A small company DeepL was able to surpass Google, Microsoft and Facebook in the field of machine translation - with equal speed, the program achieves a more accurate result while preserving the shades of meaning.

According to journalists Tech Crunch, who conducted the tests of an automated translator, DeepL copes with the task better than Google Translate and Bing. "While Google Translate often chooses very literal values ​​that do not take into account some of the nuances and idioms (or fundamentally misunderstand their meaning), DeepL usually offers a more natural option that is closer to how a professional translator could do it, - says the article. And if you speak English, French, German, Spanish, Italian, Dutch or Polish, you can assess the possibilities of DeepL yourself. Russian language yet.

The predecessor of DeepL was the Linguee project, which existed for several years, but never achieved the results of Google. His co-founder, Gerone Fraling, used to work for Google Research. His team had many years of experience in machine translation, but only a year ago they began to seriously develop an entirely new system and company, which was called DeepL. "We built a neural network for translations, which absorbs most of our developments, to which we added new ideas," says Fraling.

A huge database of more than a billion translations and queries plus a method of searching for similar semantic elements in the network has become a solid basis for learning a new model. And the computing power is provided by the supercomputer located in Iceland, the 23rd in the world rating.

hightech.fm 

© All rights reserved. Citing to www.ict.az is necessary upon using news