DeepL: Google Translator’s newest rival
Although it may appear otherwise in the eyes of many users, the truth is that Google Translator has been obtaining the best results in many different types of text within the field of machine translation. Regardless of the technology employed, to date, nobody has been able to compete with the linguistic corpus managed by Google, one of the cornerstones of statistical machine translation. Without a doubt, Google is at the forefront in this field.
DeepL: the David of machine translation
Since the emergence of a new mathematical model based on neural networks which can be applied to the field of machine translation, Google Translator’s results have improved significantly. However, the new paradigm has made it possible for the company DeepL to introduce its machine translator
to the market this past August. Developed in the past year, it offers spectacular results that often improve those produced by Google, according to information from the company itself
. For more information, please refer to the following article published in Tech Crunch: DeepL schools other online translators with clever machine learning.
What does seem to be clear is that now users of Google Translator will have a second alternative when they don’t achieve satisfactory results.
Linguee: David’s predecessor
DeepL is the new strategic name of the company that for years now has been the largest translation search engine on the Internet, Linguee
, which is widely recognised and used among professional translators. Linguee is able to align translations found on the Internet, applying an algorithm to evaluate the translation and to offer the translator information on the source from which the translation was taken. So, while DeepL’s machine translator is still in its initial stages, the company behind its creation has been developing linguistic solutions for over a decade.
How is it possible for DeepL to offer such good results?
This question can lead to speculation, taking into account that, for strategic reasons, the company has not commented on this matter when asked. However, it is safe to say that there are at least two variables that have enabled this surprising progress:
Quality over quantity
One of the parameters that plays an important role in the results obtained by a machine translation system is the quality of the translations introduced into the system. In this sense, it is worth pointing out that one of the reasons why Linguee is so successful is precisely the fact that it offers texts that have achieved a minimum score in the algorithm developed by DeepL to evaluate the quality of the translations. In other words, compared to the large quantity of data used by Google Translator, DeepL has applied a prior filter to the translations introduced into its machine translation tool.
Application of the new paradigm of neural networks
The emergence of this new paradigm applied to machine translation is relatively recent. Therefore, it is possible that DeepL’s programmers have been able to develop a tool that, to date, is able to efficiently use the virtues of the learning paradigm based on neural networks.
To find out more about the concept of neural networks applied to translation, please see the following blog entry by Lionbridge, the world’s largest professional translation and localization company: Neural Machine Translation: How Artificial Intelligence Works When Translating Language