in new quality
DeepL: Automatic translation in new quality
When I compare the results of DeepL with those of Google Translate, I come to the conclusion that a quick translation of blog posts, landing pages etc. into other languages will be much easier in the future. I’m impressed.
DeepL is the work of the creators of Linguee, which has been a useful tool on the web so far, when it came to producing good-sounding sentences in English as a non-native. This is because it is not only possible to look up the translations of individual terms there, but also to display the searched translations in context.
And it is precisely the context that has been the central weakness of automatic translations so far. Anyone who occasionally uses Google Translate to translate entire passages of text knows that the results are usually not particularly useful. The grammar is usually not even close to correct, not to mention subtleties such as expression, melody and metrics.
This will probably not change with DeepL either. But a first, quick test is promising: The new service could make it easier to translate blog posts, news articles or landing pages into other languages within a very short time.
DeepL, the translation service, is confident enough to announce on its own website:
100 sentences have been translated by DeepL translators, Google, Microsoft and Facebook. Professional translators assessed these translations without knowing which system they came from. DeepL’s translations were chosen three times as often as those of the other systems.
The English version of this post has been translated 100% machine translated
I’m doing the test right now and offer readers the opportunity to judge for themselves: The English version of this post, which can be accessed by clicking on the flag at the top, was created 100% by DeepL and was not revised by me afterwards. I haven’t edited a single character of it manually. Therefore, I apologize to English-speaking readers, if not all the phrases sound perfect. I am looking forward to receiving comments on the quality of the translation.
By the way, heise. de asked DeepL which technologies are behind the new translation service: Heise Online reports that Convolutional Networks would be used, as they are common for image recognition. “They process all the words in parallel, and there are also highly optimized libraries for the calculation.”
I agree wholeheartedly with you, Bernhard. I have tested DeepL on a difficult philosophical article and on the text of a children’s book. (E->F in both cases). I find DeepL absolutely tremendous. It doesn’t think, of course, but that’s not really the point. Its output is way superior to Google’s or Bing’s and even to the translation of the published children’s book.
I am a professional translator (retired from the UN after a little over a quarter-century in the salt mines :).