You are here

The power of machine translation

By Jean-Claude Elias - Nov 19,2020 - Last updated at Nov 19,2020

Photo courtesy of intonation.com

Putting to good use the power of computers, artificial intelligence (AI), neural networks and the various digital tools available, to do language translation, has reached a milestone. Gone are the days where completely manual, traditional translation was done.

This is true at least for contents like business, news, websites, technical documentation of all kinds, user manuals, brochures, emailing, online chats and the like. Such contents probably constitute the bulk of all text generated these days. As for literature like novels and poetry, where specific style is of prime importance, manual translation is still irreplaceable at this point in time.

Making the computer translate is the way to go today. The simplest, cheapest method is through systems like for instance Google Translate. The limitations, shortcomings and the weaknesses are well known, and the result in some extreme cases is just a laughable matter. Yet, it often helps to communicate quickly and give a reasonable idea of the original material. Best of all it is a huge time and money saver.

Better than the above are DeepL and Linguee. Though not perfect they use AI to analyse the context and produce a translation of good-enough to real-good quality. DeepL does direct translation whereas Linguee provides a list of already translated examples, in the source and the target language, leaving you the choice of selecting and deciding what seems the best to you.

DeepL uses what is referred to as Machine Translation (MT), whereas Linguee taps into what the company calls the largest available database of Human Translation (HT) already done. Both acronyms, MT and HT, are now part of the field. Both are software products by the German company DeepL and are available free online.

Professional translators have even more powerful MT tools. They use software that relies on a concept called Translation Memories (TM) along with glossaries that pertain to the very subject they are translating: medicine, architecture, music, business, tourism, software and so forth. This dramatically increases the precision and the quality of the output. In the best cases the result is “almost” as good as pure HT.

Naturally, these professional tools are not free. SDL Trados is one of the suppliers of paid software that blends all the digital means available: MT, TM, AI and specialised glossaries, to help the professional linguists generate perfectly translated text, by leaving to them to put the final touch to the work, in a personalised, customised manner.

By combining the digital tools while still letting the translator decide what is the best text in the end, SDL Trados allows for very fast yet accurate translations. By tapping on machines and memory resources, the product avoids wasting time translating what others have translated before. In a certain way it is about not reinventing the wheel.

Another method of obtaining perfectly translated text used in the professional field is the MTPE, which stands for Machine Translation Post Editing. It consists of starting with a completely computerised translation, a huge time saver since it produces a virtually instant result, at least when done with fast computers, and then giving it to a specialised linguist to review, edit and correct. Here too the result is usually excellent and there is invaluable time saved.

Whether with the free online tools like DeepL or Google Translate, or with paid software like SDL Trados, the world of language translation has undergone a real revolution over the past 10 years. The whole concept of getting help from the machines, their memories, AI, and the networks, is referred to as CAD (Computer Aided Translation).

CAD smartest aspect perhaps is that it lets you balance how much you are willing to accept as purely automated translation, how much human input you want to involve in the end, and how much you are ready to pay, so as to reach the quality of translation you are targeting.

CAD is more important than ever, given the flabbergasting amount of text generated every day, and the need to translate it quickly into the main languages spoken in the world.

up
66 users have voted.


Newsletter

Get top stories and blog posts emailed to you each day.

PDF