quite expensive. "A professional translator would be best suited to post-edit and improve texts of this nature, and in the case of translation products of a particularly low quality, one may ask whether such intervention would be worth it, or whether it would not have been more cost-effective to use a professional translator in the first place."

A future for translators?

According to Alta van Rensburg (US) the future of language practitioners and professional translators is undoubtedly safe. "Although permanent work is somewhat limited, there is a huge demand for well-experienced freelance translators. Of the 31 translators that recently participated in an empirical study 19 were full-time freelance translators, seven were elsewhere employed and delivered language services in their free time and five were permanently employed as language practitioners in a language centre," she says.

Language Inc. has recently become South Africa's first company to fully comply with the world's only acknowledged quality standard for translators and make use of about 350 freelance translators per month.

A degree in languages, language science, translation and/or language practice forms the prerequisite for a career in translations. "Post graduate training in translation theory is also invaluable," says Alta van Rensburg. "Another option might be a different degree like law with a post graduate diploma in translation theory. These translators then specialises as they already have in-depth knowledge of the specific terminology."

Van Rensburg and Nortjé agree that extensive experience and continued training are vital. According to Nortjé he appoints translators by virtue of their qualification as well as proven track record of experience (other than being a mother tongue speaker of the target language). Van Rensburg proposes that a mentor should monitor the work of a new translator for at least a year.