Recently a lot has been said about this concept: neural translation. Translating documents from one language into another is not an easy task. Most translation courses usually last 4 years or more. This gives students time to gain in-depth knowledge of a language and translation techniques, and at the same time absorb the culture of the languages they are studying and acquire essential skills, such as writing, summarising and paraphrasing sentences, words and complete paragraphs.
It's no secret that technology is advancing rapidly, and every new development is more surprising than the last. As with other sectors, technological advancements also affect the world of translation.
Today's translators are now debating whether it is possible to develop a system capable of translating as concisely and as accurately as a professional translator. While some think that a machine will never be able to attain the quality of a human translation, others disagree.
This is where neural machine translation comes in. What is it?
It is a translation system created via a large artificial neural network that imitates human learning, currently known as ‘machine learning’. According to Gereon Frahling, former employee of Google and Managing Director of DeepL: “This service, as with other similar technologies, finds its growth engine in “machine learning” based on artificial neural networks. Thus, this recent revolution, applied in the field of robotics, together with “deep learning” models, enables artificial intelligence to progress quickly and "learn", according to the values entered".
What does this mean? That, based on previous translations and trial and error, the machine evolves and increases its code, permitting the same rules to be applied to future translations.
This means that, with specific and extensive learning, the system could replace human translators. However, Frahling also states: “We expect this to happen within a few years, but we can't be sure yet. Sometimes, however, the context is not enough: professional translators translate according to the client's requirements, the public to which the translation is intended, the field of the speciality, the purpose of the translation or a product, a specific register, the age of the target audience, etc. These are details that a neural network cannot take into account or, for now, overcome.
What about you? What do you think? Do you think that in the future we will resort to machine translation systems to translate our documents? In the meantime, if you need any type of translation, why not have a look at the different services that AT Language Solutions offers?