BUSINESS GUIDE DISTRIBUTED WITH


12TH FEBRUARY 2024


5


How do machines


translate language?


Creative Translation is a global translation agency providing transcreation, cultural consultancy, typesetting, voice-over, video editing and market research


Believe it or not, AI has been used in translation for many years. Te earliest form of AI translation was NMT (neural machine translation), which is the foundation on which the Chat GPTs and LLMs are built today.


So how do they do it? First, algorithms analyse the grammar, vocabulary and structure, and then neural networks recog- nise patterns — sifting through huge amounts of language that have been previously catalogued. (Tis is similar to how we learn language as children — listening to people talking around us.) Te system then works out — using


probability calculations — how to express the same meaning in the new language. Te trouble is, languages rarely have one-to-one translations, and the context, or a cultural glitch, can change the meaning. All of this happens in a flash and the translated text is produced.


AI is always learning And the more AI absorbs, the more it learns, the more it can reference. Tis is especially true if the user corrects any translations. Machine learning has become


remarkably good, especially for straightforward text; it’s when the message needs to connect emotion- ally that a more creative approach is called for. And that’s when transcreation


steps in. Wherever text needs an emotional


interpretation, a living, breathing human being must be there to refine the meaning. Plus, there are phrases and expres-


sions (idioms) that are unique to every country. Te


very English ‘Everyone


makes mistakes’ translates as ‘Even monkeys fall from trees’ in Japanese. ‘Raining cats and dogs’ in English is ‘raining ropes’ in French. But what machine would, or could,


know that? Visit: creativetranslation.com


T: 020 7138 3810 E: info@creativetranslation.com


Creative Translation 111 Power Road, London W4 5PY


Tese idiomatic differences still


present a challenge for machines, even though they are improving.


Two different systems Currently there are two technolo- gies leading the charge in transla- tion. Te tried-and-trusted NMTs deliver accuracy and consistent terminology,


while LLMs can


make a machine translation sound more human. But LLMs can’t be trusted, they’re


only as good as the input they’ve been given to index. Tey can even ‘hallucinate’, creating random content from any unfiltered internet ramblings that have been stored.


Te trick is to get the balance right, using each technology to its strengths — and have an experienced pair of eyes overseeing things


Human being required In short, NMTs are great for accuracy, while LLMs are good at sounding natural.


Te trick is to get the balance


right, blending new and old, using each technology to its strengths — and have an experienced pair of human eyes overseeing things. AI is perfect at speeding up the


translation process, but it needs a flesh and blood, coffee-drinking translator to make sure the work is pin sharp. And expert transla- tors and gifted writers are exactly who you’ll find working at Creative Translation (CT).


So, why choose CT? Global transcreation agency Crea- tive Translation has been at the fore- front of translation for 20 years, and


is a leader in exploring and adopting new technologies. Tat’s why it can offer a range of


proven, integrated language services to both agencies and brands, for every industry category. All this is supported by a special-


ised, in-market talent pool of linguists. If you need anything to do with translation, perhaps you should have a word with Creative Translation?


CT HAS THE MACHINES. AND


THE PEOPLE Creative Translation has over 3,000 in-market translators, editors and cultural consultants, capable of adapting campaigns in every language from Azerbaijani to Zulu. Global marketing has changed. Let’s do this. Together.


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24