Deepl vs Google translate: A modern day David and Goliath

In only 6 months DeepL has become indispensable for aspiring translators and is now teaching giants like Google Translate a lesson. It must be said that free and powerful online translation tools are few and far between. This means that as soon as a player with potential enters the online translation market, observers scrutinize its each and every move. In this case, the player is located in Cologne, Germany, and is run by about only 20 employees.

The community in the service of online translation

DeepL is a free translation tool from DeepL GmbH company which uses the famous multilingual dictionary database, Linguee. Consequently, the real power behind Deep’l is Linguee. First and foremost, DeepL’s Artificial Intelligence is “learning” thanks to the millions of texts and translations posted on Linguee. DeepL is an abbreviation for “Deep Learning”. However, one would expect Google AI to learn even more. Let’s not forget that
the keyword “Google translate” is searched 45,500,000 times a month. Furthermore, DeepL AI and Google Translate AI both use neural network technology, and yet DeepL AI
is, nevertheless, better. How does DeepL work better when using a similar technology? The answer is simple: the performance of a neural network depends on the quality of its learning 1 materials and not just on quantity. DeepL’s neural networks evolve 2 from billions of high-quality translation segments from the Linguee database. Its main contributors are reliable sources: the European Parliament, Unesco patents and literary works.

15 times fewer languages available, but 3 times more effective than Google translate

Google Translate is a leader in automated translation. 1 While DeepL only translates 7 languages, the Google tool is able to translate 103 languages. Although DeepL’s technical director, Jaroslaw Kutylowski, stated in August his “willingness to take things further in a few months”…. nine months later: nothing has changed.
In order to realise its ambitions, the company, funded largely by advertising, has built in one of the world’s best supercomputers in Iceland. And why in this country? Because Iceland uses renewable energies on a massive scale, which allows DeepL to manage its neural networks in highly beneficial way. Moreover, this supercomputer is actually the world’s 23rd most powerful supercomputer, with a capacity to perform more than 5,100,000,000,000,000,000,000,000 floating point operations per second. As a result, its computing power translates 1 million words in less than one second.
Therefore, the breathtaking prowess of its supercomputer, combined with the quality and quantity of its learning material in Artificial Intelligence, allows DeepL to be the most efficient on the market. The most wondrous thing is that DeepL proves it: its online translation service has gained results 3 times better than Google Translate on
the BLUE test. BLUE stands for Bilingual Evaluation Understudy, a standard quality evaluation test in the field of automated translation. It compares computer-translated texts with those of a human translator. In addition, during a blind test carried out by the company, DeepL was chosen far more often than its competitors by professional translators.
Translated by Margaret Ubels-Rattray
From :

Leave a Reply

Bouton retour vers haut de page